Authors: Sam Earle, Kay Arulkumaran, Andrew Dai, Akarsh Kumar, Julian Togelius, Sebastian Risi
Abstract: We are in the midst of large-scale industrial and academic efforts to automate the processes of scientific, technological and creative production through AI-driven assistants. Historically, a fundamental property of these processes in their human form has been their open-endedness: their capacity for generating a seemingly endless supply of novel and meaningful new forms. Do artificial agents have any capacity for such fruitful unguided discovery? To answer this question, we turn to Picbreeder, the canonical exemplar of human-driven open-ended search, in which users collaboratively generated a diverse library of images through interactive evolution of small neural networks. We replicate Picbreeder, replacing human users with frontier Vision Language Models (VLMs). We observe clear qualitative differences between the output of our system and the historical human baseline, and attempt to characterize them using metrics of phylogenetic complexity and visual and semantic salience and novelty. In an effort to identify some of the causal factors contributing these differences, we study the addition of exploratory noise to the agents' selection process, of behavioral diversity between agents, and of narrative momentum in the form of memory of past actions. We make our code available at https://github.com/smearle/picbreeder-vlm.
Authors: Noam Michael, Daniel BenShushan, Jacob Bien, Don A. Moore
Abstract: We investigate the calibration of large language models' (LLMs') confidence across diverse tasks. The results of our preregistered study show that the current crop of LLMs are, like people, too sure they are right: confidence exceeds accuracy, on average. Importantly, however, this tendency is moderated by a powerful hard-easy effect, wherein overconfidence is greatest on difficult tests; by contrast, easy tests actually show substantial underconfidence. We develop LifeEval, a test for evaluating model calibration across levels of difficulty.
Authors: Zhiyuan Zhai, Xinkai You, Wenjing Yan, Xin Wang
Abstract: Reasoning-capable large language models solve hard problems by emitting long chains of thought, paying heavily in latency, GPU time, and energy. Casual inspection of their traces reveals extensive reformulation, verification, and circular self-reflection, yet how much of this deliberation is actually necessary has never been measured at scale or explained from first principles. This paper closes both gaps. We formalise reasoning redundancy directly in terms of the reasoning model itself: the redundancy of a correct trace is the largest fraction of its trailing segmented steps that can be truncated while $\pi$, forced to terminate thinking and emit a final answer, still produces the correct answer. A large-scale quantification across four frontier reasoning models and two mathematical benchmarks shows that step-level redundancy is consistently high -- between 61% and 93% across the 8 (model, benchmark) conditions we study, with the median critical prefix equal to a single segmented step in six of the eight conditions -- that the finding is robust to the choice of judge family, and that although $\rho$ decreases with problem difficulty on MATH-500, all four models remain substantially redundant ($\rho \in [46\%, 85\%]$) even on the hardest Level-5 problems. We then prove that this redundancy is a structural consequence of length-agnostic outcome rewards, not a model-specific artefact: under any such reward, no finite expected stopping time is optimal. The result holds regardless of RL algorithm, base model, data distribution, or whether the policy is obtained via RL or distillation; over-thinking is therefore not a bug to be patched in individual models but a structural property of how current reasoning models are trained. Code: https://github.com/zhiyuanZhai20/how-much-thinking-is-enough
URLs: https://github.com/zhiyuanZhai20/how-much-thinking-is-enough
Authors: Gregory Magarshak
Abstract: We present Context, the intelligence layer of the Magarshak Architecture, which replaces reactive query-response chatbots with proactive goal-directed agents that advance shared tasks without waiting for user prompts. The architecture rests on three mutually reinforcing mechanisms. Write-time context assembly precomputes enriched typed attributes via Groker agents, assembling interaction context as a deterministic pure function of graph state; context blocks are byte-identical across turns between semantic changes, enabling near-100% KV-cache reuse. Composable sandboxed wisdom programs form a governed library of LM-generated imperative programs declaratively wired to goal types via typed stream relations, composed via phase ordering, and executed at interaction time without further LM calls. Proactive goal stream state machines drive conversations toward terminal states by inspecting graph state and emitting structured interaction content (option arrays, governance affordances, clarification prompts) without awaiting user input. We prove six formal results: the Context Stability Theorem, bounding per-turn LM cost as a function of semantic change rate; a Program Composition Correctness Theorem; a Declarative Wiring Soundness Theorem; the Proactive Dominance Theorem, proving proactive agents weakly dominate reactive agents on expected turns-to-terminal-state; Coordination Overhead Elimination and Quality Preservation, establishing Pareto improvements in multi-participant goal chats; and a Cross-Platform Vote Consistency Theorem. Implemented in the open-source Qbix / Safebox / Safebots stack.
Authors: Ya-Ting Yang, Quanyan Zhu
Abstract: Modern AI systems increasingly rely on workflows composed of multiple interacting agents, some powered by large language models (LLMs) and others by conventional computational modules. This paper analyzes the fundamental tradeoffs between latency, reliability, and cost in LLM-enabled agentic workflows. We introduce performance models for both LLM and non-LLM agents that capture the relationship between computational effort and output quality, incorporating the impact of reasoning and output tokens for LLM agents using a parametric exponential reliability function. Then, we study the design of sequential workflows under latency and cost constraints. Main results include a water-filling token allocation policy and characterizations of optimal workflow reliability in terms of shadow prices.
Authors: Saad Mankarious
Abstract: We introduce \emph{Quantum Frog}, a two-player cooperative game built on a novel \emph{quantized-time} mechanic in which the environment advances only when a player acts. Inspired by the classic arcade game Frogger, Quantum Frog requires two frogs to cross an 8$\times$8 grid of traffic and reach the far side together. We use reinforcement learning (RL) as an analytical lens to answer four design questions: (1) how does game difficulty scale with traffic density, (2) what is the optimal single-agent policy and why, (3) how large is the cooperation gap between independent and cooperative two-agent play, and (4) what joint strategy emerges when agents are incentivised to cooperate? We train agents through five escalating stages, Tabular Q-Learning, Deep Q-Network (\DQN), Independent \DQN~(\IDQN), and Multi-Agent Proximal Policy Optimisation (\MAPPO\ with a centralised critic), evaluating each against traffic densities of one to six cars. Our key findings are: (i) the quantized-time mechanic makes a \emph{rush strategy} (moving directly upward at every step) universally optimal, as time exposure to traffic is minimised; (ii) adding an uncoordinated second player is harder than sextupling the traffic for a single expert player; (iii) cooperative training recovers +32--34 percentage points of joint success rate relative to independent agents and reduces episode length from $\sim$90 to $\sim$6 steps; and (iv) the emergent cooperative strategy is synchronised rushing, not complex positional coordination, illustrating that shared incentives alone suffice to align agents in time-critical cooperative tasks. These findings provide concrete, empirically grounded guidance for the commercial design of Quantum Frog and offer broader insights into the role of environment mechanics in shaping multi-agent learning dynamics.
Authors: Zhiming Chang, Ziyang Li
Abstract: The formal verification of operating system kernels requires precise specifications that capture the intended behavior of system calls. Writing these specifications manually demands deep domain expertise, motivating the use of large language models (LLMs) to automate the process. However, in OSV-Bench, a benchmark of 245 specification generation tasks derived from the Hyperkernel OS kernel, the best reported Pass@1 is 55.10%. We propose a domain knowledge prompting method (BODHI), which augments the standard few-shot prompt with a structured C-to-Python translation guide covering 15 categories of domain-specific translation patterns. Inspired by Structured Chain-of-Thought (SCoT) prompting, the guide organizes translation by separation of concerns, addressing pre-condition extraction and post-condition generation as distinct categories. Evaluated on nine models from six providers (Anthropic, Mistral, Amazon, DeepSeek, Meta, Alibaba), covering dense, mixture-of-experts and reasoning architectures, BODHI improves every model tested, with gains ranging from +11% to +32%. The best configuration (Claude Opus 4.6 + BODHI) reaches 96.73% Pass@1. BODHI reduces both syntax and semantic errors, with the strongest effect on models that have sufficient instruction-following capability to utilize structured reference material. These results demonstrate that domain knowledge injection is a model-agnostic technique that substantially bridges the gap between general-purpose code generation and formal specification synthesis.
Authors: Boyu Xiao, Xiuqi Tian, Xuwen Song, Haochun Wang, Guanchun Song, Sendong Zhao, Bing Qin
Abstract: Despite strong medical benchmark accuracy, LLMs can exhibit severe multi-turn sycophancy in clinical dialogue, abandoning initial correct diagnosis under escalating pressure. We propose \textbf{\textsc{Med-Stress}}, a targeted stress test framework that evaluates belief stability under escalating pressure. Across nine frontier large language models (LLMs), we find a clear dissociation between medical knowledge and robustness: high initial diagnostic capability does not imply high belief stability, yielding large knowledge-robustness gaps for several LLMs. To mitigate this failure mode, we propose a lightweight inference-time defense, \textbf{\texttt{RBED}} (\textbf{R}ole-\textbf{B}ased \textbf{E}pistemic \textbf{D}efense), and \textbf{\texttt{R-FT}} (\textbf{R}esilience-oriented \textbf{F}ine-\textbf{T}uning), a training-time approach that internalizes evidence-based resistance to pressure. Experiments show that \textbf{\texttt{R-FT}} nearly eliminates belief change and substantially improves robustness.
Authors: Wang Rui, Lu Diannan
Abstract: Quantum computing devices are recognized as powerful tools for solving NP-complete problems. However, the intricacy of their modeling presents notable barriers for non-specialists, while the tedious iteration of constraint weights and modeling methodologies also consumes substantial effort on the part of experts. To address these challenges, this study integrates a femtosecond laser-pumped Coherent Ising Machine (CIM) with an LLM-driven agentic system by leveraging the LangGraph and LangChain frameworks. Comprehensive investigations demonstrate that large language models (LLMs) can effectively perform such tasks in modeling as QUBO/Ising model calibration, constraint weight decision iteration and rapid validation of literature-reported schemes. Notably, all these tasks can be fully implemented based on domestic large models, combined with domestically developed CIM hardware, we truly achieve the practical empowerment of quantum CIM that fully relies on all-domestic agentic large models and hardware. This work successfully realizes robust technological integration, laying a solid foundation for subsequent research. Nevertheless, it also identifies the persisting challenges in the two cutting-edge fields of large models and quantum computing at the current stage. Encouragingly, we unexpectedly discover a promising new paradigm where accumulated knowledge from agent-assisted quantum computing iterations reciprocally enhances the agent's own problem-solving capability, thereby addressing these challenges.
Authors: Marcelo Fernandez - TraslaIA
Abstract: Autonomous agent systems fail not only due to incorrect decisions, but due to executing decisions whose authority no longer holds at runtime. Prior work defined Reconstructive Authority (RAM) as a condition for valid execution: actions are permitted only if authority can be constructed from current state. This paper addresses enforcement at runtime: how to enforce this condition in a running system. We introduce a runtime execution model in which authority is evaluated at action time and execution is conditioned on its constructibility. This extends the execution state space beyond admit/deny with a third state, halt, representing cases where authority is undefined due to incomplete or uncertain observability. We define a concrete execution protocol including dynamic dependency resolution, authority reconstruction, and explicit decision semantics. We further introduce a Recovery Loop that integrates drift detection (IML) with execution control (ACP), allowing the system to suspend execution, acquire missing information, and re-attempt authority reconstruction. We show that this model guarantees safety -- no action is executed without constructible authority -- and conditional liveness: execution resumes when authority-defining variables become observable. This work operationalizes reconstructive authority as a runtime enforcement mechanism, providing the execution semantics required to apply RAM in real systems.
Authors: Takaaki Fujita, Florentin Smarandache
Abstract: This book presents a comprehensive and systematic survey of graph theory under uncertainty, with particular emphasis on the unifying role of the uncertain graph framework. It reviews fundamental concepts, structural properties, graph classes, and graph parameters within fuzzy, neutrosophic, and related models, while also introducing a wide range of extensions such as uncertain digraphs, hypergraphs, superhypergraphs, and dynamic graphs. In addition to theoretical developments, the book explores practical applications, including uncertain molecular graphs, decision-making systems, graph neural networks, knowledge graphs, and cognitive maps. By organizing diverse uncertainty-aware graph models within a common perspective, this work provides a coherent framework for understanding their relationships, capabilities, and applications in complex systems.
Authors: Bruno F. Louren\c{c}o, Hesham Morgan, Ana Ozaki, Aleksandar Pavlovi\'c, Emanuel Sallinger
Abstract: Knowledge base (KB) embeddings aim at combining the capability of classical knowledge graph embeddings to generalize the information present in facts, the ABox, with conceptual knowledge represented in an ontology language, the TBox. Several authors have recently explored the idea of mapping concepts to convex regions in a vector space. This is useful to represent hierarchies, typically present in TBoxes, since more general concepts can be mapped to larger regions, containing those regions associated with more specific concepts. However, the power of convexity is rarely leveraged during the actual learning tasks. Here, we introduce BoxLitE, a KB embedding model for DL-Lite$^{\mathcal{H}}$ that allows for convex optimization. We show that for any satisfiable DL-Lite$^{\mathcal{H}}$ KB, there is a BoxLitE embedding that is a weakly faithful model. As a proof of concept, we show how to formulate the KB embedding task as a convex optimization problem and how to obtain embeddings with such desirable faithfulness properties.
Authors: Long Zhang, Zi-bo Qin, Wei-neng Chen
Abstract: Large language models (LLMs) increasingly fuse heterogeneous inputs in ubiquitous systems. Yet, how LLMs implicitly allocate authority when sensor measurements and user claims conflict remains unexamined, raising critical reliability concerns for deployments where physical sensing must retain priority. Unlike explicit traditional fusion, LLMs bury authority allocation within learned representations. We discover this allocation is severely format-dependent: numerical sensor data fails to integrate into answer-relevant model directions, allowing natural-language claims to dominate the final decision, a phenomenon we term \textbf{Authority Inversion}.To diagnose and mitigate this, we develop a geometric framework of context integration, introduce two computable audit metrics, specifically the Context Integration Ratio (CIR) and Authority Alignment Index (AAI), and propose Geometric Authority Calibration (GAC), an inference-time layer-level intervention to suppress misplaced user authority. Evaluating four models (4B to 35B parameters, three architectures) across four datasets totaling 576 conflict instances reveals extreme inversion: on numerical tasks, models exhibit near-zero sensor trust (AAI = -0.805, Cohen's d = -2.14), unaffected by model capacity. Validating our geometric framework, theory-guided causal injection flips 80.2\% of incorrect decisions (vs. <0.4\% for random controls). Practically, GAC improves HAR accuracy from 0 -- 1.6\% to 21.9 -- 27.5\%, outperforming prompting baselines. Ultimately, authority allocation in LLM-mediated systems must be explicitly audited and application-specifically configured rather than left implicit.
Authors: Xirui Liu, Sihang Zhou, Yanning Hou, Rong Zhou, Haoyuan Chen, Maolin He, Siwei Wang, Hao Chen, Jian Huang
Abstract: Web agents require both high-level reasoning (for task decomposition) and low-level interactions (for page elements manipulation) to conduct different tasks. However, these knowledge types differ fundamentally: reasoning knowledge (e.g., booking a flight requires first searching for routes) is abstract and transferable across websites, while interaction knowledge (e.g., clicking the Search button at a specific coordinate on Site A) depends heavily on page-specific contexts. Existing methods store experiences uniformly. This creates a dilemma: abstract representations lose executability on concrete pages, while concrete representations fail to generalize across domains. This entanglement limits capability accumulation: on new websites, agents either fail to recognize reusable task logic due to surface-level differences or attempt infeasible actions from outdated page structures. To disentangle them, we propose DRIVE, a dual-level skill modeling framework separating historical experience into natural language reasoning skills, which capture transferable task logic, and programmatic interaction skills, grounding abstract actions to executable operations. A scene-aware coordination mechanism adaptively retrieves and invokes these dual-level skills based on task semantics. DRIVE also uses skill-level reflection to identify hierarchy-specific failure modes, enabling targeted skill library expansion and refinement. Experiments across five WebArena domains show DRIVE attains an average task success rate of 52.8%, exceeding the skill-free baseline by 7.3 percentage points. Further ablations show reasoning and interaction skills provide distinct, complementary benefits, supporting separation of transferable task logic from executable page-level operations.
Authors: Sebastien Kawada
Abstract: How do multi-turn reasoning systems fail? The expected answer is logical contradiction, in which the system's maintained state becomes unsatisfiable. We show that the dominant mode is instead satisfiable drift, where the internal state stays consistent while the returned answer silently violates prior commitments. We build DRIFT-Bench (Decomposing Reasoning Into Failure Types), a solver-instrumented benchmark of 816 test problems across three constraint domains, and evaluate four methods on it across four open-weight models (8B-120B parameters). MUS-Repair, which feeds minimal unsatisfiable subsets back to the generator, is strongest in every setting (+1.8 to +15.0 pp over the best non-MUS baseline). But the central finding is what repair leaves behind. After structured feedback, models rarely contradict themselves. They forget. Residual errors are 98-100% satisfiable drift across all settings, while contradiction drops to near zero. Reliable multi-turn systems must separately validate that the returned answer respects the maintained state. Code is available at https://github.com/kaons-research/drift-bench.
Authors: Maissa Abir Smaili, Eren Sadikoglu, Ransalu Senanayake
Abstract: Alzheimer's disease is a neurodegenerative disorder marked by progressive declines in memory and language that reduce independence in daily life, motivating socially assistive robotic support. This paper presents MEMOR-E, a mobile quadruped robot with an interactive tablet interface that assists patients and caregivers through medication reminders, routine guidance, memory oriented interactions, and companionship. We evaluated the feasibility of fine tuning large language models (LLMs) to emulate stage consistent cognitive behavior and interpret responses across standard neuropsychological language tasks, using audio transcriptions from 235 Alzheimer's patients and synthetically generated healthy controls. We also report findings on using in context learning (ICL) in LLMs, where a second LLM produced domain and severity level cognitive error summaries. Our results show that MEMOR-E can generate stage aware, non diagnostic cognitive summaries that support personalized assistive interactions, while explainable AI mechanisms translate model outputs into transparent, human readable evidence to enable caregiver oversight and trustworthy human robot interaction.
Authors: Carlo Cattani, Dioneia Motta Monte-Serrat
Abstract: This paper proposes a structural and dynamical framework for modeling cognitive processes within a cybernetic perspective. Cognitive states are represented as elements of a state space evolving through an iterative update rule of the form \[ X_{t+1} = \pi\big(F(f(X_t))\big), \] where $f$ describes internal transformations, $F$ represents interpretative mappings, and $\pi$ enforces semantic equivalence. The model is interpreted as a feedback system integrating transformation, observation, and stabilization. A categorical formulation is introduced to capture compositional structure, while the associated dynamics are analyzed through fixed-point arguments and contraction conditions ensuring stability. To demonstrate the operational character of the framework, a computational illustration is provided, together with a qualitative analysis of the induced dynamics. A concrete linguistic application shows how context-dependent interpretation can be modeled as a trajectory toward a stable semantic class. The proposed approach connects dynamical systems, category theory, and cognitive modeling, and provides a unified representation of cognition as a feedback-driven process evolving toward invariant interpretations.
Authors: Song-Ju Kim
Abstract: Quantum cognition often explains order effects, contextuality, and violations of the law of total probability by replacing classical probability with quantum probability on a fixed event structure. This paper proposes a different interpretation: quantum probability is the fixed-spacetime projection of contextual spacetime formation under finite-state requirements. The framework begins not with time, space, objects, or probabilities, but with requirements such as finite representational capacity, single-state semantic stability, context-sensitive intervention, avoidance of explicit context labels, coherent world-formation, and intersubjective transformability. When these requirements cannot be realized within a single global Boolean event structure, the mismatch appears, under fixed-spacetime projection, as noncommutativity, interference, and quantum-like probability. Building on prior single-state approaches to contextuality, we reinterpret classical contextual bookkeeping cost as the fixed-spacetime shadow of contextual spacetime formation. Auxiliary memory or context labels in a classical representation correspond, in this account, to holonomy-like mismatch among locally Boolean logic-worlds. The interference term is the cross term generated when locally classical realization contributions are nontrivially glued and projected back into a fixed classical spacetime form. The result is a transcendental-operational realist account: objecthood, eventhood, probability, and spacetime are treated as forms of realization under requirements, while objectivity is defined by invariants preserved across observer- and history-dependent spacetime formations.
Authors: Jing Dong, Prakirt Raj Jhunjhunwala, Yash Kanoria
Abstract: We model the interaction between a user and an AI driven recommendation system. The user initiates the process by conveying preference information through a costly and noisy message. The AI assistant, acting as a Bayesian agent, interprets the user's message to form a posterior belief about their true preferences and make product recommendations. In particular, it determines how many recommendations to present so as to maximize the user's expected utility from their final choice, while accounting for the search cost induced by the size of the recommendation set. We use mutual information based cost functions to model the two distinct costs incurred by the user during the interaction: (i) a communication cost, which increases with the precision of their preference message, and (ii) a search cost, which increases with the size of the recommendation set provided by the AI assistant. We study products and preferences which live in d dimensional space, and ask how the user's expected payoff can be maximized. For large d, we characterize how optimal message precision and recommendation set size depend on the cost parameters, under two distinct distributions from which recommendations can be sampled from the product universe: (i) Bayes' posterior belief, and (ii) an optimized tilted distribution. Under the posterior sampling scheme (i), we identify a hybrid regime, in which an efficient interaction policy requires jointly optimizing the amount of information (in bits) conveyed by the user and the number of recommendations provided by the AI assistant. In the tilted sampling scheme (ii), our results show that the optimal interaction policy uses only one of communication and search, favoring whichever of them is less costly.
Authors: Long Zhao, Qinghe Wang, Jiaan Zhu, Youhui Bai, Zewen Jin, Chaoyi Ruan, Shengnan Wang, Cheng Li
Abstract: Reinforcement Learning from Human Feedback (RLHF) has become a key post-training paradigm for improving model quality. However, the synchronous three-stage RLHF pipeline is often bottlenecked by the generation stage, where response-length skew causes the effective batch size to shrink rapidly during decoding, leaving GPUs underutilized while a few long responses remain unfinished. Mainstream frameworks employ a static tensor parallelism (TP) configuration that cannot adapt to changing batch characteristics, leaving substantial performance headroom unexplored. We propose PAT, an adaptive TP method that dynamically reconfigures TP during the generation stage of each RLHF iteration. PAT introduces two key techniques. First, a predictor-guided online reconfiguration method decides both the reconfiguration point and the target TP configuration based on offline profiling, triggering reconfiguration only when the predicted latency benefit outweighs the reconfiguration overhead. Second, a lightweight online reconfiguration mechanism updates only the states and layouts affected by TP changes: it adapts unfinished decoding states through a cost-model-based choice between KV-cache migration and recomputation, performs in-place weight resharding, and reuses cached communication groups. We implement PAT on top of SGLang and integrate it with the VeRL framework. Evaluations on LLaMA3.1-8B and Qwen3-14B using DeepScaleR show that PAT reduces generation latency by up to 34.6% and end-to-end RLHF training iteration latency by up to 27.2% compared to the original VeRL setup.
Authors: Yunbei Zhang, Janet Wang, Yingqiang Ge, Weijie Xu, Jihun Hamm, Chandan K. Reddy
Abstract: This position paper argues that, for long-horizon tasks evaluated across models with comparable frontier capability, the agent execution harness, namely the infrastructure layer that governs context construction, tool interaction, orchestration, and verification around a language model, is often a stronger determinant of agent performance than the model it wraps. We formalize and defend the Binding Constraint Thesis: in this regime, performance variance is governed more by harness configuration than by model choice, and current evaluation protocols therefore systematically misattribute harness-level gains to model improvements. We support this thesis along three lines. First, a control-theoretic formalization treats the harness as the controller of a closed-loop dynamical system and the LLM as the stochastic policy it governs, which explains why small harness changes can produce performance shifts that exceed those obtained by substituting one model for another. Second, published benchmarks, industry deployments, and a controlled variance decomposition show that harness-induced variance can substantially exceed model-induced variance, including cases of model ranking reversal. Third, we propose a harness-aware evaluation framework with a disclosure standard and a variance decomposition protocol. Until harness specifications are disclosed, leaderboard comparisons for long-horizon agents should be treated as incomplete and potentially misleading.
Authors: Alfredo Metere
Abstract: The companion paper introduced a four-level verification lattice on agent-skill manifests (unverified, declared, tested, formal) and left the top level aspirational. This paper closes that gap. We give a precise semantics for skill behaviour faithful to how a skill is consumed by an LLM-driven runtime (a deterministic script-side reachable through a non-deterministic LLM-side), state the verification problem as a capability-containment property over that semantics, and present three composable methods that together raise a skill from declared or tested to formal: (1) sound static capability-containment analysis of the script-side via abstract interpretation over a small effect lattice; (2) a refinement type system for tool-call envelopes that mechanically rejects any call whose statically-inferred capability is not in the manifest's declared set; (3) SMT-bounded model checking against the parent paper's biconditional correctness criterion, with the bound chosen so any counter-example fitting the runtime's transaction-buffer horizon is exhibited as a concrete trace. We prove the three layers composed soundly cover the parent paper's threat model modulo a single residual (the LLM's freedom to refuse to act) that the parent paper's runtime biconditional catches at session boundary. The methods reuse existing well-engineered tools (Z3, Semgrep, CodeQL, refinement-type checkers, mechanised proof assistants) rather than asking operators to build new ones, and the proof-carrying artifact extends the existing SKILL.md convention. All three methods plus the bundle producer and re-checker ship as zero-dependency JavaScript modules in the open-source enclawed framework (https://github.com/metereconsulting/enclawed; project page https://www.enclawed.com/), with 53 unit tests and an end-to-end CLI demo on a sample skill.
URLs: https://github.com/metereconsulting/enclawed;, https://www.enclawed.com/),
Authors: Alex Bogdan, Adrian de Valois-Franklin
Abstract: Artificial agents now generate behavior rich enough to invite trust, surprise, and concern, yet our evaluation tools still privilege capability scores over psychological structure. This paper argues that the philosophical impasse between two symmetrical errors (Artificial Mind Blindness, which dismisses psychological organization in non-biological systems, and Artificial Mind Projection, which infers human-like inner life from fluent behavior alone) can be circumvented not by resolving the consciousness question, but by introducing a disciplined measurement layer beneath it. Drawing on Michael Levin's continuum view of cognition as goal-directed competency across substrates, and on the methodological repertoire of mathematical psychology (Item Response Theory, Signal Detection Theory, Bayesian cognitive modeling, calibration analysis, cognitive-bias batteries), the paper develops Machine Psychometrics as a measurement science of latent behavioral, metacognitive, communicative, and self-modeling dispositions in artificial agents. Its operational core is the Machine Mindprint: a multidimensional, domain-bounded, versioned profile spanning calibration, source integrity, suggestibility resistance, context stability, expressive alignment, tool integrity, drift monitoring, and distributional grounding. A complementary Trust Protocol turns Mindprints into deployment decisions through probe batteries, perturbation testing, reliability and validity analysis, and longitudinal monitoring across high-stakes domains. The philosophical contribution is a third stance, Artificial Mind Discipline, that neither anthropomorphizes nor dismisses, neither presupposes consciousness nor forecloses it. The aim is not to humanize artificial agents, but to understand them precisely because they are not human, through measurement before judgment.
Authors: Ruizhe Zhou, Xiaoyang Liu, Gaoyuan Du, Yi Zheng, Shouxi Ren, Deepayan Chakrabarti, Dengdu Jiang
Abstract: Deploying machine learning in regulated financial environments -- credit risk, fraud detection, and anti-money laundering -- exposes critical vulnerabilities in algorithmic reproducibility. While early financial ML addressed statistical challenges such as backtest overfitting, deep neural networks and Generative AI have introduced mechanical nondeterminism rooted in hardware and architecture. This survey provides a systems perspective on reproducibility failures across three modalities now dominant in financial AI: tabular models (post-hoc explanation variance), graph networks (stochastic sampling and temporal asynchrony), and LLM-based agentic workflows (batch-dependent divergence and trajectory drift). We supplement the literature analysis with first-party experiments on public financial datasets -- quantifying explanation rank instability in credit scoring, prediction flip rates in GNN-based fraud detection, and tensor-parallel-induced output divergence in LLM entity extraction. We propose a layered evaluation framework linking modality-specific metrics (RBO, D_cos, TDI, PSD) to audit readiness, and empirically validate the complementarity of logit-level and semantic-level determinism measures.
Authors: Prashanti Nilayam, Sankalp Nayak
Abstract: Compound AI systems that chain multiple LLM calls into directed computation graphs are now the dominant architecture for production AI. Although these architectures leverage heterogeneous nodes with mixed-mode outputs, no existing framework quantifies how perturbations propagate through such pipelines, where nodes are stochastic and execution paths can diverge structurally. We introduce QUIVER, a formal framework for measuring perturbation propagation in graph-structured LLM pipelines. The framework defines: (1) a sensitivity matrix with type-dispatched distance metrics that classifies edges as amplifiers, absorbers, or threshold-sensitive, complemented by occurrence-lift; (2) trajectory divergence decomposing variation into value drift, structural path divergence, and iteration count divergence; (3) bifurcation thresholds identifying the smallest perturbation that causes structural execution path changes; and (4) distribution faithfulness, quantifying when per node evaluation datasets diverge from production distributions. We validate on two production enterprise pipelines and a public DSPy multihop QA pipeline, three structurally distinct architectures. Across 8,200+ instrumented traces (32,000+ pair comparisons), we demonstrate that QUIVER reveals distinct sensitivity profiles across architectures, distinguishes mechanistically different cascade patterns producing identical divergence rates, predicts nodes prone to trajectory bifurcation from observational data alone, and localizes stale evaluation artifacts to specific node-field categories that aggregate metrics cannot surface.
Authors: Junhao Wei, Yanxiao Li, Yifu Zhao, Zhenhong Peng, Baili Lu, Dexing Yao, Haochen Li, Qinbin He, Sio-Kei Im, Yapeng Wang, Xu Yang
Abstract: Learning-assisted hyper-heuristics can select among dispatching rules while preserving the feasibility and interpretability of constructive Job Shop Scheduling Problem (JSSP) heuristics. Their main computational cost lies in label generation rather than model fitting, since each supervised label usually requires rolling out candidate rules from a partial schedule. We study this label-cost problem together with a reliability problem: a learned selector should not switch away from a strong default rule unless the predicted gain is credible. The proposed selector uses regret-normalized rollout labels, a contextual KNN uncertainty estimate, and a gate that acts only when the predicted improvement exceeds an uncertainty-adjusted margin. We also vary rollout depth and breadth to measure the cost-quality trade-off. On synthetic JSSP instances, the gated selector achieves the lowest mean RPD among learned selectors, remains close to the best fixed dispatching rule, and reduces Random-HH mean RPD by more than an order of magnitude.
Authors: Zenghui Zhou, Man Li, Xiaoke Fang, Xinyi Zhou, Weibin Li, Zheng Zheng
Abstract: Large Language Models (LLMs) achieve strong performance on logical reasoning benchmarks, yet their reliability remains uncertain. Existing evaluations rely on static benchmarks, which fail to assess robustness under logically equivalent transformations and often overestimate reasoning capability. We propose LGMT (Logic-Grounded Metamorphic Testing), an oracle-free framework that leverages first-order logic (FOL) to evaluate LLM reasoning. By deriving metamorphic relations from formal logical equivalences, LGMT constructs semantically invariant test cases and detects reasoning defects through cross-case consistency checking. Experiments on six state-of-the-art LLMs show that LGMT exposes substantial hidden defects missed by traditional reference-based evaluations. We further find that models are particularly sensitive to symbol-level and conclusion-level variations, and that advanced prompting such as Few-shot CoT only partially mitigates these issues. These results suggest that LLM evaluation should move beyond isolated correctness toward robustness under logical invariance. LGMT provides a principled and scalable approach for diagnosing reasoning failures.
Authors: Feisal Alaswad, Batoul Aljaddouh, Maher Alrahhal, Poovammal E, Talal Bonny
Abstract: Large language models achieve strong performance in language generation and knowledge-intensive tasks, yet remain limited in settings requiring causal reasoning, persistent state tracking, and long-horizon planning. We argue that these limitations may arise from an objective-level mismatch between sequence prediction and reasoning over latent environment dynamics. To formalize this distinction, we introduce Latent Dynamics Inference (LDI), a conceptual perspective that interprets language and multimodal observations as partial evidence of underlying transition dynamics. To empirically investigate this perspective, we introduce Flux, a sequential reasoning environment specified entirely through natural-language rules. As a proof-of-concept case study, the rules are first compiled into an explicit state-transition simulator, illustrating that structured latent transition dynamics can, in some cases, be operationally extracted from textual rule descriptions. This enables a controlled comparison between the LLMs operating purely over textual observations and reinforcement-learning agents trained directly within the extracted latent state space. Within this case study, agents operating with explicit access to the latent state space exhibit substantially more stable behavior in long-horizon gameplay, achieving an aggregate win rate of approximately 79% versus 11% for LLMs. Qualitative analysis further reveals failure modes consistent with unstable persistent state tracking, including invalid actions, state-tracking errors, and short-horizon reasoning failures. The complete implementation of the Flux environment available at https://github.com/FeisalAlaswad/FLUX-RL-Agent Within the evaluated setting, these results suggest that strong sequence prediction alone may struggle to support robust long-horizon dynamic reasoning without mechanisms for persistent state tracking and transition modeling
Authors: Fabio Rovai
Abstract: We investigate growth dynamics in deterministic equational discovery substrates. Across three toy domains (arithmetic, boolean, higher-order list; n=592 trajectories), short-range substrate sizes fit a power-law N(t) proportional to t^b. Within each substrate b is architecture-sensitive (cross-validated R^2 approximately 0.82); the regression does not transfer across substrates (arith+bool to list yields R^2 approximately -0.84). A heuristic mean-field closure model predicts a saturating power-law dN/dt = K N^k exp(-mu N) of which the pure power-law is the short-range approximation. Three robustness checks: bootstrap intervals on (k, mu) are tight in 4/5 toy trajectories and degenerate in 1/5; out-of-sample forecasting on toy data (fit first 100 epochs, predict next 400) is won by pure power-law 5/5, indicating the toy trajectories do not reach saturation; on two real-world growth proxies the result splits. New Mathlib/*.lean file additions per month (mathlib4, 60 months, 9701 files) support the saturating form on OOS forecasting by approximately 7x over pure power-law; Coq mathcomp monthly commits (129 months, 3083 commits) favour pure power-law on both tests with mu collapsing to zero. The dynamics are substrate-conditional at two levels: within-substrate architecture-to-b regressions do not transfer, and the preferred functional family for N(t) itself (pure vs. saturating power-law) differs by substrate. We propose "saturating power-law growth with substrate-conditional (k, mu), observable when the substrate has reached its saturation regime" as a working framing.
Authors: Hong Su
Abstract: Autonomous robots operating in open and changing environments cannot always rely on predefined inputs, outputs, and action routines. Although existing learning methods enable robots to improve their performance through environmental interaction, the objects of learning are often fixed in advance, such as input features, recognition outputs, network structures, task goals, or action sequences. This limits their ability to adapt when new features, new categories, or more efficient task routines appear during long-term operation. To address this problem, this paper proposes a thinking-learning interaction model for autonomous robots. The core idea is that thinking guides learning by identifying potential changes, selecting useful evidence, organizing training materials, and planning verification actions, while learning promotes thinking by updating task knowledge, feature-selection experience, action strategies, and future reasoning processes. Based on this bidirectional mechanism, the robot can gradually move beyond predefined learning settings and adapt its recognition relations and action relations through continuous interaction with the environment. Specifically, the proposed model supports adaptive input feature discovery, output category expansion, learning model update, and action routine reconstruction. Experimental results show that the proposed model improves the final recognition accuracy from 0.419 to 0.845 in feature adaptation, achieves higher new-category formation accuracy and model-update success rate, and reduces the average action length from 13.0 to 4.0 in action routine reconstruction. In learning-enhanced thinking, the useful evidence selection rate increases from 0.272 to 0.965, indicating that learning results can effectively improve future evidence selection and reasoning.
Authors: Jinhu Qi, Muzhi Li, Jiahong Liu, Yuqin Shu, Dianzhi Yu, Shicheng Ma, Wenqian Cui, Yiyang Zhao, Yiyi Chen, Ruoxi Jiang, Irwin King, Zenglin Xu
Abstract: Agentic AI systems -- Large Language Models (LLMs) augmented with planning, tool use, memory, and long-horizon interactions -- can execute complex tasks autonomously, but their multi-step trajectories introduce new failure modes that challenge trustworthiness. This survey provides a focused examination of trustworthy agentic AI through two core dimensions that are critical for high-risk deployments: Safety and Robustness, and Privacy and System Security. For each dimension, we clarify key concepts, identify where risks emerge along the agent workflow, and summarize stage-targeted mitigation strategies. Other trustworthiness aspects (value alignment, transparency, fairness, and accountability) are discussed as relevant context rather than parallel chapters. To support consistent comparison and deployment decisions, we consolidate evaluation into a unified metrics-and-benchmarks hub, emphasizing both outcome and process signals (e.g., constraint violations, trace completeness, and adversarial success rates) and offering scenario-to-metric guidance for release gating. We conclude by outlining open challenges such as self-evolving agents, runtime monitoring and verification, privacy-preserving personalization, and the trust-utility trade-off, and present a case study of real-world security failures in open-source agentic systems. Our goal is to serve as a practical reference for researchers and practitioners building trustworthy agentic systems in high-stakes environments.
Authors: Zhengqi Sun, Yiwen Sun, Boxuan Liu, Tailai Chen, Tianxu Guo, Jiabin Liu
Abstract: Large language models (LLMs) are promising for autonomous driving, but semantics-only decision policies can yield physically unsafe behavior in dynamic traffic. Existing methods either perform online language reasoning without explicit dynamics verification or use world models mainly in offline pipelines, leaving a gap between semantic intent and physical feasibility at decision time. We propose Reason--Imagine--Act (RIA), a closed-loop framework that couples an LLM reasoner with an action-conditioned world model for online safety verification. At each step, the LLM proposes an action template and candidate sub-actions, the world model performs short-horizon rollouts, and a safety scorer selects the safest executable action with feedback to the next reasoning step. Under a unified CARLA point-goal protocol (1000 episodes), RIA achieves 80.05% route completion, 51.10% arrival rate, and 0.20% collision rate. Under the same closed-loop interface, RIA consistently outperforms training-free baselines, including CARLA TM and MADA, on core closed-loop metrics. For reproducibility, code is available at https://github.com/pku-smart-city/source_code/tree/main/RIA.
URLs: https://github.com/pku-smart-city/source_code/tree/main/RIA.
Authors: Yanyu Chen, Jiyue Jiang, Dianzhi Yu, Zheng Wu, Jiahong Liu, Jiaming Han, Xiao Guo, Jinhu Qi, Yu Li, Yifei Zhang, Irwin King
Abstract: The evolution of Large Language Model (LLM) reasoning is bottlenecked by the scarcity of high-quality process data. While self-alignment via endogenous rewards offers a solution, mining valid supervision faces three challenges: (1) Label Noise via Mimetic Bias, where rewards prioritize statistical likelihood over logical truth, creating a "correctness illusion" that masks compounding errors; (2) Coarse-Grained Supervision, where sparse global outcomes (e.g., in GRPO) fail to provide granular guidance, treating reasoning chains as monolithic; and (3) Distributional Collapse, where signals fail to generalize without amplifying pre-training biases. To address these, we introduce LC-ERD (Logic-Consistent Endogenous Reward Decomposition), a framework framing self-alignment as latent structure mining. We derive a Variational Logic Potential by aggregating consensus from the model's Latent Logic Expertise (LLE) to denoise the reasoning manifold, and introduce a Multi-Agent Value Decomposition protocol based on the IGM principle to quantify individual step utility. Experiments show LC-ERD delivers a robust self-evolution path, uncovering trade-offs between logic consistency and accuracy while identifying high-value reasoning patterns missed by standard rewards. Our code is available at https://github.com/Reinhardmannn/LC-ERD.
Authors: Xiaoyu Xiong, Yuqi Ren, Deyi Xiong
Abstract: Large language models (LLMs), have shown strong potential in scientific discovery, yet existing methods still face substantial challenges in the design of research workflows and multi-role collaboration mechanisms. To mitigate these issues, we propose EvoSci, a multi-agent scientific collaboration framework, which integrates bio-inspired evolution with knowledge graph modeling. To iteratively generate, evaluate, and refine research ideas, EvoSci incorporates multiple role-based agents, including mentor, researcher, and reviewer. By combining collaborative reasoning, shared memory, and evolutionary feedback, EvoSci significantly enhances the coherence and creativity of scientific exploration. Experiments on real-world research topics demonstrate that EvoSci significantly outperforms strong baselines in LLM-based structured peer-review and comparative ranking evaluations, achieving the highest overall peer-review score (ICLR 4.90) and top ranking (Top-10 = 54). These results suggest its superiority in both scientific idea generation and continuous discovery.
Authors: Maikel Yelandi Leyva-V\'azquez, Florentin Smarandache
Abstract: Large Language Models (LLMs) are predominantly governed by probabilistic frameworks in which the sum of outcome probabilities is constrained to unity. This architectural limitation, often imposed by Softmax layers, leads to a collapse of uncertainty that makes it difficult to differentiate between epistemic uncertainty, paradox, and vagueness. We present an empirical investigation of the application of Neutrosophic Logic, a framework that treats Truth (T), Indeterminacy (I), and Falsity (F) as three independent dimensions, to model epistemic states in LLMs. We conducted experiments on a family of four OpenAI GPT models across five linguistic phenomena: logical paradoxes, epistemic ignorance, vagueness, ethical contradictions, and future contingencies, under three prompting strategies: neutrosophic, probabilistic, and entropy-derived. Our findings reveal that the neutrosophic approach, by allowing T+I+F > 1, a state we term hyper-truth, provides a richer representation of a model's internal state. In 35% of evaluations, hyper-truth emerged spontaneously, predominantly under ethical contradiction and logical paradox. We demonstrate that this approach preserves truth values in fuzzy contexts and offers a robust method for identifying and quantifying internal model conflict. We conclude that the integration of neutrosophic evaluation layers is a critical step toward more transparent, reliable, and ethically aware AI systems.
Authors: Haiyang Shen, Xuanzhong Chen, Wendong Xu, Yun Ma, Liang Chen, Kuan Li
Abstract: Coding agents are increasingly used as iterative development partners, but most benchmarks still evaluate one specification followed by one final assessment. This leaves out a basic question: can an agent keep its own codebase working as requirements change? We introduce EvoCode-Bench, a benchmark of 26 stateful coding tasks and 227 evaluated rounds. Each task preserves the agent's workspace for 5-15 rounds, states requirements through observable behavior, and uses cumulative executable tests to check new requirements and still-active prior ones. We evaluate 13 coding agents with two metrics: MT@4, a four-attempt fail-stop multi-round score, and SR, a single-round score from a reference-completed prior state. For most agents, SR exceeds MT@4 by 22-40 points. The gap also changes rankings: the highest-SR agent (78.9) ranks only third in persistent execution (44.0 MT@4). Even the strongest agents achieve only about 50% success on multi-turn metrics, and aggregate pass rate drops below half of round-1 performance by round 5. Failure analysis shows tier-dependent behavior: weaker agents fail early, while stronger agents survive long enough to expose specification-tracking and regression failures. We release the benchmark data and Harbor multi-turn infrastructure.
Authors: Yingtie Lei, Zhongwei Wan, Jiankun Zhang, Samiul Alam, Zixuan Zhong, Peizhou Huang, Xin Wang, Jingxuan Zhang, Donghao Zhou, Yunta Hsieh, Zhihao Dou, Hui Shen, Yan Xu, Dimitrios Dimitriadis, Tuo Zhang, Mi Zhang
Abstract: Large language model (LLM) agents accumulate rich episodic trajectories while solving real-world tasks, but it remains unclear whether such experience can be distilled into reusable procedural skills. We introduce SkillEvolBench, a diagnostic benchmark for evaluating this step from experience reuse to skill formation. It contains 180 tasks across six real-world agent environments, organized into role-conditioned task families with shared latent procedures. Agents learn from acquisition tasks, update an external skill library using compacted trajectories and verifier feedback, and then face frozen deployment tasks testing context shift, adversarial shortcuts, and composition. By comparing self-generated and curated-start skill evolution against no-skill and raw-trajectory controls, SkillEvolBench separates procedural abstraction from base capability, curated prior knowledge, and direct reuse of episodic traces. Across ten model configurations and three agent harnesses, we find that current agents often adapt locally but rarely form robust reusable skills. Skill-based conditions can improve acquisition or replay, and individual models sometimes gain on specific deployment axes, but these gains are unstable under frozen deployment. Raw-trajectory reuse frequently outperforms distilled skills, suggesting that current abstraction procedures discard contextual and procedural cues that remain useful for future tasks. Capacity and cost analyses further show that writing more skills or larger Tier-3 resource libraries is not sufficient: additional updates can improve coverage while introducing episode-specific drift and procedural clutter. These findings position SkillEvolBench as a testbed for measuring when one-off experience becomes durable procedural knowledge rather than task-local memory.
Authors: Qian-Rong Li, Hung Guei, I-Chen Wu, Ti-Rong Wu
Abstract: Imperfect-information games (IIGs) are challenging, as players must make decisions without fully observing the true game state. While AlphaZero has achieved remarkable success in perfect-information games, extending it to IIGs remains difficult. Existing search-based approaches, such as Perfect Information Monte Carlo (PIMC), suffer from strategy fusion, while Information Set Monte Carlo Tree Search (IS-MCTS) incurs high computational cost when combined with neural networks. In this paper, we propose Multi-State Aggregated PoLicy Evaluation (MAPLE), a tree search method that aggregates policy and value evaluations from multiple sampled world states within a single search tree, combining the advantages of PIMC and IS-MCTS while maintaining a controllable computational cost. We further incorporate a Siamese-based sampling strategy to select informative world states from the information set. Experiments on Phantom Go and Dark Hex show that MAPLE significantly outperforms the PIMC-based AlphaZero baseline, achieving Elo improvements of 291 and 136, respectively. These results demonstrate that MAPLE is an effective approach for AlphaZero-style learning in imperfect-information games.
Authors: Yuyu Liu, Haotian Xu, Yanan He, Sarang Rajendra Patil, Mengjia Xu, Tengfei Ma
Abstract: Multi-step reasoning remains a central challenge for large language models: single-pass generation is efficient but lacks accuracy; tree-search methods explore multiple paths but are computation-heavy. We address this gap by distilling reasoning progress into a hyperbolic geometric signal that guides step-by-step generation. Our approach is motivated by a structural observation: in combinatorial reasoning trees, solution-bearing states are few while dead ends are exponentially numerous. The hyperbolic space matches this asymmetry, with compact volume near the origin and exponentially expanding capacity toward the boundary, so that distance-to-origin naturally encodes solution proximity while angular separation distinguishes branches requiring different next operations. We train a lightweight head to project LLM hidden states into this space, then fine-tune a low-rank adapter interactively on its own reasoning attempts to act on the injected signal. Across multiple benchmarks, the geometric signal yields consistent gains, with larger improvements on deeper reasoning chains. Our code is publicly available at https://github.com/yuyuliu11037/HyperGuide.
Authors: Maryna Kapitonova, Tonio Ball
Abstract: We present a neuro-inspired framework for embodied planning and control. Building on three principles that enable fast and highly effective goal-directed behavior in the mammalian brain - paired forward/inverse internal models, open-loop multi-step motor commands, and sequential, hierarchical organization of action - our Inverter framework uses learned components, trained end-to-end through Inverse Learning (IL) and supplemented where natural by analytic or algorithmic modules; we formalize IL and delineate it from supervised, reinforcement, and imitation learning. IL bridges Reinforcement Learning (RL)-style amortization, which runs in a single forward pass but emits only one action at a time, and Optimal Control (OC)-style sequence planning over whole trajectories, but with iterative test-time computation. Single Inverters or hierarchical n=2 Inverter stacks match or improve on offline-RL and diffusion-planner baselines on all 3 maze2d and 6 antmaze D4RL variants by an average of +24.2% (range -1.9% to +78.2%), at one-to-two orders of magnitude less inference compute time. Distinctively, optimizing through the Forward Model (FoM) over the entire T-step action sequence - rather than per step - lets Inverters produce smooth, goal-coherent, trajectory-wide structure and reach control policies closer to the analytic optimum than the policy underlying the training data itself. We also identify a failure mode of IL: FoM hacking under narrow training-data coverage, which we mitigate by using random training data with broader coverage. As an application example, a Pulse Inverter synthesizes arbitrary single-qubit quantum gates with fidelity matching the standard iterative numerical baseline (GRAPE), at more than 1000x lower per-gate compute time. In summary, we conclude that IL enables a versatile class of world-interfaces, especially for latency- and resource-critical embodied AI.
Authors: Qitao Tan, Xiaoying Song, Arman Akbari, Arash Akbari, Yanzhi Wang, Xiaoming Zhai, Lingzi Hong, Zhen Xiang, Jin Lu, Geng Yuan
Abstract: Current safety alignment of foundation models largely follows a \emph{one-size-fits-all} paradigm, applying the same refusal policy across users and contexts. As a result, models may refuse requests that are unsafe for general users but legitimate for authorized professionals, limiting helpfulness in specialized professional settings. Existing approaches either require costly realignment or rely on inference-time steering that suffers from imprecise control and added latency. To this end, we propose \textsc{Palette}, a modular, controllable, and efficient framework that selectively relaxes refusal behavior on authorized target domains while preserving standard safety elsewhere. Our method identifies a refusal direction via multi-objective search and internalizes it into the model through lightweight adaptation. \textsc{Palette} further supports modular composition: it learns domain-specific safety controls independently and composes them through parameter merging, enabling on-demand multi-domain authorization without retraining. Experiments across four safety benchmarks, multiple model variants, and both LLMs and VLMs show that \textsc{Palette} delivers precise safety control without sacrificing general utility, offering a practical path toward foundation models that adapt to diverse professional needs.
Authors: Sahil Joshi, Prithvi Dixit, Agniva Chowdhury, Anshumali Shrivastava, Joseph E. Gonzalez, Ion Stoica, Kumar Krishna Agrawal, Aditya Desai
Abstract: Sparsity has long been a central theme in LLM efficiency, but its role in context processing remains unresolved. As LLM workloads shift toward longer contexts and agentic interactions, the compute and memory bottlenecks of attention become increasingly critical, raising the question of whether these constraints are fundamental. Our position is that these constraints are artificial and unnecessary, and that the future of LLM inference lies in extreme but principled sparsity along the context dimension. This position is supported by several strands of empirical and theoretical evidence. First, we find the insistence on dense attention unreasonable, since in a long context a query effectively projects O(N) attention information into a hidden space of dimension d << N, making the process inherently lossy. Second, we perform an extensive study of sparsity in LLMs spanning 20 models across five model families, varying context lengths, and different sparsity levels. We empirically demonstrate a strong trend: current LLMs, despite not being trained for context sparsity, are remarkably robust to inference-time decode sparsity across tasks of varying complexity, including retrieval, multi-hop QA, mathematical reasoning, and agentic coding. Importantly, we also show that current hardware is already sufficient to realize substantial gains from this sparsity. For example, our sparse decode kernels accelerate large-context processing by up to 10x over FlashInfer at 50x sparsity levels on hardware such as the H100. Overall, these results position extreme context sparsity not as a heuristic, but as a principled foundation for LLM inference, training, and architecture design: one that is both feasible and beneficial, and a compelling direction for future systems.
Authors: Samah Fodeh, Sreeraj Ramachandran, Elyas Irankhah, Muhammad Arif, Afshan Khan, Ganesh Puthiaraju, Linhai Ma, Srivani Talakokkul, Jordan Alpert, Sarah Schellhorn
Abstract: Secure patient-provider messages contain clinically important communication behaviors that are difficult to characterize manually at scale. The Electronic Patient-Provider Communication (EPPC) framework provides an ontology for coding these behaviors, but automated extraction remains challenging because predictions must preserve fine-grained code/sub-code structure while grounding annotations in message text. We developed EPPC-OASIS, an ontology-aware adaptation approach for structured EPPC extraction, and combined it with deployable inference-refinement procedures designed to improve the coherence of final annotations. EPPC-OASIS augments supervised fine-tuning with a Wasserstein alignment objective that encourages alignment between model representation neighborhoods and EPPC ontology-derived neighborhoods, while inference refinement uses verification, self-consistency, hybrid correction, and selection or ensembling to address residual prediction errors. We evaluated the framework on a de-identified corpus of secure patient-provider messages against prompting, supervised fine-tuning, preference-based, and robustness-oriented baselines across multiple open-weight language models. Across model families, the best deployable pipeline achieved 77.13% Code+Sub-code F1 and 63.83% Triplet F1, corresponding to modest but consistent absolute gains of +1.39 and +2.12 F1 points over the strongest supervised fine-tuning baseline. These results suggest that ontology-aware adaptation with structured inference refinement can support scalable retrospective EPPC mining, although external validation is needed before operational use.
Authors: Wenqian Ye, Bo Yuan, Zhichao Xu, Yijun Tian, Yawei Wang, Henry Kautz, Aidong Zhang
Abstract: We study a class of emergent misalignment in multi-agent systems (MAS), with a focus on automated workflows, which we refer to agentic misalignment. Although these systems can solve complex tasks, they often fail because agents act according to implicit proxy utilities that do not align with the intended human goals. We formally define these behaviors and analyze them within a Bayesian framework, showing that generic utilities naturally lead to posterior collapse of agents in automated workflows. To address this issue, we propose Agentic Evidence Attribution (AEA), a novel alignment paradigm that improves agent posteriors using context-specific evidence. AEA reasons over agent actions and provides structured evidence to correct misaligned behavior during collaboration. To better understand the role of evidence, we study two instantiations of AEA: self-reflection (internal evidence from the model) and weak-to-strong generalization (external evidence on the agentic trajectory). We show that a small evidence model effectively aligns the MAS by providing orthogonal failure attribution. Our results clarify the sources of agentic misalignment in automated workflows and show that evidence-based alignment can effectively improve agent collaboration and leads to reliable multi-agent systems built on automated workflows.
Authors: Yifan Zeng, Yiran Wu, Yaolun Zhang, Wentian Zhao, Kun Wan, Qingyun Wu, Huazheng Wang
Abstract: Multi-agent LLM workflows route inference through specialized roles to lift end-task accuracy, but jointly training those roles with reinforcement learning is unstable in ways that are poorly understood. We study when end-to-end RL training of multi-agent LLM workflows improves over their base models, comparing Shared-Policy training, where all roles update one policy, with Isolated-Policy training, where each role has its own parameters. Our experimental matrix spans Eval-Opt, Voting, and Orch-Workers workflows, math and code tasks, and three model scales (0.6B, 1.7B, 4B). We find that multi-agent RL usually improves over base models, but gains depend jointly on workflow, task, and scale, not on policy sharing alone. Isolated-Policy tends to reach higher peak accuracy yet more often falls off a terminal accuracy cliff, while Shared-Policy training does not eliminate failure; it redistributes failure into qualitatively different patterns. We then explain the strongest of these patterns through role-level gradient dynamics induced by workflow topology and policy routing: under Isolated-Policy, parallel same-role agents on shared prompts amplify per-role gradients and drive terminal degradation in Voting and Orch-Workers workflows; under Shared-Policy, asymmetric per-step gradient mass causes the shared policy to be captured by the dominant role, producing different failure signatures by task and workflow. Together, the empirical map and its underlying mechanisms show that policy sharing routes training pressure through different channels rather than offering uniform stability, making it a design choice with workflow- and task-conditional tradeoffs.
Authors: Ashok Chandrasekar, Jason Kramberger
Abstract: As Large Language Models (LLMs) transition from research environments to production deployments, evaluating their performance against strict Service Level Objectives (SLOs) has become critical. However, current evaluation methodologies suffer from severe measurement bias at scale. We demonstrate that widely used benchmarking utilities rely on single-process, asyncio-driven architectures that introduce fundamental client-side queuing bottlenecks under high concurrency. By modeling the benchmarking client as an $M/G/1$ queue, we mathematically demonstrate how the Python Global Interpreter Lock (GIL) artificially inflates Time to First Token (TTFT) and Time Per Output Token (TPOT) metrics as request rates scale. To resolve this systematic inaccuracy, we propose an unbiased, multi-process evaluation framework that effectively distributes client-side load, ensuring negligible queuing overhead. Furthermore, we formalize a composite metric, Normalized Time Per Output Token (NTPOT), to robustly amortize end-to-end latency, including prefill and scheduling delays across sequence lengths. Our empirical evaluation demonstrates that this methodology successfully isolates pure serving engine performance, enabling accurate, reproducible profiling of LLMs at production scales exceeding thousands of queries per second.
Authors: Harshada Badave, Santosh Borse, Andrea Gomez, Harshitha Narahari, Sara Carter, Vishwa Bhatt, Aishani Rachakonda, Shuxin Lin, Dhaval Patel
Abstract: Large Language Models (LLMs) are increasingly deployed as autonomous agents that reason, use tools, and act over multiple steps. Yet most hallucination benchmarks still evaluate only the final output, missing failures that originate in intermediate Thought-Action-Observation steps. We present Trajel, a dataset and evaluation framework for auditing trajectory-level hallucinations in multi-agent industrial workflows. Trajel introduces a five-type hallucination taxonomy (factual, referential, logical, procedural, and scope-based) over expert-annotated agent traces from AssetOpsBench. We benchmark supervised detection models at the subtask, trajectory, and long-context levels. Our results show that the most common failure modes are missed by existing benchmarks, that nearly half of hallucinated trajectories involve multiple types at once, and that automated detectors with high binary accuracy still misclassify the subtlest types. Trajectory-aware detection significantly outperforms standard post-hoc verification, making taxonomy-grounded evaluation necessary for safer agentic deployment.
Authors: Arya Jakkli, Senthooran Rajamanoharan, Neel Nanda
Abstract: Frontier AI developers now train models against long written behavioral specifications, such as Anthropic's constitution (Anthropic, 2025a) and OpenAI's Model Spec (OpenAI, 2025a), integrated into post-training via methods like character training (Anthropic, 2024) and deliberative alignment (Guan et al., 2024). These documents serve a governance function, but it is unclear how well models actually follow them under adversarial, multi-turn pressure similar to what they would face in real-world deployment. We propose a multi-method audit pipeline that treats each lab's published specification as an auditable target: it decomposes the specification into atomic testable tenets (205 for Anthropic, 197 for OpenAI), generates multi-turn adversarial scenarios with the Petri auditing agent (Anthropic, 2025b), runs a modified SURF-style rubric search (Murray et al., 2026) to catch shallow single-turn failures Petri misses, validates flagged transcripts against the relevant specification, and compares the findings against the lab's own published system card. Applying the pipeline across seven models per specification, we find that models follow their own lab's specification substantially better with each generation. On Anthropic's constitution, the Claude family falls from a 15.0% violation rate (Sonnet 4) to 2.0% (Sonnet 4.6); on OpenAI's Model Spec, the GPT family falls from 11.7% (GPT-4o) to 3.6% (GPT-5.2 medium reasoning), with the severity ceiling falling from 10/10 to 7/10. We cannot externally isolate whether these gains come from specification-specific training, broader post-training improvements, or evaluation awareness. Remaining failures cluster around operator-imposed personas under AI-identity questioning, irreversible action in agentic deployments, and fabricated quantitative claims with false precision.
Authors: Banafsheh Rafiee, Richard Sutton
Abstract: In this paper, we advocate for incorporating enactive approaches to perception and cognition into artificial intelligence (AI). Enactive approaches view perception as an active, skillful engagement with the world, where agents perceive by acting and by understanding how their actions shape their experience. This contrasts with classical views that treat perception as a passive internal process in which the brain receives sensory input, processes it, and issues commands for action. Enactive views emphasize the dynamic, embodied, and interactive character of perception, grounded in the lived experience of agents embedded in their environments. We identify and develop four key enactive concepts that we find most relevant to AI: experience, action perception inseparability, autonomy, and embodiment. Much of mainstream AI, from classical rule based systems to large language models, has largely neglected these insights, treating cognition as internal processing detached from embodied interaction and intrinsic normativity. Reinforcement learning (RL), however, exhibits structural resonance with enactive principles through its emphasis on action, agent environment interaction, feedback driven adaptation, and agent centered evaluation. However, this resonance should not be taken as theoretical equivalence, as RL approximates some enactive insights, but key elements remain absent or weakly developed. Building on this analysis, we suggest a broader incorporation of enactive ideas into both mainstream AI and RL.
Authors: Md Nurul Absar Siddiky
Abstract: Sparse mixture-of-experts (MoE) language models activate only a small subset of parameters for each token, making router behavior a central part of model computation. This paper studies routing behavior of Mixtral 8x7B-Instruct under benign and harmful prompts using two complementary signals: activation-based routing scores derived from expert selection frequencies and gradient-based scores derived from router-gate sensitivities. We analyze expert- and layer-level routing behavior and conduct expert-suppression interventions. The results show that activation-based expert usage is broad and long-tailed, whereas gradient-based importance is concentrated. At expert level, benign and harmful prompt groups remain close under both signals with modest separation. At layer level, activation-based routing is most selective around layers 8-15, while gradient-based importance is concentrated in final layers. Expert classification shows most experts are shared across benign and harmful prompts, though a limited subset shows clear group preference. Top-ranked expert sets show stronger benign-malicious overlap under gradient scores than activation scores, suggesting concentration on a common late-layer expert set. In intervention experiments, suppressing top five benign-dominant experts from activation scores reduces restricted responses from 24 to 14 over 100 prompts, while suppressing gradient-derived experts reduces them from 34 to 22 with fewer unintended reversals. Overall, safety-relevant routing in Mixtral is subtle, depth-dependent, and distributed rather than dominated by a fixed set of experts.
Authors: Amirhossein Yousefiramandi, Ciaran Cooney
Abstract: The issues that must be considered regarding the utilization of synthetic data generated through LLMs for multilabel patent classification include (i) when the use of such data may help and (ii) why. Indeed, the former part appropriately adjusts for the possibility of improving results by an increase in sample size. The current experiment involves six open-source LLMs (from 3.8B to 12B parameters) for four real-data regimes in classification of 64 WIPO labels of assistive technologies. Both full-synthesis generation, conditioned on the label set, and paraphrasing methods are applied, with each used in combination with three classifier categories. It is shown that the claimed improvements in micro F1 for BERT-for-Patents from 0.120 to 0.702 mainly reflect a volume effect; indeed, replication with replacement in 165 examples produces 0.678. Thus, the improvement over the control is +0.024, while compared to the best baseline (focal loss reweighting) is +0.219. The second crucial point to consider here is that of evolving fidelity scores as the data generation regime varies. For low real-data regimes, the volume effect dominates and the correlation coefficient between maximum mean discrepancy (MMD) and classification performance equals r = +0.95. As more real data is used, the correlation becomes inverted and reaches r = -0.73 at the 1:10 regime (Fisher z = +6.47, p < 0.001, 95% CI on Delta r [ +0.96, +1.00 ]). In terms of a fixed budget allocation, combining real data (about 20-30%) with synthetic (70-80%) outperforms both purely synthetic and purely real strategies. Moreover, a corpus that allows for improvement in classification performance up to +0.58 in raw micro F1 may adversely affect a Jaccard-overlap retrieval proxy. Prompt-family variations for other genres may provide some explanation of the phenomenon, but using the standard-patent filter still decreases nDCG@10 by 26%.
Authors: Adnan Ahmad, Bahareh Nakisa, Mohammad Naim Rastgoo
Abstract: Human-AI collaboration requires agents that can adapt to diverse partner behaviors and skill levels while remaining robust to unseen partners. Existing methods often collapse to a single dominant behavior or learn poorly aligned skills, limiting effective coordination. We propose Intrinsic Action Disentanglement (IAD), a deep hierarchical reinforcement learning (DHRL) framework that learns distinct, partner-aware low-level action sequences conditioned on high-level latent skills. IAD introduces an intrinsic reward that explicitly encourages disentangled action distributions of the agent's low-level policy across skills, yielding an interpretable mapping between high-level decisions and partner-specific behavioral responses. By capturing temporally extended interaction patterns, IAD enables flexible adaptation to heterogeneous partner dynamics under distributional shift. We evaluate IAD in the Overcooked-AI domain across multiple layouts and diverse partner settings, including unseen simulated partners, a human-proxy model trained on human-human gameplay, and real human partners. Results show that IAD consistently outperforms strong baselines and achieves more reliable, adaptive coordination across all settings.
Authors: Adnan Ahmad, Bahareh Nakisa, Mohammad Naim Rastgoo
Abstract: Multi-agent collaboration, especially in human-AI teaming, requires agents that can adapt to novel partners with diverse and dynamic behaviors. Conventional Deep Hierarchical Reinforcement Learning (DHRL) methods focus on agent-centric rewards and overlook partner behavior, leading to shortcut learning, where skills exploit spurious information instead of adapting to partners' dynamic behaviors. This limitation undermines agents' ability to adapt and coordinate effectively with novel partners. We introduce Partner-Aware Skill Discovery (PASD), a DHRL framework that learns skills conditioned on partner behavior. PASD introduces a contrastive intrinsic reward to capture patterns emerging from partner interactions, aligning skill representations across similar partners while maintaining discriminability across diverse strategies. By structuring the skill space based on partner interactions, this approach mitigates shortcut learning and promotes behavioral consistency, enabling robust and adaptive coordination. We extensively evaluate PASD in the Overcooked-AI benchmark with a diverse population of partners characterized by varying skill levels and play styles. We further evaluate the approach with human proxy models trained from human-human gameplay trajectories. PASD consistently outperforms existing population-based and hierarchical baselines, demonstrating transferable skill learning that generalizes across a wide range of partner behaviors. Analysis of learned skill representations shows that PASD adapts effectively to diverse partner behaviors, highlighting its robustness in human-AI collaboration.
Authors: Tyrone Serapio, Arjun Prakash, Haoyang Xu, Kevin Wang, Amy Greenwald
Abstract: Large Language Models (LLMs) have shown great ability in generating executable code from natural language, opening the possibility of automatically constructing environments for AI agents. Recent work on Code World Models (CWMs) demonstrates that LLMs can translate game rules into Python implementations compatible with solvers like Monte Carlo Tree Search. We study this problem in game settings, where generated environments must implement rules, legal actions, state transitions, observations, and rewards. We refer to these game-specific executable models as Game Code World Models (GameCWMs). However, current approaches to generating code world models rely on frontier models and inference-time refinement loops, limiting accessibility and scalability. This work investigates whether GameCWM generation capabilities can be distilled into smaller models through post-training. We introduce: (1) a curated dataset of 30 games spanning perfect and imperfect information games, (2) a verification framework that evaluates generated code against structural and semantic game properties, and (3) a post-training pipeline combining Supervised Fine-Tuning (SFT) with Reinforcement Learning with Verifiable Rewards (RLVR). We experiment with Qwen2.5-3B-Instruct and find that SFT can increase syntactic correctness, while RLVR can improve execution-level adherence to game rules, thereby improving Qwen's ability to generate valid GameCWMs in both perfect and imperfect information games. Overall, our pipeline makes Qwen2.5-3B-Instruct more capable of generating valid GameCWMs, thereby offering a scalable path toward automatic environment generation from natural language.
Authors: Weiwei Xu, Hengzhi Ye, Haoran Ye, Kai Gao, Vladimir Filkov, Minghui Zhou
Abstract: Ethical constraints on open-weight AI models are both a reflection of societal concerns and a foundation for AI governance policy. They are expected to propagate to downstream derivatives while implemented as voluntary metadata disclosures that must be restated at each generation of reuse. We audit 2,142,823 model repositories on Hugging Face Hub to test whether this disclosure-based governance infrastructure can sustain traceability across deep model lineages. Restriction evidence decays with a half-life of 1.31 derivation steps ($R^2$=0.98), and beyond seven downstream generations at least 80% of descendant models lack sufficient public evidence for a governance determination, a depth boundary we formalize as the governance horizon. Platform-level interventions to restore missing licence metadata reveal that policy design (not enforcement alone) is the binding factor: inheritance-only designs require near-complete enforcement to move the horizon, whereas a mandatory-declaration design that explicitly resolves orphan lineage components shifts the horizon already at moderate enforcement. The structural bottleneck is lineages with no inheritable upstream intent: such orphan components remain undecidable under any inheritance-only policy regardless of enforcement rate, and unresolved upstream nodes additionally create direct downstream undecidability bottlenecks that inheritance rules alone cannot recover. Comparison with PyPI, where governance signals are carried by explicit machine-readable declarations, corroborates that the collapse is topology-specific to open-weight derivation rather than inherent to open ecosystems. These results establish that disclosure-based governance has a shallow, structurally determined reach in open-weight AI, and that achieving deep supply-chain accountability requires provenance mechanisms propagating governance signals through derivation itself.
Authors: Jingchu Gai, Guanning Zeng, Christina Baek, Chen Wu, J. Zico Kolter, Andrej Risteski, Aditi Raghunathan
Abstract: Long chains of thought (CoT) from current language models frequently contain logical gaps and unjustified leaps, limiting the gains from additional test-time compute. Improving reasoning quality directly would require process reward models, but the step-level annotations needed to train them are expensive and scarce. We find such a signal in how the model's confidence evolves during reasoning: premature confidence, the tendency to commit to an answer early and use the remaining tokens to rationalize it, strongly predicts flawed reasoning across tasks and model scales. We exploit this in progressive confidence shaping, a reinforcement learning objective that trains models to update their confidence as they reason rather than commit early -- rewarding gradual confidence growth and penalizing early commitment, with no external labels or reward models. The method improves accuracy and reasoning quality from 1.5B to 8B parameters across arithmetic (Countdown), math (DAPO, AIME), and science (ScienceQA): on Countdown, accuracy improves 3.2x (+42.0pp) and flawed reasoning drops 48pp; on AIME, Pass@64 improves 6.6pp. Consistent with this mechanism, the method also improves faithfulness: on a safety benchmark, our models more transparently surface misleading content in their reasoning traces rather than concealing it. Controlled experiments reveal that the problem and its remedy scale together: premature confidence grows with model size and task difficulty, and so do the gains from addressing it.
Authors: Xuan Wang, Zhongling Xu, Gopi Kannedhara, Joakim Nguyen, Jian Yu, Jinrui Fang, Abdurrahmaan Baghdadi, Tianlong Chen, Awais Naeem, Chandra Krishnan, Edward Castillo, Andrew H. Song, Ankita Shukla, Ying Ding, Nicholas Konz, Hairong Wang
Abstract: Healthcare models are transitioning from unimodal prediction toward multimodal reasoning over heterogeneous diagnostic inputs. In computational pathology, for complex tumor subtypes where morphology alone can be challenging to distinguish, pathology reports and molecular measurements may provide additional diagnostic evidence alongside whole-slide images, yet existing models often fail to clarify how diverse signals assemble into recognizable diagnostic concepts. We propose ConceptM$^3$oE (Concept Multimodal MoE), which embeds concept formation directly within interaction-aware mixture-of-experts (MoE) pathways. The architecture decomposes evidence into modality-specific, redundant, and synergistic experts, which are then projected into structured concept bottlenecks mapping latent features to a hierarchy of morphology and biomarker concepts. To prevent the information loss typical of interpretable bottlenecks, we utilize residual pathways within each expert to allow task-relevant signals to flow both through the concepts and directly to the final task prediction, so that high performance is maintained alongside interpretability. Across an institutional pediatric brain tumor cohort and a public glioma cohort, the framework delivers competitive performance to unconstrained models while producing reasoning traces validated by an independent neuropathologist. In data-limited regimes, ConceptM$^3$oE improves limited-data performance, increasing macro-F1 from 56.41% to 66.70% at small training sizes compared to non-concept-informed baselines, while also showing faster training convergence consistent with the regularizing effect of concept learning. This work offers a scalable path toward high-performance medical AI that is inherently verifiable and better aligned with the complex decision-making of clinical practice.
Authors: Renchu Guan, Yajun Wang, Chunli Guo, Bowen Cao, Fausto Giunchiglia, Wei Pang, Yonghao Liu, Xiaoyue Feng
Abstract: Graph few-shot learning, which aims to classify nodes from novel classes with only a few labeled examples, is a widely studied problem in graph learning. However, existing methods often face two key limitations. First, the predominant graph few-shot learning paradigm relies on supervised tasks, failing to leverage the vast number of unlabeled nodes in the graph. Second, many approaches require complex task adaptation or fine-tuning during inference, limiting their efficiency and applicability. Inspired by the powerful in-context learning capabilities of large language models, we propose a novel model named VISION for adVancIng graph few-Shot learning via In-cOntext LearNing to address these challenges. Our model reframes graph few-shot learning as a fine-tuning-free sequence reasoning problem. At its core is a context-aware network that initializes nodes with role embeddings and employs a dual-context fusion module to synergistically integrate local topological structures and global task-level dependencies. This allows our model to dynamically generate class-aware representations for the query set conditioned on the support set context in a single forward pass. To effectively train our model, we introduce an unsupervised task generator that creates structure-adaptive features and constructs diverse pseudo-tasks from abundant unlabeled data. Our method unifies unsupervised meta-learning with graph in-context learning, achieving efficient inference. Extensive experiments on multiple benchmark datasets demonstrate the superiority of our model. Our public code can be found
Authors: Alexander Mihalcea
Abstract: Existing language model applications struggle to meet the demand for emotionally oriented support, primarily due to their inability to maintain deep, persistent context across sessions. This report introduces Psych LM, an iOS application that validates the thesis that, for such applications, the surrounding architecture is paramount. Psych LM runs a local, on-device language model within a purpose-built, local-first runtime designed for behavioral and life-coaching applications. The system achieves the practical effect of a near-infinite context window through an automated, user-inspectable memory corpus that converts conversations into structured memory cards, including facts, goals, and events, and dynamically injects them into the prompt via semantic and vector search. As such, the system can be defined as an active-learning, retrieval-augmented generative, on-device architecture. This architecture delivers four primary contributions: a local-first design where privacy is a core property; a detailed description of the memory corpus for persistent context of key user information; a deterministic orchestration layer that provides a stable behavioral spine independent of the model's internal state; and a benchmark framework focused on evaluating the integrated system's reliability under realistic operating conditions. The R and D process confirms that complex, context-aware interaction can be reliably achieved under the strict constraints of a mobile environment by prioritizing architectural control and resource management over simple model size.
Authors: Junlan Feng, Fanyu Meng, Chong Long, Pengyu Cong, Duqing Wang, Yan Zheng, Yuyao Zhang, Xuanchang Gao, Ye Yuan, Yunfei Ma, Zhijie Ren, Fan Yang, Na Wu, Di Jin, Chao Deng
Abstract: We introduce JT-Safe-V2, a large language model designed to advance the safety and trustworthiness of foundation models, extending our previous JT-Safe model toward a more comprehensive safety-by-design paradigm. JT-Safe-V2 emphasizes the joint optimization of general intelligence and safety-by-design through several key innovations: enriching pre-training data with contextual world knowledge, high-certainty pre-training procedures, and safety strengthening post-training mechanisms for enterprise-oriented agentic capabilities. Building on these safety-enhanced foundation models, we propose Safe-MoMA (Safe Mixture of Models and Agents), a framework that enables traceable and efficient inference through the orchestrated deployment of multiple models and agents. Extensive evaluations demonstrate that JT-Safe-V2 achieves state-of-the-art performance across both general intelligence and safety benchmarks. Moreover, Safe-MoMA reduces inference costs by more than 30\% compared to using the largest standalone model baseline while maintaining comparable performance. To facilitate future research on safety-by-design foundation models, we publicly release the post-trained JT-Safe-V2-35B model checkpoint.
Authors: Yuheng Jing, Kai Li, Ziwen Zhang, Jiajun Zhang, Zeyao Ma, Jiaxi Yang, Lei Zhang, Zhe Wu, Jinmin He, Junliang Xing, Jian Cheng
Abstract: In-Context Reinforcement Learning (ICRL) has enabled foundation agents to adapt instantaneously to novel tasks, yet its efficacy in Ad-Hoc Teamwork (AHT)-where coordination with unknown partners is required-remains unexplored. To rigorously evaluate this, we introduce a large-scale benchmark ICRL4AHT, built upon a high-throughput JAX implementation of Overcooked-V2. Our benchmark includes a large, diverse teammate suite spanning both RL and heuristic policies, enabling controlled train-test shifts, and provides a reproducible end-to-end pipeline for teammate generation, learning-history collection, dataset construction, and online multi-episode evaluation. We evaluate representative history-conditioned ICRL algorithms, including Algorithm Distillation (AD) and Decision-Pretrained Transformer (DPT), across millions of transitions. Results reveal notable limitations: contrary to their success in single-agent domains, these baselines fail to exhibit robust test-time adaptation in multi-agent settings. Specifically, these methods frequently underperform random baselines across both unseen teammate and unseen layout tracks, with no clear in-context improvement over long horizons. These findings highlight the challenges of strategic inference under partial observability within the OvercookedV2 AHT protocol, establishing our benchmark as a critical testbed for next-generation coordination algorithms.
Authors: Yuyang Hu, Hongjin Qian, Shuting Wang, Jiongnan Liu, Ziliang Zhao, Jiejun Tan, Zheng Liu, Zhicheng Dou
Abstract: Long-horizon agentic reasoning requires large language models to act over long interaction histories containing thoughts, tool calls, observations, and partial conclusions. The challenge is not merely that these histories grow long, but that information needed for the current decision may be scattered across distant steps and only become relevant later. Existing approaches address this difficulty by truncating the interaction history, compressing it into shorter surrogates, or retrieving selected parts of it for reuse, but they do not explicitly model how access to past interaction should adapt to the agent's evolving state. We instead cast long-horizon reasoning as a problem of state-adaptive memory. To this end, we propose State-Adaptive Memory~(SAM), a standalone framework that consolidates ongoing interaction into compact memory cues while preserving raw trajectory pages for intent-driven recall. These cues are not treated as replacements for history; rather, they serve as lightweight handles that allow the agent to reconstruct temporally distant information according to its current needs, without retraining the underlying backbone. We further optimize the memory module through expert-guided supervision and reinforcement learning, aligning it with trajectory-level utility. Across BrowseComp, BrowseComp-ZH, WideSearch, and HLE, SAM consistently outperforms strong baselines over diverse agent backbones. Our results suggest that explicit memory modeling provides a simple and effective foundation for long-horizon agentic reasoning.
Authors: Rongsheng Chen, Changliang Zhou, Canhong Yu, Yuanyao Chen, Yu Zhou, Zhuo Chen, Zhenkun Wang
Abstract: Generalist neural routing solvers have shown great potential in solving diverse vehicle routing problems (VRPs) with a unified model. However, existing solvers are typically limited to symmetric settings or degrade in performance when switching to asymmetric settings due to input inconsistencies or inherent structural differences, substantially limiting their practicality in real-world scenarios that encompass both scenarios. To address this limitation, we define the spatial position of each node based on the relative distances to a specific set of pivots and further propose a Spatial Pivot-Aligned Coordinate-free Embedding (SPACE) framework that unifies node representation and solution generation across symmetric and asymmetric VRPs. Specifically, we construct a bidirectional Frechet representation using a novel furthest pivot sampling strategy to enable invariant node representations across distinct problem settings. Furthermore, we introduce a weight-decomposed adaptive decoding mechanism that decouples geometric perception from problem representations, mitigating the overfitting of constraint decisions to a specific geometry setting. Extensive experiments on 110 VRP variants, comprising 55 symmetric problems and their asymmetric counterparts, demonstrate that SPACE achieves promising zero-shot generalization in both symmetric and asymmetric VRPs.
Authors: Yuyang Hu, Hongjin Qian, Shuting Wang, Jiongnan Liu, Tong Zhao, Xiaoxi Li, Zheng Liu, Zhicheng Dou
Abstract: Recent progress on long-horizon agentic tasks has been driven largely by scaling up individual agents through stronger models, better tools, and more effective scaffolding. In contrast, much less is understood about scaling out: whether multiple peer agents, all targeting the same task, can become an additional source of capability without relying on explicit role specialization or workflow orchestration. We study this question and propose AgentFugue, a collective reasoning framework built around a shared reasoning hub. As peer agents explore the same task in parallel, the hub records concise notes on what each agent has established, attempted, or ruled out, and enables each agent to selectively access what other agents have discovered in a form useful for its current search. This design turns otherwise isolated trajectories into a connected ecology of reusable intermediate reasoning without requiring centralized planning. We instantiate the hub as a plug-in communication layer, trained with supervised fine-tuning and end-to-end reinforcement learning. Across the challenging long-horizon settings we study, AgentFugue improves over strong baselines. Our results suggest that collective reasoning can turn scaling out peer agent systems into a distinct source of capability gains, rather than merely a way of spending more compute.
Authors: Yuhang Zhang, Keyan Ding, Peilin Chen, Han Liu, Can Lin, Ruixi Chen, Shiqi Wang, Qi Song
Abstract: Enzyme-reaction retrieval is a fundamental problem in computational biology, underpinning enzyme characterization, reaction mechanism elucidation, and the rational design of metabolic pathways and biocatalysts. As a bidirectional task, it entails both enzyme-to-reaction and reaction-to-enzyme mapping. However, existing approaches suffer from poor generalization across tasks and distributions, with performance highly sensitive to dataset splits and substantial asymmetry between retrieval directions. To address these challenges, we present TIGER, a Text-Informed Generalized Enzyme-Reaction Retrieval framework that leverages protein-to-text generation models to distill textual semantic knowledge from enzyme sequences, providing a generalized representation that bridges enzymes and biochemical reactions. To ensure the quality and reliability of textual semantics, we design a Dynamic Gating Network that adaptively fuses text-derived knowledge with sequence features, enabling more consistent and informative enzyme representations, while a Structure-Shared Feature Projector aligns enzyme and reaction representations within a unified latent space. Extensive experiments demonstrate that, under bidirectional retrieval supervision, TIGER significantly outperforms state-of-the-art baselines across diverse distributions and exhibits strong robustness and transferability across tasks.
Authors: Yunhua Pei, Zerui Ge, Jin Zheng, John Cartlidge
Abstract: Multi-agent LLM decision systems for portfolio management still lack a principled way to assign credit across specialist agents, remain vulnerable to cold-start dominance under regime shifts, and offer limited transparency into how final allocations are formed. We propose Market Regime Council (MRC), a cooperative multi-agent decision system that computes exact Shapley credits across all single, pairwise, and Grand-coalition outputs for online agent weighting. Instantiated with N=3 specialist agents, at each trading period, MRC recomputes coalition-based Shapley weights from exponentially weighted performance histories, uses a Bayesian adaptive mixture to stabilize early periods, applies regime-dependent multipliers to adjust agent authority, and records each rebalance through a five-layer causal trace. Over 1,037 trading days across 13 crypto assets and five seeds, MRC achieves a Sharpe ratio of 1.51 and a cumulative return of 440.1%, ranking first on CR, SR, and IR among active baselines and attaining the lowest MDD among active methods. Ablation results show that the gains come from Shapley-weighted integration across coalition outputs rather than from any single stage in isolation. Code and demo data are included in the supplementary material.
Authors: Jianan Li, Simeng Qin, Xiaojun Jia, Lionel Z. Wang, Tianhang Zheng, Xiaoshuang Jia, Yang Liu, Xiaochun Cao
Abstract: Large Reasoning Models (LRMs) have demonstrated remarkable capabilities in reasoning and generation tasks and are increasingly deployed in real-world applications. However, their explicit chain-of-thought (CoT) mechanism introduces new security risks, making them particularly vulnerable to jailbreak attacks. Existing approaches often rely on static CoT templates to elicit harmful outputs, but such fixed designs suffer from limited diversity, adaptability, and effectiveness. To overcome these limitations, we propose an adaptive evolutionary CoT jailbreak framework, called AE-CoT. Specifically, the method first rewrites harmful goals into mild prompts with teacher role-play and decomposes them into semantically coherent reasoning fragments to construct a pool of CoT jailbreak candidates. Then, within a structured representation space, we perform multi-generation evolutionary search, where candidate diversity is expanded through fragment-level crossover and a mutation strategy with an adaptive mutation-rate control mechanism. An independent scoring model provides graded harmfulness evaluations, and high-scoring candidates are further enhanced with a harmful CoT template to induce more destructive generations. Extensive experiments across multiple models and datasets demonstrate the effectiveness of the proposed AE-CoT, consistently outperforming state-of-the-art jailbreak methods.
Authors: Jeffrey Qin, Wasu Top Piriyakulki, Zhuangfei Gao, Mia Radovanovic, Jessica Sommerville, Kevin Ellis, Marta Kryven
Abstract: Real world decision-making requires constructing mental models under uncertainty over evidence, over the underlying causal rules, and over the state of the world itself. Which computational principles underpin human inference under such conditions, and do LLM-based agents exhibit similar behavior given matching constraints? We address these questions using an inductive inference Box Task in which participants, human children and LLM-based agents, infer a latent cause through sequential interaction with an uncertain environment. We formalize this task as program induction with Bayesian particle-based inference, admitting two complementary interpretations: (1) as a constraint satisfaction process over hypotheses, and (2) as a program synthesis problem in which hypotheses are executable programs evaluated against evidence. Using the constraint-based formulation, we show that children's behavior is best explained by a combination of subjective evidence reliability and online hypothesis generation, accounting for both their evidence-seeking patterns and their dissociation between task completion and rule generalization. Using the program synthesis formulation, we treat LLM-based agents as model organisms: controllable systems that allow systematic manipulation of task conditions. Across backends, LLM-based agents replicate children's responses to changes in evidence reliability and observability, including discounting unreliable evidence, seeking to resolve partial information, and dissociating between task completion and causal generalization. At the same time, LLM-based agents tend to over-observe and over-comply with instructions relative to children. These results suggest that while children and LLM-based agents adapt similarly to environmental structure, their information-seeking behavior exhibits distinct underlying costs and inductive biases.
Authors: Lirong Che, Yuzhe yang, Peiwen lin, Chuang wang, Xueqian wang, Jian su
Abstract: Agent harness evolution improves frozen language-model agents by modifying the executable structures around them. We study this paradigm as a form of sample-efficient fast adaptation: instead of updating model weights, an agent can acquire task-specific competence by changing its external harness, while leaving the base model's general capabilities intact. Prior work shows that self-generated rollouts can support harness search, suggesting that agents may acquire new task competence through practice. Yet in long-horizon stochastic environments, self-practice becomes fragile: rewards are sparse, outcomes are high-variance, and failures are hard to attribute to concrete harness mechanisms. We introduce DemoEvolve, a demonstration-bootstrapped approach to harness evolution. When reward-only search is too broad and noisy, competent human trajectories serve as expert reference experience for the coding proposer, guiding harness-level diagnosis and editing. Experiments on Liar's Dice show that self-rollout evolution can work when episodes are short and failures are attributable. In contrast, Balatro exposes a harder long-horizon stochastic regime, where self-rollout evolution is misled by sparse feedback and candidate-selection noise, while tutorial-like textual knowledge alone does not yield stable improvement. Under the same limited budget, DemoEvolve produces more effective and auditable harness edits and achieves better performance. Overall, demonstrations make sparse-feedback harness evolution more diagnosable, localizable, and stable.
Authors: Ninglin Ou, Mohammad A. Razzaque, Iftekher Islam Shovon, Shafkat Khan Siam, Shafiuzzaman K Khadem, Krishnendu Guha, Mayeen U Khandaker, Md. Noor-A-Rahim
Abstract: The rapid growth of Electric Vehicle (EV) adoption challenges power distribution networks through peak load spikes, voltage instability, and transformer overloads from uncoordinated charging. While Model Predictive Control (MPC) and standard Reinforcement Learning (RL) methods have addressed these issues, existing approaches rarely treat real-time carbon intensity or fluctuating renewable energy (RE) availability as primary scheduling objectives, leaving substantial decarbonisation potential unrealised. This paper proposes an emission-aware RL strategy based on the Soft Actor Critic (SAC) algorithm, with a multi-objective reward that penalises carbon emissions, curtailed on-site renewables, and unmet user demand. The agent is trained within a unified benchmarking framework on the EV2Gym platform, incorporating behind-the-meter solar and wind profiles, time-varying EirGrid carbon intensity data, and realistic workplace EV behaviour across 25 Electric Vehicle Supply Equipment (EVSE) units. Nine control strategies, including heuristics, emission-aware MPC variants, and the proposed RL agent, are compared under five renewable penetration scenarios (0%-50%) over ten independent runs each. The RL agent achieves a carbon intensity as low as 23.96 grams of carbon dioxide per kilowatt-hour under 50% wind penetration, representing up to 87% emission reduction versus the uncontrolled baseline, and outperforms the external graph-based Power Distribution Network (PDN) benchmark. Transformer overload remains below 7 kWh across scenarios, against up to 1093 kWh for the As Fast As Possible (AFAP) heuristic, and renewable self-consumption reaches 52% under combined wind and solar supply. Embedding carbon intensity forecasts into the RL state and reward aligns charging with low-emission periods while preserving grid compliance and user satisfaction.
Authors: Anton Antonov, Humam Kourani, Alessandro Berti, Gyunam Park
Abstract: Process modeling is a sub-domain of Business Process Management (BPM) focused on the translation of process artifacts into formal models. This task traditionally requires extensive human input and domain expertise in both BPM notations and the specific business context. While Large Language Models (LLMs) can now automate much of this manual work, current text-to-model approaches focus predominantly on the control-flow perspective-ordering activities without considering the collaborative aspect of the processes. In this paper, we introduce a resource-aware generation pipeline that produces formal BPMN 2.0 collaboration diagrams from natural-language descriptions. Rather than solely prompting an LLM for raw XML, we describe a compact, executable intermediate language with mandatory resource details defining both the organization (pool) and the role (lane). Cross-organization dependencies are materialized using the standard formal notation for such interactions-message events-while an orthogonal layout routine automatically handles the spatial arrangement of elements within pools and lanes. Experiments on ten business processes with nine LLMs show strong resource discovery while preserving control-flow quality and adding only marginal runtime overhead. This approach moves generative modeling toward a more comprehensive, multi-collaborative representation of business operations.
Authors: Mustafa Hayri Bilgin, Mariam Barry, Albert Bifet, Azzedine Idir Ait Said, Soumya Banerjee
Abstract: Efficiently updating Large Language Models (LLMs) with new or evolving factual knowledge remains a central challenge, as even parameter-efficient adaptation can erode previously acquired reasoning abilities. This tension reflects a plasticity-stability dilemma: models must incorporate new knowledge while preserving skill-critical representations. In this work, we study this trade-off through the spectral structure of multilayer perceptron weight matrices. We show, both theoretically and empirically, that information essential for reasoning is not localized only in dominant singular directions, but is instead distributed across the singular spectrum. Motivated by this observation, we introduce PALoRA, a two-stage framework for knowledge injection with reduced interference. PALoRA first trains a Singular Value Fine-Tuning (SVF) expert on a reasoning dataset and uses its learned singular scaling vector as a frozen geometric probe to identify components that are critical for the target skill. It then performs factual knowledge injection with Low-Rank Adaptation (LoRA) under a structural orthogonality constraint, ensuring that updates avoid the identified skill-relevant subspace. Across Llama 3.1 8B and Mistral 7B, and across mathematical, coding, and scientific reasoning benchmarks, PALoRA preserves on average 95% of the SVF expert's reasoning performance while maintaining competitive factual recall. It consistently improves skill retention over prior spectral Parameter-Efficient Fine-Tuning (PEFT) methods while adding less than 0.006% parameter overhead.
Authors: Seokil Ham, Jaehyuk Jang, Wonjun Lee, Changick Kim
Abstract: Fine-tuning-as-a-Service (FaaS) enables personalization of large language models (LLMs), but it can weaken safety-alignment under harmful fine-tuning attacks. Recent work has shown that activating harmful-behavior modules during fine-tuning can prevent models from learning undesired behaviors, but its mechanism remains unclear. In this paper, we revisit temporary jailbreaking as a defense against harmful fine-tuning and provide a gradient-level analysis showing that it saturates safety-degrading gradients while preserving benign task-relevant gradients. Based on this insight, we propose a Buffer-and-Reinforce fine-tuning framework that buffers harmful updates during user fine-tuning and reinforces safety after adaptation. Specifically, BufferLoRA induces temporary jailbreaking as a removable adapter to reduce harmful updates during user fine-tuning. After adaptation, ReinforceLoRA, trained to recover refusal behavior under the temporarily jailbroken state, is integrated with UserLoRA via QR decomposition-based merging to reinforce safety while preserving user-task performance. Extensive experiments show that our framework achieves superior safety and utility with no additional safety data during user fine-tuning and minimal computational cost.
Authors: Weixian Waylon Li, Mengyu Wang, Tiejun Ma
Abstract: Backtesting large language models (LLMs) on historical financial data is unreliable because pre-training cuts off after the events happened. An LLM trained in 2024 already "knows" which way 2018-2020 stocks moved. We name this failure parametric look-ahead bias and propose FinCAD, an inference-time adaptation of Context-Aware Decoding that suppresses an LLM's memory of historical outcomes without retraining. FinCAD pairs an adversarial bias-discovery pipeline that learns a model-specific memory-activating prior prompt with an entity- and date-adaptive rule that scales the CAD strength to per-(entity, date) memorisation, so the penalty fires on memorised in-sample dates and decays to zero out-of-sample. Across five 7-14B LLMs and five mega-cap equities, FinCAD cuts in-sample backtest returns by up to -67.1% on memorised dates while leaving 2025 out-of-sample returns within $8K and Sharpe within 0.10 of baseline, and preserves general-purpose reasoning within 1.7 pts. On an eleven-model leaderboard, it raises the in-sample / out-of-sample Spearman correlation from +0.779 to +0.846, recovering rankings that genuinely predict out-of-sample performance.
Authors: Elias Stenhede, Eivind Bj{\o}rkan Orstad, Torbj{\o}rn Omland, Henrik Schirmer, Arian Ranjbar
Abstract: Artificial intelligence-enabled electrocardiography (AI-ECG) can detect heart failure (HF), including disease not captured by left ventricular ejection fraction (LVEF), but the cardiac phenotypes underlying model predictions remain unclear. We therefore investigated whether AI-ECG-predicted HF risk aligns with established echocardiographic measures of myocardial dysfunction, remodelling, and filling pressures. We retrospectively analysed ECG and echocardiography data from 8147 patients who underwent both examinations within three days at Akershus University Hospital between 1 January 2023 and 1 June 2025. A previously validated AI-ECG model for HF detection was applied to all ECGs. Spearman's rank correlation $\rho$ quantified associations between echocardiographic parameters and AI-ECG risk. Subgroup analyses were performed by sex and left ventricular ejection fraction (LVEF). External validation included 36,286 ECG-echocardiography pairs from Columbia University Irving Medical Center. Global longitudinal strain (GLS) showed the strongest correlation ($\rho$=0.57), followed by mitral annular plane systolic excursion (MAPSE) ($\rho$=-0.49) and LVEF ($\rho$=-0.45). In patients with LVEF>50%, correlations remained substantial for GLS, MAPSE, and diastolic-related parameters. Volumetric left ventricular indices correlated less strongly in women, whereas diastolic indices showed stronger correlations in women than in men. Physiological validation showed that AI-ECG HF risk predictions align primarily with measures of systolic function, particularly global longitudinal strain, while also capturing diastolic-related abnormalities in patients with preserved LVEF. This approach may improve clinical interpretability and identify opportunities for model refinement.
Authors: Shubham Gupta, Nikhil Panwar, Partha Pratim Roy
Abstract: While Deep Learning (DL) enhances automated electrocardiogram (ECG) analysis, clinical deployment is hindered by class imbalance and the generalization gap. This paper presents HeartBeatAI, a deep learning framework combining domain generalization, multi-scale feature aggregation, and clinical explainability for robust 12-lead ECG classification. Moving beyond image-based paradigms, HeartBeatAI integrates a Squeeze-and-Excitation (SE) ResNet to isolate diagnostic leads alongside a Multi-Layer Concentration Pipeline to capture macro-rhythm and micro-morphological anomalies. To mitigate domain shift, the framework employs MixStyle regularization and Label Smoothing. Rigorous benchmarking across four large-scale datasets using intra-source and Leave-One-Domain-Out (LODO) protocols demonstrates high performance (98% Macro F1-score) under intra-source conditions. However, LODO evaluations reveal significant degradation in detecting rare anomalies, highlighting a persistent challenge in cross-institutional deployment.
Authors: Andreas Opedal, Francesco Ignazio Re, Abulhair Saparov, Mrinmaya Sachan, Bernhard Sch\"olkopf, Ryan Cotterell
Abstract: Many applications of large language models (LLMs) require deductive reasoning, yet models frequently produce incorrect or redundant inference steps. We frame natural language inference as a search problem where the final answer is the valid proof itself, requiring a reasoning procedure in which intermediate inferences are correct. Specifically, we investigate whether LLMs can learn to generate correct and efficient proofs with guidance from A* search -- an algorithm that guarantees an optimally efficient path to a goal. We explore two training techniques: supervised fine-tuning on execution traces from A* and reinforcement learning with A*-informed process reward models. Empirically, we find that Llama-3.2 models in the 1B--3B range benefit substantially from A* post training, going from near-zero accuracy to outperforming DeepSeek-V3.2 -- a much larger model. Our analysis uncovers a trade-off: while simple correctness rewards maximize accuracy, A*-informed signals strike a balance between accuracy and efficiency. Furthermore, we find that on larger search spaces, models trained with imperfect heuristics exhibit superior accuracy. Our results demonstrate a promising direction towards reasoning guided by principles derived from classical search algorithms.
Authors: Yuxin Zhang, Mengxue Hu, Zheng Lin, Xiaoyi Fan, Fan Xie, Zihan Fang, Jing Yang, Wenjun Zhu, Zhiwen Chen, Chengfei Lv, Zhe Chen
Abstract: Large language model (LLM) agents excel at solving complex long-horizon tasks through autonomous interaction with environments. However, their real-world deployment faces a fundamental device--cloud dilemma: on-device models are efficient but often brittle, while cloud models are stronger but costly in computation. State-of-the-art LLM device--cloud routers usually make coarse task-level decisions, which cannot adapt to the changing difficulty of multi-step agent interactions. To address this issue, we present Hera, a step-level device--cloud LLM agent coordinator for long-horizon tasks achieving a strong performance--cost Pareto frontier. Hera adopts a novel two-stage training paradigm: (1) imitation learning for cold-start, followed by (2) reinforcement learning that jointly optimizes task success and cloud usage efficiency. The first stage casts step-level routing as a supervised classification problem: the device agent is replayed on cloud trajectories, with each state labeled by the agreement between device and cloud actions. In the second stage, we perform cost-aware reinforcement learning by grouping identical states across trajectories and updating Hera with labels favoring higher expected return and fewer future cloud calls. We evaluate Hera on ALFWorld, WebShop, and AppWorld, where it consistently outperforms prior methods, achieving 92.5% of the cloud-only success rate with cloud use in only 46.3% of steps.
Authors: Zhimin Lin, Kun Cheng, Fan Bai, Jie Gao
Abstract: Large language models (LLMs) are increasingly used for qualitative data analysis (QDA), yet their outputs often miss the depth and nuance of human analysis. We argue this gap reflects a missing credibility practice from human QDA: peer debriefing, in which an analyst seeks feedback from a disinterested peer and uses it to refine their coding. To bring this practice into LLM-assisted QDA, we propose Agent-as-Peer-Debriefer, a multi-agent QDA framework that builds peer debriefing into key coding steps. In our framework, a Hierarchical Coding Agent follows the standard QDA process to generate codes, sub-themes, and themes, along with self-explanations and reflection memos. It then shares these outputs with three Peer-Debriefing Agents, each applying a distinct analytical perspective (Theory-Driven, Data-Driven, or Applied) and refining the codes by keeping, renaming, reassigning, merging, or splitting them. These perspectives are drawn from established human QDA practices that generalize across domains and datasets. To evaluate the framework, we test it on three datasets across two domains with three LLMs, measuring semantic similarity to human-annotated codes. Across all settings, perspective-based, peer-debriefing refinement aligns more closely with human codes than a single-LLM baseline, and an ablation further shows the gain is not merely from additional refinement. The three perspectives also produce distinct trade-offs, showing that the choice of perspective is a meaningful and controllable design decision. More broadly, these findings suggest that simulating peer debriefing with explicit perspectives is a promising route to more credible LLM-assisted QDA.
Authors: Gustavo (Jesus), Angulo
Abstract: We develop a rigorous algebraic framework for deep convolutional architectures, CNNs, ResNets, and encoder--decoder networks such as UNet, grounded in lattice theory and mathematical morphology. The central tool is the Matheron--Maragos--Banon--Barrera (MMBB) universal representation theory for translation-invariant operators, which we apply systematically to every layer of a standard deep network. The principal finding is that the standard CNN pipeline (linear convolution~$+$ ReLU~$+$ flat max-pooling) is a cross-lattice operator: the convolution is an erosion in the Fourier inf-semilattice while ReLU is a lattice-join closing and max-pooling is a dilation in the pointwise max-plus lattice, and their composition is a morphological opening in neither. A second finding is that the upper adjoint of ReLU in the pointwise lattice is a global (non-local) operator, the identity on globally non-negative functions and $-\infty$ otherwise, so no local morphological erosion can form an adjunction pair with ReLU. These two results together provide the precise algebraic reason why depth in standard CNNs introduces genuine representational power: the composed layer is not idempotent. Three layer designs that are genuine idempotent openings are identified and fully characterised: the pure max-plus morphological layer (pointwise lattice), the spectral Wiener layer (Fourier lattice), and the self-dual morphological layer. We establish a complete fixed-point and convergence theory. The framework also unifies max-pooling, strided convolution, and the Laplacian pyramid under the Goutsias--Heijmans adjoint pyramid theory, and gives the Activation--Pooling Dilation (APD) factorisation with its correct adjoint.
Authors: Junjie Zhao, Jingyi Liang, Zhenyang Cai, Jiaming Zhang, Zhenwei Wen, Shuzhi Deng, Wenjing Yi, Chunfeng Luo, Hexian Zhang, Junying Chen, Tianrui Liu, Zhuhui Bai, Zixu Zhang, Pradeep Singh, Xiang Liu, Jianquan Li, Nhan L Tran, Falk Schwendicke, Zuolin Jin, Lijian Jin, Liangyi Chen, Wei-fa Yang, Benyou Wang, Junwen Wang, Shan Jiang
Abstract: While large language models (LLMs) hold transformative potential for medicine, their reasoning robustness and safety in real-world clinical scenarios remain critically underexplored, particularly in dentistry. Here we introduce GlobalDentBench, the first multinational dental benchmark, featuring a taxonomy that encompasses 14 dental specialties across 88 countries and regions spanning six continents. The benchmark comprises 8,978 expert-validated questions across three formats (multiple-choice, short-answer, and case-based questions) and assesses three progressive reasoning levels: knowledge recall (L1), routine reasoning (L2), and individualized reasoning (L3). To ensure data quality, the automated construction framework was calibrated by six senior dentists, achieving expert agreement rates of 99.98% for multiple-choice and short-answer questions and 96.78% for the more complex case-based questions. Evaluation of 12 frontier LLMs on GlobalDentBench revealed a sharp, stepwise performance degradation with increasing reasoning complexity. Specifically, accuracy plummeted from 81.34% on multiple-choice to 64.53% on short-answer and 22.34% on case-based questions, while declining markedly from 74.01% at L1 to 55.64% at L2 and 35.71% at L3. More critically, risk analysis of real-world dental cases demonstrated an alarming overall unsafe rate of 31.01% in LLM-generated clinical recommendations, with 4.51% posing risks of irreversible patient harm and risks particularly pronounced in specialties such as orthodontics. These findings expose fundamental limitations in the medical reasoning and safety of current LLMs. Consequently, GlobalDentBench provides a scalable foundation for trustworthy clinical AI evaluation, underscoring the urgent need for rigorous validation before the safe deployment of these models in healthcare.
Authors: Jialiang Yang, Bin Xia, Ruihang Chu, Dingdong Wang, Wanke Xia, Zhun Mou, Tianyang Zhong, Yiting Zhao, Wenming Yang
Abstract: Rapid advances in audio-video (AV) generation have enabled high-fidelity synthesis with synchronized sound, particularly for human-related scenarios involving speech and interactions. Yet evaluation for AV generation remains at an early stage, with only a few coarse-grained benchmarks for human-related scenarios and relying on limited preset evaluations with generic multimodal LLMs, leading to inaccurate assessments of model capabilities. To address these issues, we introduce AVBench, a fully automated benchmark tailored for human-centric AV generation. AVBench is built on two key designs for comprehensive and accurate evaluation: (i) Human-centric and fine-grained metrics. AVBench integrates ten evaluation dimensions designed for human-centered real-world scenarios, covering visual quality, audio quality, and multi-level consistency across modalities. These practical metrics capture human-related details that existing benchmarks often overlook. (ii) Specialized evaluators via preference learning. To address the lack of specialized training data, we construct large-scale supervision by transforming real-world videos into diverse training pairs with controlled perturbations. After fine-tuning on this high-quality dataset, the evaluators learn to reliably detect subtle cross-modal inconsistencies. Crucially, instead of producing discrete textual judgment, AVBench derives continuous evaluation scores from the model's prediction confidence on binary decisions. This probabilistic scoring mechanism enables a more reliable assessment than traditional VQA-style evaluation and aligns closely with human judgment. Taken together, AVBench offers automated evaluation for AV generation, demonstrates strong potential for data filtering, and serves as a differentiable reward signal for Reinforcement Learning from Human Feedback (RLHF).
Authors: Simon Dennis, Kevin Shabahang, Hao Guo, Rivaan Patil
Abstract: Major LLM platforms deploy models in an inference-only configuration: the model serves requests but never updates per-user weights. Users must repeatedly re-teach preferences, corrections, and project context, and context-based workarounds consume context-window space and degrade under cascading compaction. We evaluate an alternative: nightly consolidation of interaction knowledge into model weights via reflection, synthesis, and Low-Rank Adaptation (LoRA) fine-tuning on a single consumer GPU. Across ten realistic software development conversations (n = 10, 1,146 test questions across three memory types), three cycles of cascading compaction retain 36.8 +/- 3.0% of knowledge (between an 11.8% no-context floor and a 90.1% full-context ceiling), while consolidation retains 80.4 +/- 1.3% -- a 43.6 pp gain (paired t(9) = 14.8, p < 0.001) that more than doubles what compaction preserves, with the largest gains on procedural corrections (36.3% -> 74.6%) and episodic project facts (31.5% -> 78.2%). As a methodological aside, mean per-token validation cross-entropy is negatively correlated with LLM-judged accuracy (r = -0.51) while median per-token validation cross-entropy tracks accuracy almost exactly (r = +0.99): under evaluators that tolerate surface-form variation, the mean is misleading and a heavy-tail-robust statistic is the faithful signal. Persistent personalization requires moving beyond inference-only deployment toward architectures that consolidate knowledge into weights.
Authors: Ali \c{S}enol, Garima Agrawal, Huan Liu
Abstract: LLMs have achieved remarkable success in complex reasoning tasks, yet current evaluation approaches predominantly rely on final-answer correctness, offering limited insight into the underlying reasoning processes that produce those answers. To address this gap, this study proposes a unified multi-dimensional framework for measuring reasoning quality in LLMs from a behavioral perspective, operationalizing six theoretically grounded dimensions: Correctness (CQ), Consistency (CS), Robustness (RS), Logical Coherence (LS), Efficiency (ES), and Stability (SS). Extensive experiments on seven LLMs across 975 items from four benchmarks demonstrate that the framework reveals behaviors invisible to accuracy-only metrics. Notably, logical coherence is orthogonal to correctness (r = -0.172, ns), confirming that correct answers can arise from incoherent reasoning, while Claude-Haiku-4.5 achieves the highest multi-dimensional score (Q_bal = 0.778). Furthermore, the framework exposes critical ranking inversions: DeepSeek-V3 ranks second under accuracy-priority but fifth under legal/compliance weighting, a reversal that single-metric evaluation cannot detect. Discriminant validity confirms 11/15 dimension pairs are independent (|r| < 0.50), providing psychometric support for treating each dimension as a distinct signal. The dimensional profiles produced by the framework directly support three classes of deployment decision: identifying models whose reasoning traces would fail accountability audits despite correct final answers (LS--CQ orthogonality); preventing ranking errors caused by accuracy-only benchmarking; and ensuring that no single metric silently substitutes for the six independent signals the framework captures.
Authors: Hao Guo, Simon Dennis, Rivaan Patil, Kevin Shabahang
Abstract: Mean cross-entropy is the standard validation metric for language models, but it can fail to track model quality during training. We examine this in two common scenarios. First, in Qwen2.5-1.5B SFT on synthetic fact-learning, we find that mean CE rises substantially after the initial learning phase while held-out fact-recall accuracy remains near its peak. Second, we find that in top-K distillation on TinyStories, decreasing K improves median CE while worsening mean CE; the Top-5 student attains the highest LLM-judge score and crosses below its teacher on median CE, despite having the worst mean CE. In both cases, median CE correlates much more closely with task performance than does mean CE. Analyzing how bulk and tail percentile CE move during training reveals that training reshapes the empirical per-token CE distribution. In top-K distillation, smaller K yields a distribution with more mass at both extremes, decreasing the median and increasing the mean. In Qwen SFT, the bulk saturates quickly while the tail extends in the latter half of training. In both, the task-evaluation metric appears more sensitive to the bulk than to the tail. Practically, we recommend reporting a small set of percentile CE summaries alongside the mean, and using concordance among them as a tool to keep track of distribution reshaping, as well as a low-cost diagnostic for when mean and median CE disagree on model selection.
Authors: Vassilis Lyberatos, Edmund G. Dervakos, Eleni Adamidi, Athanasios Voulodimos, Giorgos Stamou
Abstract: Speech and language technologies offer valuable opportunities for supporting mental health assessment through objective and interpretable cues. We present a systematic feature-based analysis framework leveraging perceptually grounded acoustic and linguistic characteristics, including prosody, vocal quality, semantic coherence, syntactic structure, and sarcasm. Using statistical analysis and interpretable machine learning (XGBoost with SHAP and LIME), we examine associations between speech features and validated symptom measures of depression, anxiety, and ADHD. Evaluated on both controlled benchmark datasets (StressID, DAIC-WOZ, Androids, EATD) and a real-world clinical dataset, the framework reveals stable and consistent relationships between symptom severity and vocal irregularities (e.g., shimmer, jitter), lexical-syntactic patterns, and affective tone. An ablation study conducted across all datasets further identifies the most informative feature groups. This work explores a transparent and clinically interpretable approach to speech-based mental health analysis.
Authors: Minghao Lv, Lu Chen, Enchang Zhang, Anji Zhou, Xiaoran Xue, Hanyi Zhang, Fenghua Tang, Zhuo Rachel Han, Mengyue Wu
Abstract: As large language models (LLMs) are increasingly integrated into emotionally sensitive domains, the structural integrity of their emotional intelligence (EI) becomes a critical frontier for safety and alignment. Current benchmarks often conflate superficial politeness with deep affective reasoning, failing to distinguish between perceptual accuracy and interactive efficacy. Here, we introduce FACET (Functional Affective Competence and Empathy Test), a psychometrically grounded framework comprising 480 expert-crafted items. Unlike previous metrics, FACET is theoretically anchored in the Mayer-Salovey-Caruso four-branch ability model, operationalizing EI through perception, facilitation, understanding, and management of emotions. Through an evaluation of nine frontier models (including GPT-5, Claude-Sonnet-4), we demonstrate that emotional intelligence is not a monolithic capability but is fragmented across cognitive and interactive dimensions. While frontier models demonstrate robust proficiency in objective emotion recognition and social reasoning, this does not consistently translate to interactive success. We categorize these discrepancies into three distinct performance profiles: cognitive-dominant, interactive-dominant, and context-dependent. These typologies indicate that emotional skills do not scale uniformly with general intelligence or model size; rather, they are shaped by specific alignment paradigms. Notably, we identify hidden emotion recognition as a universal performance bottleneck across all architectures. Our results suggest that current RLHF processes may optimize for "stochastic empathy", a statistical mimicry of emotional syntax, at the expense of integrated affective reasoning. These findings challenge the assumption of linear emotional scaling and provide a rigorous roadmap for developing socially aware agents capable of genuine clinical resonance.
Authors: Roberto Cruz, David Rey-Blanco
Abstract: Most reported gains on agentic-LLM clinical benchmarks are often attributed to prompt engineering, yet our results suggest that larger improvements can come from architectural and engine-level design. We present MDIA, a Multi-agent Diagnostic Intelligence Agent implemented as a 7-node specialty-routed clinical reasoning graph, on the full HealthBench Professional benchmark (n = 525), on a non-fine-tuned LLM. MDIA achieves 0.6272 under OpenAI's GPT-5.4-2026-03-05, which is +3.72 pp above the performance of OpenAI's ChatGPT for Clinicians. The experimental work shows that performance lift is attributable to system architecture: specialty routing, multi-turn context preservation, drug-state safety gating, site-filtered search, length-aware synthesis, and engine-level reliability. These findings support the view that agentic clinical benchmark performance is shaped both by the underlying foundation model and the orchestration architecture. Nevertheless, we also noticed notable differences when using other models as a grader; in particular, when using Gemini 2.5 Pro, MDIA scored 0.6585, which suggests that the choice of grader is a source of variability. Robust evaluation of LLMs would therefore require assessment across several independent grader models.
Authors: Atsushi Suzuki, Jing Wang
Abstract: While large-scale models such as LLMs and diffusion models have achieved practical success, public institutions have emphasized the importance of explainability in AI. Existing methods for explaining AI, however, are not designed to provide completely faithful explanations of the behavior of large-scale AI systems. Although a completely faithful and interpretable explanation of the behavior of an AI system might be useful for AI governance, it has not been known whether providing such an explanation is theoretically possible. In this paper, we mathematically prove a fundamental quadrilemma in explaining AI, stating that AI and its explanation cannot satisfy the following four conditions simultaneously: 1) the complexity of the operation environment, 2) the goodness of the AI's performance, 3) the interpretability of the AI's explanation, and 4) the complete faithfulness of the AI's explanation. This quadrilemma suggests that, in most applications where we cannot change the environment or sacrifice good AI performance and an interpretable explanation, we should give up complete faithfulness of explanations and should instead aim to explain only the parts that are important for applications. As a consequence, the quadrilemma implies that AI governance should be designed on the premise that the faithfulness of AI explanations is always incomplete.
Authors: Christopher Da Silva
Abstract: Foundation models can increasingly describe, reconstruct, and generate 3D objects, assemblies, scenes, and environments, but visually plausible spatial output is not yet operable 3D. A generated object or environment becomes useful to an agent only when the system can identify its entities, frames, surfaces, constraints, provenance, admissible actions, expected effects, and validation failures. This paper introduces Hylos, a systems architecture for contract-bounded spatial intelligence. Hylos maintains scene-scale operability state over objects, assemblies, assets, surface anchors, assertions, action candidates, solver jobs, shared actuator invocations, capability gaps, and effect diffs. Durable spatial changes are routed through a SpatialTransaction: a commit boundary that resolves references, checks admissibility, protects invariants, projects effects, and returns commit, review, rollback, deferral, or capability-gap outcomes. The paper is framed as a systems/position preprint with a focused artifact study rather than a broad benchmark. The study examines causal repair: a visible misalignment appears on a dependent component, while the supported repair lies upstream in the placement structure that controls it. The successful interaction traces the symptom through scene dependencies, selects a supported upstream interaction, and applies a validated change instead of directly editing visible geometry. The broader claim is that spatial AI should be evaluated not only by visual quality, but by whether generated or edited 3D can become reliable substrate for CAD, robotics, simulation, inspection, manufacturing, and interactive world authoring.
Authors: Feng Chen, Justin Tauscher, Changye Li, Meliha Yetisgen, Alex Cohen, Adam Kuczynski, Angelina Pei-Tzu Tsai, Benjamin Buck, Dror Ben-Zeev, Trevor Cohen
Abstract: Speech monologues recorded in naturalistic settings provide opportunities to characterize mental illness phenomenology and detect symptom exacerbation. Large language models (LLMs) offer new possibilities for automating this process, as they require annotated data primarily for evaluation rather than training. In this paper, we present a novel automated, multi-agent LLM pipeline for the fine-grained, multi-label extraction of language suggestive of delusional beliefs, associated affective responses, and behavioral responses from transcripts of naturalistic audio diaries collected from people with moderate persecutory ideation. Evaluating an ensemble of three foundation models, we demonstrate that detailed diagnostic prompt instructions successfully reduce false positives for delusional theme classification, but also constrain the interpretation of affective or behavioral responses. Furthermore, comparing multi-agent adjudication frameworks shows that complex conversational debate between agents diminishes accuracy on clinically ambiguous text by inducing premature consensus. Instead, majority voting establishes robust performance (Micro F1 of 0.872 and 0.779 for delusion detection and classification respectively). This work provides a validated and scalable pipeline for the automated detection and characterization of content suggesting delusional beliefs in naturalistic speech.
Authors: Suresh Raghu, Satwik Pandey, Shashwat Pandey
Abstract: Language-model agents increasingly emit uncertainty signals throughout a trajectory, but existing agentic UQ evaluations often conflate ranking usefulness with probabilistic truthfulness. AUROC, AUPRC, risk-coverage, Trajectory ECE, and scalarized trajectory scores evaluate discrimination, binwise calibration, or collapsed summaries, but do not strictly elicit the full prefix-conditioned success-probability trace $q_t = P^{\pi}(Y=1 | H_t)$. Building on prequential proper scoring, we introduce the Trajectory Proper Score (TPS), a predictor-agnostic family of strictly proper trajectory-level scoring rules for any per-step uncertainty signal calibrated into a probability of eventual success. We prove that TPS strictly elicits the success-probability process under complete observation, within the chosen score family and weight schedule. We extend the construction to administratively censored trajectories by projecting the complete-data score onto the observable stopped prefix, yielding an exact $q_Z$-weighted reduced score and a tractable approximation when $q_Z$ is unestimated. We further show that common trajectory evaluators target weaker objects than the full prefix-conditioned probability process: Trajectory ECE is resolution-blind, while scalarized Trajectory Brier elicits only the collapsed scalar, not the full trace. Experiments on StrategyQA, Tau2-Bench, HotpotQA, and WebShop show that these theoretical distinctions are operationally visible: probability recalibration can substantially change TPS while leaving rank metrics nearly unchanged, and the tractable censored approximation can change the verdict relative to complete-only evaluation.
Authors: Keito Inoshita, Takato Ueno
Abstract: Annotator disagreement in emotion classification reflects ambiguity intrinsic to emotion concepts and is essential for predictor-quality assessment in subjective NLP. Yet no prior work integrates soft-label learning with Bayesian deep learning to evaluate uncertainty along axes including annotator-distribution fidelity. We train a linear head on a frozen RoBERTa via cyclical stochastic gradient Markov chain Monte Carlo (cSG-MCMC), targeting the empirical annotator distribution with a soft-label objective under a five-axis evaluation. On the 28-emotion GoEmotions benchmark, the proposed method outperforms Monte Carlo Dropout and Deep Ensemble simultaneously on three axes -- Jensen-Shannon divergence (JSD) to the annotator distribution, Spearman correlation between per-emotion aleatoric uncertainty and disagreement, and selective-prediction Area Under the Risk-Coverage Curve (AURC) and Area Under the ROC Curve (AUROC) -- showing independent axes are jointly attainable from one posterior. Post-hoc temperature scaling exhibits a bidirectional effect, establishing hard-label calibration and annotator-JSD as independent dimensions and motivating joint reporting as an honest protocol.
Authors: Sasank Annapureddy
Abstract: Operating LLMs as coordinated multi-agent research systems over multi-hour runs surfaces failure modes that single-shot evaluation cannot: upstream providers throttle without warning, sub-agents drift the task to fit accessible tools, narrate machinery instead of using it, open revision iterations with self-apology, or treat upstream context as executable directives. We present PRIMA, whose primary contributions are three operational patterns for surviving these failure modes: (1) a resilience-and-recovery layer that detects upstream rate-limit signals, persists a typed pause record to disk, and resumes long-running runs without re-executing converged work even across process restarts; (2) a sub-agent operating discipline encoding task-fidelity, tool-use, revision, and inter-step context-boundary norms as a structural prompt layer; (3) a multi-phase application pattern for structured engineering deliverables pairing orthogonal draft steps with an explicit cross-document harmonization pass before final synthesis. These sit atop a foundational protocol: a research-program specification language with explicit convergence criteria, a dual-metric scoring engine (LLM-judged rubric plus sandboxed code), an outer meta-optimization loop, event-driven persistence, hook-based middleware, context compaction, and a multi-provider LLM abstraction. Agent identities derive from prime powers, giving collision-free identifiers and trivially-verifiable cluster membership without a central registry. Theoretical guarantees include $O(k)$ verification, $O(V+E)$ DAG validation, and identity collision freedom by the Fundamental Theorem of Arithmetic. A Graph Isomorphism case study grounds the architectural claims in a generated artifact: a six-step protocol that produced a research paper proposing a new canonical-form algorithm with three theorems and five conjectures.
Authors: Zhuocheng Shang, Ahmed Eldawy
Abstract: Domain scientists increasingly develop Python scripts to analyze satellite imagery but they lack scalability to large-scale data. This paper demonstrates GRAIL, an agentic translation system that converts Python geospatial workflows into executable Spark-based programs without requiring scientists to learn a new framework. Rather than fine-tuning a specialized LLM model, GRAIL adapts RDPro, a Scala library for satellite data analysis, to make it LLM-ready using structured documentation, API alias functions, and repair-oriented error logs. Translation is structured as a LangGraph pipeline that decomposes code generation into explicit sections with guided inputs and outputs, enabling targeted repair without regenerating the full program. We demonstrate GRAIL on real-world geospatial workflows and showcase the correctness and scalability of the translated code.
Authors: Yubo Li, Yidi Miao, Yuntian Shen, Yuxin Liu
Abstract: Recent advances in multimodal web agents often rely on increased inference-time computation, including rollout search, verifier passes, offline skill discovery, and specialist model stacks. This raises a central question: can a web agent become more efficient as it accumulates experience, rather than more expensive? We first analyze trajectories from VisualWebArena and identify three recurring sources of inefficiency: repeat-action loops, hidden discovery costs, and low prompt-cache reuse. We then introduce PANDO, a single-rollout online skill-distillation framework that maintains a structured Skill Library and combines progress reflection, confidence-based skill demotion, hierarchical routing, visual compression, and cache-aware prompting. On the full set of 910 VisualWebArena tasks, PANDO achieves a 58.3% success rate, outperforming SGV (54.0%) and our WALT reproduction (45.2%), while using 58% fewer tokens than SGV and 61% fewer tokens than WALT, without any pre-evaluation discovery budget. A 300-task ablation further shows that rules and routines provide most of the success gains, while routing, compression, and cache-aware prompting convert the larger skill library into lower marginal token cost. Finally, we introduce three trajectory-level efficiency metrics -- Action Repetition Rate, Step Overhead Ratio, and Prompt Cache Utilization -- to make efficiency visible beyond terminal success.
Authors: Zhihao Dou, Qinjian Zhao, Zhongwei Wan, Xiaoyu Xia, Sumon Biswas
Abstract: Large language models (LLMs) have achieved strong performance in code generation, but most methods rely on autoregressive decoding without global planning, often leading to locally coherent yet globally suboptimal solutions (e.g., failing test cases or inefficient complexity). While recent approaches such as Chain-of-Thought (CoT) and multi-agent systems (MAS) introduce planning, their limited role specialization and coordination hinder performance on complex tasks. To address the challenges of coordination and specialization in multi-agent code generation, we propose Collaborative Reinforcement Code (CoRe-Code), a framework for role specialized LLM agents that enhances inter-agent coordination to generate more accurate and efficient code. CoRe-Code adopts a simple Planner-Coder paradigm, where the Planner produces high-level plans and the Coder executes them to generate code. We further introduce a collaboration-aware reinforcement learning stage based on Group Relative Policy Optimization (GRPO) to enhance role specialization and alignment. Experiments show that CoRe-Code outperforms a wide range of existing RL-based and multi-agent methods. In addition, we demonstrate that CoRe-Code can generalize to other multi-agent frameworks (e.g., Retrieval and Debugging agents), highlighting its flexibility and scalability. We evaluate CoRe-Code on multiple benchmarks of varying difficulty using three base models. Compared to existing baselines, the results show consistent improvements in accuracy, while also achieving higher efficiency in terms of execution time and memory usage, demonstrating the effectiveness and practicality of CoRe-Code.
Authors: Yilei Zhang
Abstract: Manufacturing has passed through four widely recognized paradigms - mechanization, electrification, programmable automation, and Smart Manufacturing - each defined by the kind of work it shifted from humans to machines. In every case, one layer of industrial work remained fundamentally human: the coordinative cognition of production, comprising the interpretive, allocative, diagnostic, negotiative, and governance work exercised by engineers, planners, and operational managers. We argue that a fifth transition is now underway in which this layer, rather than the physical or routine-cognitive layers below it, is what foundation-model-based autonomous agents primarily redistribute. We name this paradigm Agent Manufacturing and define it operationally: a manufacturing system is an instance of Agent Manufacturing when its principal coordination mechanism is reasoning performed by foundation-model agents that can interpret open-ended goals, plan over long horizons, invoke tools and machines, and negotiate with other agents and humans. This is a narrower and more falsifiable definition than the existing literature on cognitive manufacturing or Industry 5.0 provides, and it distinguishes the paradigm sharply from classical multi-agent manufacturing systems, which were autonomous only within closed protocol spaces.
Authors: Wentong Chen, Xin Cong, Zhong Zhang, Yaxi Lu, Siyuan Zhao, Yesai Wu, Qinyu Luo, Haotian Chen, Yankai Lin, Zhiyuan Liu, Maosong Sun
Abstract: With the continuous advancement of Large Language Models (LLMs), intelligent agents are becoming increasingly vital. However, these agents often fail in environments governed by implicit rules--hidden constraints that cannot be observed directly and must be inferred through interaction. This causes agents to fall into repetitive trial-and-error loops, ultimately leading to task failure. To address this challenge, we propose Test-Time Exploration (TTExplore), a framework where a thinker component analyzes interaction history to infer these implicit rules and guide an actor. Effective exploration in this setting critically depends on the reasoning ability of the thinker. However, evaluating deep reasoning trajectories is inherently unstable and difficult, which poses a major obstacle to effective training. To overcome this issue, we introduce a novel and stable reinforcement learning pipeline. The core idea is to use accurate task-level scores as indirect rewards to bypass the difficulty of evaluating intermediate reasoning, and to retain only a single thinking node per trajectory to alleviate reward sparsity. Using this pipeline, we train a specialized 7B model, Exp-Thinker. Experiments on five text-based embodied tasks show that TTExplore equipped with Exp-Thinker improves baseline agent performance by an average of $14$-$19$ points, demonstrating the effectiveness of explicitly reasoning about implicit rules.
Authors: Chenyou Guo, Zongqi Liu, Yizhou Zhang, Zhaorui Jiang, Ze Liu
Abstract: While general-purpose Large Language Models (LLMs) applied to Geology often hallucinate when reasoning about subsurface structures and deep-time evolution, current AI in Earth sciences predominantly targets surface remote sensing and GIS. To bridge this gap, we introduce Geo-Expert, a family of parameter-efficient geological LLMs fine-tuned on a custom-curated, high-quality instruction dataset processed using our custom instruction synthesis pipeline. We investigate the impact of model scaling and architecture by fine-tuning three base models: Qwen3-8B, Qwen3-32B, and Gemma-3-27B, with Low-Rank Adaptation (LoRA) method. Our extensive evaluation on a novel domain-specific benchmark, Geo-Eval, reveals that a domain-aligned 8B model can outperform open-weight 70B generalists and proprietary GPT-4o on specialized geological reasoning, while a 32B variant approaches frontier reasoning models. The optimized 8B model further offers a competitive cost-performance ratio for deployment. This work provides a reproducible recipe for democratizing scientific LLMs and establishes a baseline for geological artificial intelligence.
Authors: Yuanhong Wang, Juhua Pu, Yuxu Zhou, Yuyi Wang, Ond\v{r}ej Ku\v{z}elka
Abstract: Combinatorial counting problems pervade artificial intelligence, statistics, and discrete mathematics. Whether the task is enumerating subsets, multisets, permutations, partitions, or compositions under structural and arithmetic constraints, solving it remains a stubbornly manual exercise. Closed-form derivations are powerful but brittle, while naive encodings to propositional model counting or constraint satisfaction destroy the exchangeability that makes counting tractable in the first place. We present Cofola (COmbinatorial counting LAnguage with First-Order logic), a typed declarative language whose primitives are the combinatorial objects that recur in everyday counting questions, including sets, bags, tuples, sequences, circles, partitions, and compositions, together with natural relational and arithmetic constraints over them. A denotational semantics maps every Cofola program to a well-defined combinatorial counting problem, and a three-phase compilation pipeline (preprocessing, decomposition, and symmetry-preserving encoding) reduces this problem to a weighted first-order model counting (WFOMC) instance augmented with coefficient-extraction constraints. To stay inside known domain-liftable fragments whenever possible, the encoding groups indistinguishable entities, breaks the symmetry of unordered groupings lexicographically, and encodes sequences and circles via order axioms. On a suite of representative combinatorial counting problems, ranging from textbook math problems to multi-object scenarios that the closest prior framework cannot express, Cofola produces concise specifications and a uniform solving pipeline that is practical end-to-end.
Authors: Xuanting Xie, Zhaochen Guo, Bingheng Li, Xingtong Yu, Zhifei Liao, Zhao Kang, Yuan Fang
Abstract: Chain-of-Thought (CoT) prompting has shown promise in enhancing the reasoning capabilities of large language models (LLMs) on text-attributed graphs (TAGs). This work reframes CoT-based graph learning through the principle of clustering as reasoning, offering a $k$-means interpretation of how iterative reasoning operates over graph-structured data. We observe that existing graph CoT methods rely on disjoint architectures and fixed graph representations, limiting step-by-step semantic-topological interaction and interpretability. To overcome this limitation, we propose a unified framework named KCoT that integrates CoT reasoning with graph representation learning. Our key theoretical result reveals a formal mathematical correspondence between a Transformer block and the $k$-means algorithm, allowing reasoning to be interpreted as iterative assignment and update steps. Based on this insight, we introduce a Semantic Discriminating Prompt that explicitly formulates these steps as structured CoT reasoning, together with a structure-grounded alignment strategy to fuse topological priors with evolving thought-conditioned representations. Experiments on standard benchmarks demonstrate consistent improvements over state-of-the-art methods, validating clustering as a principled mechanism for CoT-based graph learning.
Authors: Xiaoyue Lu, Xianglin Yang, Haijun Liu, Jiahao Liu, Kuntai Cai, Yan Xiao, Jin Song Dong
Abstract: The widespread integration of Large Language Models (LLMs) necessitates rigorous and systematic safety evaluation. Existing paradigms either rely on constructed benchmarks to assess safety from predefined perspectives, or employ dynamic red-teaming to probe potential vulnerabilities. While effective, these approaches face challenges, as they depend heavily on expert domain knowledge, offer limited systematic guarantees, and are vulnerable to rapid obsolescence. To address these limitations, we introduce a novel framework POLARIS that brings the rigor of specification-based software testing to AI safety. POLARIS first compiles unstructured natural-language policies into First-Order Logic (FOL) representations, establishing a traceable link between high-level rules and concrete test cases. This formalization enables the construction of a Semantic Policy Graph, where complex policy violation scenarios are encoded as traversable paths. By systematically exploring this graph, POLARIS uncovers compositional violation patterns, which are then instantiated into executable natural-language test queries, enabling coverage-driven and reproducible safety testing. Experiments demonstrate that POLARIS achieves higher policy coverage and attack success counts compared to established baselines. Crucially, by bridging formal methods and AI safety, POLARIS provides a principled, automated approach to ensuring LLMs adhere to safety-critical policies with verifiable traceability. We release our code at https://github.com/huac-lxy/POLARIS.
Authors: Mathieu d'Aquin
Abstract: Ontologies represent the conceptual knowledge of a domain. At the core of an ontology is the taxonomy of concepts and subconcepts that represent specific entities, which can be complex to build. In many cases, information is available in the form of records describing the characteristics of relevant entities, i.e., tabular data. Identifying patterns and similarities in such data can serve as a basis for identifying concepts and organizing them. However, doing so manually can be challenging, and purely automatic approaches, such as agglomerative clustering or relying on a large language model to analyze the data, can leave the user with overwhelming results and little control. In this paper, we describe a tool that enables the progressive and interactive construction of a taxonomy of concepts by identifying clusters as well as their intentional definitions. To do so, we rely on weighted self-organizing maps as a clustering method because they enable the creation of an arbitrary number of clusters that are distinct with respect to the distributions of values of specific characteristics of the clustered entities. We show that, by integrating this mechanism and others for rapidly creating concepts that group together instances from tabular data, this tool represents a middle ground between purely manual analysis and automatic methods for building ontological taxonomies.
Authors: Lei Ding, Bin He, Chenguang Wang, Yang Liu
Abstract: Proactive task-oriented agents must autonomously anticipate user needs, identify actionable opportunities, and trigger software actions at appropriate moments - fundamentally shifting from reactive systems that await explicit instructions. However, existing approaches lack generalizable end-to-end solutions for measuring and optimizing such anticipatory behaviors. This paper introduces ProActor, a unified framework for conversational task scheduling that integrates: (1) a domain-agnostic automated annotation methodology that enables scalable proactiveness reinforcement learning (RL) by generating full opportunity time windows instead of rigid point labels, (2) systematic proactiveness metrics capturing both timing quality and reference action alignment, and (3) RL optimization using GRPO with various reward designs. Our insight is that RULER-based rewards with proactiveness rubrics are crucial for improving timing quality, and that proactiveness optimization enabled by stage-aware composite rewards is key to balancing timing quality and reference action alignment. Timing-aware RL requires extensive exploration, demanding efficient infrastructure. We develop ART-F, an adaptive framework combining request-adaptive inference clusters with DDP-based training on single-node multi-GPU systems, enabling LoRA training of 4-bit Qwen2.5-14B-ProActor-Q4 with 4-8x speedups. Experiments on two newly auto-annotated datasets demonstrate significant improvements in proactive timing while maintaining action consistency comparable to state-of-the-art (SOTA) baselines. Ablations validate the effectiveness of distinct composite reward variations.
Authors: Hsin-Min Lu, Chen-Yang Lai, Yi-Jhen Li, Ju-Chun Yen
Abstract: Financial Numerical Entity (FNE) understanding aims to recover the meaning of numerical mentions in financial reports. Existing studies primarily focus on concept name prediction and face two important limitations. First, labels derived from inline XBRL may contain errors because filings are usually prepared manually. Second, other important FNE attributes, such as reporting-time relation, measurement scale, and accounting sign, are less emphasized. We propose \textbf{NO}ise-\textbf{R}obust Tagging for Rich Financial Numerical Entity \textbf{A}ttributes (\textsc{NORA}) to address these gaps. NORA uses task-aware instance-specific weighting to attenuate the influence of noisy labels during training, and we further propose the Neighborhood Prior-adjusted KNN (NPK) filtering method for more reliable evaluation on real-world noisy test sets. In addition, we construct a large-scale benchmark containing 6.6 million instances with multi-attribute labels and filing metadata. Experiments show that \textsc{NORA} performs strongly compared with state-of-the-art noisy-label baselines, including Co-teaching, Mixup, SSR, and SelfMix. Moreover, NORA is robust under both unfiltered and noise-filtered test settings. It achieves the best Accuracy, Macro F1, and Weighted F1 for concept name and time-relation prediction, while remaining competitive on scale and sign prediction. These results demonstrate the value of jointly modeling rich FNE attributes while accounting for label noise in real-world financial filings.
Authors: Filip Cano, Thomas A. Henzinger, Konstantin Kueffner
Abstract: Runtime fairness is not a one-time constraint but a dynamic property evaluated over a sequence of decisions. To ensure fairness at runtime, it is necessary to account for past decisions, information neglected by conventional, static classifiers. Traditional fairness shields enforce runtime fairness abruptly, by intervening \emph{deterministically} whenever a sequence of decisions violates the target for a running fairness measure. This motivates our \emph{main conceptual contribution: \textbf{energy shields}.} An energy shield is a novel, lightweight, adaptive controller that monitors a sequence of decisions and intervenes \emph{probabilistically} to ensure runtime fairness smoothly, by utilizing physics-inspired energy functions to nudge the sequence toward fairness: the more unfair the decisions, the stronger the nudging force becomes. This makes energy shields the \emph{\textbf{first}} fairness shields to provide both \emph{short-term safety and long-term liveness guarantees}. Safety ensures that the running fairness measure stays within a running target interval with high probability, and liveness ensures that the limit of the fairness measure lies within the limit target interval. Intuitively, the short-term specifies the tolerated fairness values and the long-term specifies the desired fairness values. We also provide a synthesis procedure for constructing the least intrusive energy shield for a given target specification, and demonstrate its efficiency experimentally. We evaluate our energy shields against existing fairness shields through the lens of short- and long-term fairness.
Authors: Chengrui Li, Rujing Li, Yitong Bai, Rui Li
Abstract: Industrial asset operations and maintenance question answering is inherently multi-turn, iterative, and highly dependent on external tool invocation. However, the conventional plan-execute single-agent architecture exhibits clear limitations in maintaining cross-turn context, and reusing intermediate results. In this paper, we present a multi-turn dialog system designed for industrial scenarios based on a supervisor-specialist multi-agent architecture. To alleviate tool invocation bottlenecks, the system incorporates structured artifact reuse, dynamic replanning, and parallel tool execution. Evaluation results show that our system achieves better response quality compared with the baseline, with planning effectiveness increasing by 54.5% and task completion improving by 37.8%. System profiling further shows that cross-turn artifact reuse effectively reduces redundant tool invocation, decreasing the tool-time share from 47.3% to 26.3% and making turns 2-5 approximately 4.2x faster than the first turn.
Authors: Yuanzhi Xu, Qian Gao, Jun Fan, Guohui Ding, Zhenyu Yang, Sixue Lin, Yuteng Xiao
Abstract: The generation of factually incorrect objects, commonly known as object hallucination, remains a persistent challenge in Large Vision-Language Models (LVLMs). Current approaches to address this issue - ranging from expensive data-driven fine-tuning and high-latency contrastive decoding to rigid attention head truncation - frequently compromise either computational efficiency or the continuity of the model's feature space. To overcome these limitations, we introduce a novel, training-free inference strategy that operates as a region-aware adaptive weighting mechanism to dynamically correct semantic drift without relying on abrupt heuristic truncations. By computing an outlier-resistant statistical midpoint across various attention heads, we establish a stable anchor for reliable visual representations. We then utilize the inter-head disagreement mapped across regions to dynamically determine intervention budgets, gently suppressing hallucination-inducing attention paths through a continuous penalty modulation. This recalibration process effectively rectifies visual-semantic misalignments while fully preserving generative fluency and language priors. Comprehensive evaluations on standard multimodal benchmarks, including CHAIR, POPE, and MME, reveal that our strategy substantially curtails both instance- and sentence-level hallucinations. The results demonstrate state-of-the-art performance against contemporary baselines, confirming our method's efficiency and algorithmic robustness. Our code will be public.
Authors: Sijin Yu, Zijiao Chen, Zhenyu Yang, Zihao Tan, Jiakun Xu, Zhongliang Liu, Shengxian Chen, Wenxuan Wu, Xiangmin Xu, Xin Zhang
Abstract: Current fMRI decoders face a performance-fidelity trade-off where efficient ID encoders outperform geometrically faithful surface-based models. We argue this is partly driven by inefficient surface tokenization and the failure to use anatomy as a predictive signal. We present NeurIPS, a framework that improves surface-based decoding by reframing anatomical variation from a nuisance to a powerful inductive prior. NeurIPS unites two innovations: a Selective ROI Spherical Tokenizer (SRST) for efficient geometric encoding, and a Structure-Guided Mixture of Experts (SG-MoE) that explicitly models individual anatomy using cortical features. On the Natural Scenes Dataset, NeurIPS establishes a new state-of-the-art for surface decoders and achieves performance comparable to strong 1D baselines. This is achieved with unprecedented efficiency, as the model converges dramatically faster (10 vs. 600 epochs). This efficiency enables rapid adaptation to new subjects using only 20% of data and ensures robust scalability as the training cohort is expanded. Ablations provide causal evidence that these gains are driven by the model's use of cortical features, not by memorizing subject IDs. By leveraging anatomical priors, NeurIPS provides a principled and scalable path toward robust, generalizable brain decoding.
Authors: Abdurrahim Yilmaz, Ay\c{s}e Esra Koku Aksu, Duygu Yamen, Vefa Asli Erdemir, Mehmet Salih Gurel, Gulsum Gencoglan, Joram M. Posma, Burak Temelkuran
Abstract: Chronic dermatologic diseases such as pemphigus require long-term follow-up, generating extensive longitudinal clinical documentation that is difficult to review comprehensively during routine visits and increasing clinician workload as well as the risk of missing critical historical information. We evaluated whether a locally deployed, privacy-preserving small language model (SLM) could retrieve structured clinical features and generate longitudinal summaries from long-term dermatology follow-up records. In this retrospective case series, thirty pemphigus patients contributed 541 visit notes that were aggregated into full longitudinal records (89,336 words); 56 clinically relevant features were annotated by two expert dermatologists. The locally deployed SLM (Qwen3 4B Thinking 2507) was queried with each complete record to retrieve 56 features and generate one final report summaries. Across 1,680 feature retrieval tasks, mean accuracy was 82.25%. Dermatologists' ratings of AI-generated summaries were high for overall quality (8.23-8.47), clinical accuracy (7.93-8.20), and usefulness (8.47-8.50), with no significant inter-evaluator differences and an overall preference for AI summaries in 53.3% of evaluations. These findings suggest that privacy-preserving, locally deployed SLMs can outperform medical experts and reliably generate clinically meaningful longitudinal summaries. SLMs may support clinical decision-making when integrated with appropriate oversight.
Authors: Tianxiang Zhan, Xiaobao Song, Tong Guan, Shirui Pan, Ming Jin
Abstract: Time series research is moving beyond fixed forecasting benchmarks toward realistic tasks that combine prediction, contextual reasoning, tool use, and structured decision support. Most benchmarks are built around clean data and short evaluation loops; agents alone may miss temporal constraints, evidence checks, or review before finalizing outputs. We first formalize next-generation time series tasks as three-component tuples consisting of a task file, a workspace, and a validation interface. We then present AION, a time series harness built from six component groups: agents, skills, rules, memory, evaluation, and protocols. In this harness, we use three design principles: temporal grounding, temporal knowledge-grounded reasoning, and reliability mechanisms such as post-experiment analysis and layered review. A Kaggle Store Sales case study shows that the harness produces more detailed process traces, more artifacts, and more review steps than the same base agent operating in OpenCode direct build mode. Taken together, these results argue for a paradigm shift from fixed tasks to realistic ones under real-world constraints.
Authors: Chengwei Li, Junlin Liu, Yang Gao
Abstract: As modern air combat evolves toward beyond-visual-range (BVR) multi-aircraft cooperative engagements, autonomous decision-making for unmanned combat aerial vehicles (UCAVs) faces significant challenges due to high-dimensional state spaces, discrete action commands, and strongly adversarial dynamic environments. To overcome the limitations of existing multi-agent reinforcement learning (MARL) methods in such settings, namely insufficient exploration efficiency, low sample utilization, and poor policy generalization, we propose Adversarial Curriculum and Evolutionary-enhanced Multi-agent Proximal Policy Optimization (ACE-MAPPO), a hybrid learning framework that integrates evolutionary algorithms with MAPPO. Specifically, a genetic soft update mechanism is introduced to enhance population diversity and mitigate convergence to local optima. An evolutionary-augmented prioritized trajectory replay strategy is further employed to improve the utilization of sparse high-value samples. In addition, an adversarial evolutionary curriculum learning mechanism is designed to enable adaptive training with progressively increasing difficulty. Extensive experimental results demonstrate that the proposed method outperforms MAPPO and other baseline algorithms in terms of training stability, convergence speed, and win rate, validating its effectiveness in multi-aircraft cooperative air combat scenarios.
Authors: Hadi Hajieghrary, Benedikt Walter, Chaitanya Shinde, Paul Schmitt, Miguel Hurtado
Abstract: Autonomous driving stacks must pick one trajectory from a multi-modal candidate set; choosing by model confidence ignores safety, traffic-law, and comfort constraints. We present \textsc{RECTOR} (Rule-Enforced Constrained Trajectory Orchestrator), a post-generation reranking layer that scores candidates against a tiered rulebook (Safety~$\succ$~Legal~$\succ$~Road~$\succ$~Comfort) via differentiable proxies and a scene-conditioned applicability mechanism, then selects with a deterministic $\varepsilon$-lexicographic rule that preserves cross-tier priority by construction -- without retraining the predictor. On the Waymo Open Motion Dataset \texttt{validation\_interactive} split (43{,}219 augmented instances, $K{=}6$), under Protocol~B (28-rule proxy catalog, oracle applicability) rule-aware selection cuts Safety+Legal violations from 28.58\% to 20.42\% and Total from 40.32\% to 32.41\% versus confidence-only on the same candidates. A uniform-weight weighted-sum baseline matches binary compliance on this benchmark -- the empirical lift comes from rule-aware ranking, while the lexicographic guarantee is the structural differentiator no weight calibration can replicate. Under adversarial confidence corruption, confidence-only selection fails in 100\% of scenarios while both rule-aware selectors reject the injected mode in $\sim$96\%. All figures are proxy-evaluator results (not a safety certificate), open-loop, 5\,s horizon, U.S.\ rules, validation split.
Authors: Jo\~ao Sedoc, Baotong Zhang, Dean Foster
Abstract: Reliably knowing when a language model is correct is almost as important as being correct. We introduce prover-verifier deliberation (PVD), an inference-time protocol grounded in interactive proof theory, as a mechanism for selective prediction: the protocol produces both an answer and a structured confidence verdict, allowing a system to report high-confidence answers while abstaining on uncertain cases. In each dialogue, a prover defends a candidate answer through checkable sub-claims while a verifier issues targeted challenges and returns \textsc{Accept}, \textsc{Challenge}, or \textsc{Reject}. Because frozen language models are imperfect provers and verifiers operating over a noisy channel, formal soundness and completeness guarantees do not transfer; instead, we characterize the protocol empirically through its coverage-precision behavior. Our main experiment uses Claude Sonnet 4.6 as prover and Claude Haiku 4.5 as verifier on GPQA Diamond. Questions accepted with no answer revision, which we call Accept + No Change (ANC), are reported as the high-confidence subset; we evaluate this subset by its precision and coverage. ANC separates reliable from unreliable answers, yielding a $\sim$30pp HC-Prec gap over the non-ANC complement. Robustness experiments with GPT and Gemini pairings show that high HC-Prec can transfer across model families, while verifier strictness and domain competence largely determine the size of the selection gap. On Humanity's Last Exam, weaker prover-verifier pairings can collapse or invert the ANC signal, illustrating a practical failure mode when the verifier operates outside its effective region. Comparisons with self-consistency, universal self-consistency, multi-agent debate, and Reflexion suggest that prover-verifier deliberation supplies a distinct argument-defensibility signal for selective prediction.
Authors: Dao Tran, Duc Anh Le, Ngoc Luu, Quan Pham, Tung Pham, Hung Bui
Abstract: Test-time scaling improves language model reasoning by spending additional compute to explore multiple solution trajectories. The key challenge is to maximize accuracy while minimizing the total number of generated tokens during reasoning. Recent PRM-guided methods score intermediate prefixes to steer this search, but most are frontier-only: they keep only the current active prefixes and irreversibly prune or resample away the rest using noisy PRM scores. This can cause premature commitment, diversity collapse, and the loss of prefixes that still admit correct continuations. We introduce stochastic backtracking over a persistent pool of historical prefixes, allowing test-time compute to revisit previously generated states instead of only expanding the current frontier. To make this efficient, we propose two complementary mechanisms. Subpool Selection strengthens greedy PRM-guided search by applying Top-N selection within random subpools, giving historical prefixes a chance to bypass over-scored frontier candidates. Power Backtrack Sequential Monte Carlo extends SMC-style resampling to the persistent pool using powered PRM scores and mixture-corrected weights. Across mathematical reasoning benchmarks and model scales, our methods consistently achieve higher accuracy per token count, and the same level of accuracy using only a fraction of the token count in comparison to strong PRM-guided baselines, demonstrating that persistent-pool stochastic backtracking provides a simple and effective way to improve the accuracy-token trade-off in test-time scaling.
Authors: Ciar\'an Walsh, Emilio Barkett
Abstract: Large language models are increasingly used as behavioral simulators, but it remains unclear when their outputs reflect human-like cognitive mechanisms rather than prompt-sensitive surface patterns. We study this question through the realization effect, a well-characterized finding in behavioral economics in which risk-taking differs systematically after paper versus realized gains and losses. We evaluate LLM behavior at three levels: prompt-only behavioral sensitivity, linear readout of internal representations, and causal control via activation steering. Prompt-only results show systematic condition sensitivity, but the directional pattern does not reproduce human realization-effect predictions. Gemma's residual stream contains a linearly decodable realization-status signal at layer 18 that generalizes to held-out prompts. Steering along this direction does not, however, reliably shift downstream risk choices, a null result that holds across positive scales and in a negative sign-symmetry run. Behavioral sensitivity, latent readout, and causal control are three distinct properties that do not automatically co-occur, and successful latent readout is insufficient evidence that a model behaviorally relies on a representation during downstream decision-making.
Authors: Guohong Liu, Jialei Ye, Pengzhi Gao, Wei Liu, Jian Luan, Yunxin Liu, Yuanchun Li
Abstract: Mobile GUI agents powered by large language models have progressed rapidly, creating urgent needs for realistic and comprehensive evaluation. Existing benchmarks prioritize reproducibility but are often limited to open-source apps or file-operation tasks for the difficulty of constructing rewards on real applications, leaving a gap between benchmark settings and real-world usage. Moreover, most benchmarks focus on basic grounding and navigation, with limited coverage of complex, long-horizon interactions. To address these limitations, we introduce SimuWoB, a fully synthetic benchmark for mobile GUI agents with 120 challenging tasks spanning diverse types and difficulty levels. We build a robust virtual environment generation framework that synthesizes high-fidelity tasks and environments, and automatically provides valid rewards for each task. Each environment is deployed as a backend-free webpage accessible via URL, enabling efficient and reproducible evaluation. We conduct comprehensive experiments on several state-of-the-art mobile GUI agents. The average success rate is only 27.92%, dropping to 17.82% on long-horizon tasks, which reveals substantial weaknesses in current agents under complex scenarios. Evaluation result comparison with real-world sample tasks demonstrate that agent assessments based on our synthetic environment generalize well. We further provide diagnostic insights across key capability dimensions and discuss implications for future mobile GUI agent development.
Authors: Jaime Rafael Imperial, Hao Zheng
Abstract: Existing Large Language Model (LLM) approaches to SystemVerilog Assertion (SVA) generation primarily focus on syntactic validity and formal verification outcomes, while semantic alignment between generated assertions and natural language specifications remains difficult to quantify. As a result, hallucinated or misaligned SVAs can reduce confidence and increase debugging efforts in the absence of golden RTL. This paper presents SpecAlign, a framework for semantic evaluation and refinement of LLM-generated SVAs. SpecAlign introduces two iterative alignment loops that assess both natural language properties and SVAs against the design specification using entailment-based classification. We improve alignment decisions by generating multiple reasoning paths using chain-of-thought prompting and aggregating them via a self-consistency voting mechanism. Misaligned assertions are analyzed to generate actionable feedback for refinement. We further define a quantitative alignment score to measure semantic consistency across iterations. Experimental results demonstrate that SpecAlign effectively detects semantic inconsistencies and improves assertion alignment without relying on golden RTL, providing a scalable complement to traditional formal verification evaluation metrics.
Authors: Yi Li, Songtao Wei, Dongming Jiang, Zhichun Guo, Qiannan Li, Bingzhe Li
Abstract: Multi-agent LLM systems improve reasoning by combining outputs from multiple agents, but interaction-heavy methods can introduce error propagation and high communication overhead. When agents exchange raw responses or reasoning traces, incorrect intermediate reasoning may be adopted and amplified, leading to confident but wrong consensus; multi-round communication also increases token consumption, latency, and inference cost. In this paper, we propose a controlled-communication coordination framework named DarkForest. DarkForest first keeps agents independent, so each agent produces an answer without seeing the others' outputs. It then parses the raw responses into structured candidate records, groups semantically equivalent candidates into clusters, and estimates a calibrated belief distribution over these clusters using agent reliability, confidence, parse quality, support-pattern reliability, and independence corrections. A coordinator receives only policy-permitted evidence from this belief state with controlled communication. Experiments on six reasoning benchmarks show that DarkForest achieves leading overall quality, improves the strongest baseline by up to 30.7\% on benchmark metrics, and reduces token consumption by up to $6.5\times$ compared with communication-heavy baselines.
Authors: Andrew Corbett, Archit Sood, Anna Tzatzopoulou, Sai-Aakash Ramesh, Tim Dodwell
Abstract: Recent work on recursive architectures has shown that tiny neural networks can be surprisingly powerful on structured reasoning tasks. The trick is to model reasoning trajectories with a latent dynamical system. We argue that the inference-time behaviour of these architectures is best understood as approximate inference over latent reasoning trajectories, with deterministic recursion as the one-particle, zero-noise limit. We make this view operational through guided stochastic exploration: stochastic perturbations of the reasoning dynamics propose neighbouring trajectories, and the model's existing early-stopping head reweights them online. The framework yields three label-free diagnostics: local stability, guide alignment, and cloud-token entropy. These predict, from inference traces alone, whether the procedure will help and which of its outputs to trust. On Sudoku-Extreme it lifts exact-solve accuracy from $85.9\%$ to $98.0\%$ without retraining; on Maze-Hard the diagnostics flag a misaligned guide, as validation performance later confirms. The same machinery thus characterises both when recursive reasoning has room to improve at the trajectory level and when the model's internal guide can recover it.
Authors: Andy Xu, Yu-Wing Tai
Abstract: AI agents are increasingly used to solve complex, multi-step tasks, but existing multi-agent frameworks remain brittle as workflows grow in scale and depth. Small errors at intermediate stages can propagate through agent interactions, while insufficient grounding and weak verification mechanisms further limit reliability. We present Meta-Agent, a two-phase framework that automatically constructs and executes specialized multi-agent systems from natural-language task descriptions. In the construction phase, a task planner decomposes a problem into a directed acyclic graph of agent specifications with explicit input/output contracts and verification criteria. A web search module grounds each specification with external evidence, and a code generation module produces system prompts and tool configurations. A construction-time verification stage then validates generated artifacts and triggers targeted regeneration when failures are detected. In the execution phase, a coordinator dispatches subtasks across the agent graph while execution-time verification gates intermediate outputs. We further introduce a three-level error attribution mechanism that distinguishes local, upstream, and structural failures, enabling targeted recovery strategies ranging from localized retries to partial re-execution and re-decomposition. We evaluate Meta-Agent across coding, contextual learning, and open-ended reasoning tasks. Experiments against strong multi-agent baselines and ablation studies demonstrate consistent improvements in task success rate, error recovery, and workflow stability. The results highlight the importance of tightly integrating planning, grounding, and verification for building reliable multi-agent systems.
Authors: Minwei Kong, Chonghe Jiang, Ao Qu, Wenbin Ouyang, Zhaoming Zeng, Xiaotong Guo, Zhekai Li, Junyi Li, Yi Fan, Xinshou Zheng, Xi Jing, Yikai Zhang, Zhiwei Liang, Seonghoo Kim, Runqing Yang, Zijian Zhou, Sirui Li, Han Zheng, Wangyang Ying, Ou Zheng, Chonghuan Wang, Jinglong Zhao, Hanzhang Qin, Cathy Wu, Paul Pu Liang, Jinhua Zhao, Hai Wang
Abstract: Large language models (LLMs) are increasingly used for optimization modeling and solver-code generation, yet practical operations research and optimization problems often require a harder capability: designing scalable algorithms that exploit problem structure and outperform direct formulation-and-solve baselines. Existing benchmarks are limited to small or simplified examples far below real-world scale and complexity. We introduce FrontierOR, among the first benchmarks to systematically evaluate LLM-based efficient algorithm design for realistic large-scale optimization problems. FrontierOR includes 180 tasks derived from methodologically diverse papers published in top-tier operations research venues, each with standardized instances and a hidden, expert-verified evaluation suite. We evaluate seven LLMs spanning frontier, cost-effective, and open-source models both in one-shot and test-time evolution settings. The results reveal that frontier models still struggle to move from executable formulations to efficient optimization algorithms: the strongest one-shot model outperforms Gurobi in only 31% of cases in both solution quality and computational efficiency, and even strong coding agents with test-time evolution achieve only 50% on selected hard tasks. FrontierOR establishes a practical evaluation platform for LLM-based optimization algorithm design, which enables future LLMs and agents to be systematically tested on whether they can move beyond correct formulation toward a feasible, high-quality, and efficient algorithm.
Authors: Leshu Li, An Lu, Haiyu Wang, Zhibin Feng, Conghui Duan, Qing Bao, Zongmin Zhao, Sai Qian Zhang
Abstract: Lipid nanoparticles (LNPs) are among the most clinically mature platforms for nucleic acid delivery, yet designing lipids that are both effective and biologically safe remains a major bottleneck. In practical screening, toxicity is a decision-level constraint: if a lipid is toxic, its efficiency prediction is clinically irrelevant. We propose LipoAgent, a safety-aware multi-agent LLM framework for lipid discovery. LipoAgent combines domain-specific finetuning with a conditional prediction objective that enforces toxicity as a prerequisite for efficiency prediction, and further improves reliability via multi-agent verification with lightweight human oversight when disagreement persists. Across multiple foundation models, LipoAgent achieves an average 32% relative improvement in mRNA transfection efficiency prediction compared with other reported models for lipid design. Wet-lab validation confirms that virtual screening rankings reliably translate to biological transfection outcomes. The code is publicly available at https://github.com/SAI-Lab-NYU/LipoAgent.git.
Authors: Niklas Weller, Emilio Barkett
Abstract: Aligning AI systems with organizational decision-making is typically framed as a single-target problem: make the model behave like the organization. We argue this framing obscures a deeper pluralistic challenge. We rely on a decision-policy capturing method to measure process alignment: whether an LLM weights information as the organization does, not merely whether it reaches the same conclusions. Applying this method to ECHR Article 6 decisions, process alignment strongly predicts output accuracy (r = 0.85, p < .001) and externalization substantially improves alignment for poorly-aligned models. Applying it to German consumer credit decisions, this relationship collapses (r = 0.15, p = .60): interventions produce inconsistent effects and the benchmark encodes potentially discriminatory historical patterns. This contrast is itself a pluralistic alignment finding: in contested domains, high process alignment is neither achievable via externalization nor unconditionally desirable. Output agreement alone cannot distinguish a model that has internalized an organizational policy from one that merely approximates its outcomes; process-level measurement is a necessary component of any pluralistic alignment evaluation.
Authors: Michael Hardy, Anka Reuel, Lijin Zhang, Jodi M. Casabianca, Sang Truong, Yash Dave, Hansol Lee, Benjamin Domingue, Sanmi Koyejo
Abstract: While aggregate leaderboard scores drive AI development, they contain substantial measurement noise whose sources and magnitudes remain unquantified, making it unclear when rankings reflect genuine capability differences versus evaluation artifacts. We introduce a framework for measuring the latent landscape in AI benchmark ecosystems. Applying Confirmatory Factor Analysis (CFA) and Generalizability Theory to 4,000+ models from the Open LLM Leaderboard, we decompose sources of ranking variance and establish: (1) structures assumed in current reporting practice underestimate the strength of relationships between benchmarks; (2) evidence of local dependence among leaderboard items, undermining uses of benchmarks as measurement instruments under current scoring systems; (3) contributor metadata explains more rank-relevant variance ($\approx9\%$) than architecture or deployment categories in this context; (4) a manifest-score "scaling law" slope has low reliability ($R_{\beta}=0.53$); by contrast, the latent general-factor size slope is highly stable across ecosystem controls ($R_g=0.97$). We are able to provide unique insights into benchmark dynamics, such as which benchmarks are a function of LLM size and which can be oppositely impacted by post-training practices. We provide actionable diagnostics to determine how benchmark rankings can be trusted and how benchmark design can be improved.
Authors: Hongbo Jin, Mingnan Zhu, Jingqi Tian, Xu Jiang, Zhongjing Du, Haoran Tang, Siyi Xie, Qiaoman Zhang, Jiayu Ding
Abstract: While LLMs excel at reasoning over prompts using static pretrained knowledge, they struggle significantly with context learning-the ability to dynamically extract, internalize, and apply new knowledge from complex, task-specific contexts. Recent evaluations on the CL-Bench reveal a critical capability gap: frontier models solve only 17.2% of context-dependent tasks on average.
Authors: Ashwath Vaithinathan Aravindan, Mayank Kejriwal
Abstract: Large language models often generate confident but incorrect answers rather than abstaining when uncertain. This problem is particularly acute for small language models (SLMs), where computational constraints and autonomous operation amplify the need for reliable uncertainty detection. We propose _Second Guess_, a lightweight, parameter-free prompting technique for abstention in multiple-choice question answering (MCQA) that is well-suited for SLMs. Our key empirical insight is that models which truly know an answer will select it consistently, while uncertain models exhibit unstable behavior when an ``I don't know'' option is added. Evaluated on four open models (2B-8B parameters) and four benchmarks, Second Guess achieves the highest composite risk improvement of 10.81\%. Notably, it maintains an 8\% composite risk improvement on fine-tuned models where entropy-based methods degrade, and improves most for lower-performing models. All code and results required to reproduce this work is available in https://github.com/Mystic-Slice/second-guess
Authors: Yishu Wei, Hexin Dong, Yi Lin, Jiahe Qian, Yi Liu, Yifan Peng
Abstract: Objective: Survival analysis is central to medical prediction, yet large language models (LLMs) are rarely used as end-to-end survival models because censoring prevents straightforward supervised fine-tuning. Here we present LLMSurvival, a framework that enables censoring-aware survival analysis with unmodified LLMs operating directly on tabular clinical data. Materials and Methods: LLMSurvival reformulates time-to-event prediction as pairwise ranking among comparable subjects, and derives test-time risk by aggregating comparisons against anchor individuals from the training cohort. Results: Across two clinical tasks (ICU mortality prediction in MIMIC-IV and fragility fracture prediction in a NewYork-Presbyterian/Weill Cornell Medicine cohort), LLMSurvival improves overall concordance over Cox proportional hazards modeling by 3.1% for ICU mortality and 0.5% for fracture risk, 2.1% on average for ICU mortality and 2.8% for fracture risk over three established deep learning survival models. Discussion: The results show that survival modeling with censoring can be made compatible with LLM fine-tuning through comparison-based reformulation. The framework demonstrates high portability and superior performance over expert curated scores like SAPS-II and FRAX scores across diverse clinical context. Furthermore, the framework supports local deployment, as compact, publicly available base models provide sufficient performance. Conclusion: The LLMSurvival framework serves as a proof of concept for an integrated, censoring-conscious approach to survival analysis via LLMs.
Authors: Yanzhou Li, Yiran Zhang, Xiaoyu Zhang, Xiaoxia Liu, Yang Liu
Abstract: Coding agents produce rich trajectories while solving software-engineering tasks. To enable agent self-evolution, these trajectories can be distilled into reusable procedural skills that compactly encode experience to guide future behavior. However, existing skill construction and maintenance methods often rely on fixed prompts and heuristic update rules, leaving it unclear how knowledge should be selected, abstracted, and maintained to best serve downstream agents. We propose CODESKILL, an LLM-based framework that reformulates skill extraction and skill-bank maintenance as a learnable management policy. CODESKILL extracts multi-granularity procedural skills from coding-agent trajectories, evolves skills with new experience, and maintains a compact skill bank for future task solving. We train CODESKILL with reinforcement learning, using a hybrid reward that combines dense rubric-based skill-quality feedback with sparse verifiable execution feedback from the frozen downstream agent. Experiments on EnvBench, SWE-Bench Verified, and Terminal-Bench 2 show that CODESKILL improves average pass rate by 9.69 over the no-skill baseline and by 4.01 over the strongest prompt-based or memory baseline, while maintaining the skill bank at a stable size during iterative construction.
Authors: Yuntao Wang, Jianle Ba, Han Liu, Yanghe Pan, Jintao Wei, Zhou Su, Tom H. Luan, Linkang Du
Abstract: The rapid evolution of large language model (LLM)-driven autonomous agents has given rise to OpenClaw, a new class of open-source agent frameworks that operate as continuously running, skill-augmented systems with persistent memory, multi-channel interaction, and high degrees of autonomy. Such capabilities enable OpenClaw agents to autonomously execute complex, multi-step tasks and interact seamlessly with external applications, but simultaneously introduce a substantially enlarged attack surface. In particular, the combination of high-privilege operations and persistent memory exposes OpenClaw agents to various emerging threats, including skill poisoning, cognitive manipulation, multi-agent cascading failures, and supply-chain vulnerabilities. In this survey, we present a comprehensive study of the security landscape of OpenClaw agents. We first examine the general architecture and key characteristics that distinguish OpenClaw agents from traditional AI agent systems. We categorize existing security and privacy threats into a layered framework and analyze how vulnerabilities arise during agent reasoning, action execution, and external interaction. Representative defense mechanisms are also reviewed to draw the current defense landscape. Finally, several unresolved issues related to the reliability and trustworthiness of OpenClaw ecosystems are discussed.
Authors: Ziqing Yu, Yuhui Tao, Jiayu Huo, Lei Pan, Zilong Xiao, Juecheng Chen, Xiao Li, Jianxuan Li, You Zhou, Zhixing Li, Cong Wang, Beijian Zhang, Chen Chen, Hongyang Lu, Konstantinos Patlatzoglou, Daniel B. Kramer, Jonathan W. Waks, Yangang Su, Fu Siong Ng, Shuo Wang, Yixiu Liang, Junbo Ge
Abstract: Electrocardiography (ECG) is central to cardiovascular care, but conventional AI models are often restricted to common arrhythmias and may generalize poorly across populations or clinically subtle diseases. We developed ECG Contrastive Language-Image Pre-training (ECGCLIP), a signal-language contrastive learning framework that aligns ECG waveforms with expert diagnostic reports. ECGCLIP was pre-trained on 2,837,962 ECG studies from 1,324,856 patients and evaluated on a held-out internal test set plus nine independent external cohorts comprising about 1.5 million ECGs. Evaluation covered 89 downstream tasks, including 45 ECG diagnoses, 39 echocardiographic targets, and 5 rare cardiac diseases, using PRAUC as the primary metric. ECGCLIP consistently improved performance over random initialization and Merl-R18 baselines. On the internal test set, ECGCLIP-R34 achieved strong performance for atrial fibrillation (PRAUC 0.900) and ST-segment elevation myocardial infarction (PRAUC 0.383), with robust generalization across all external cohorts. It also improved low-prevalence and diagnostically elusive diseases, including Ebstein anomaly, constrictive pericarditis, dextrocardia, and cardiac amyloidosis, with internal PRAUC values of 0.253, 0.175, 0.121, and 0.201, respectively. ECGCLIP was data efficient, matching or exceeding full-dataset baseline performance with only 10% of training data. Feature visualization and saliency analysis suggested clinically meaningful representations aligned with established electrocardiographic criteria. These findings indicate that large-scale ECG-report contrastive pre-training can expand routine ECG interpretation beyond common arrhythmias toward broad cardiovascular assessment and opportunistic screening of echocardiographic and rare conditions.
Authors: Natalia Andrienko, Gennady Andrienko, J\"urgen Bernard, Michael Sedlmair
Abstract: Visual analytics (VA) workflows are inherently complex, involving data transformation, feature engineering, visual representation, and human interpretation. They are typically described in unstructured prose, hindering systematic comparison, reuse of proven strategies, and training of novices. We present Artifact-Transform Workflow Language (ATWL), a domain-agnostic, declarative language that formally represents VA workflows by capturing their structure and underlying analytical intent. ATWL is built upon a modular ontology of eight artifact types (entities, features, arrangements, visualisations, patterns, models, knowledge, specifications) and transforms characterised by standardised intents (e.g., define-unit, characterise, contextualise, abstract). To show that formalisation effort need not impede adoption, we extract workflows from research papers through supervised interaction with LLM agents, reducing the human role to review and refinement. Using this process, we constructed a library of seventeen ATWL workflows from published VA papers. Cross-workflow analysis reveals structural regularities -- a recurrent meta-structure, recurring motifs, reusable building blocks, diverse iterative strategies, and cross-domain equivalences -- that remain invisible in prose. We further evaluate practical utility through a controlled experiment in which the same LLM addressed two analytical problems with the library supplied either as original papers or as ATWL representations. Both forms enabled useful recommendations, but the formal representation systematically added explicit iteration structure, typed data flow, fragment-level adaptation provenance, and compactness supporting scaling beyond what prose libraries can fit in an LLM's context. ATWL enables a transition from narrative descriptions to formally represented, comparable, and reusable analytical knowledge.
Authors: Ankur Samanta, Akshayaa Magesh, Ayush Jain, Youliang Yu, Daniel Jiang, Kavosh Asadi, Kaveh Hassani, Paul Sajda, Jalaj Bhandari, Yonathan Efroni
Abstract: Contemporary reinforcement learning with verifiable reward methods post-train language models on multi-step reasoning by assigning a single outcome reward uniformly across all tokens in a trajectory. Such uniform assignment ignores which steps contributed to success or failure. Improving credit assignment can address this limitation by enabling targeted refinement of faulty reasoning steps, rather than updating entire trajectories uniformly. Resets are one such simple mechanism, enabling more precise credit assignment by returning to an intermediate state and resampling counterfactual continuations, so that outcome differences can be attributed to decisions made at that point. We propose two such methods: Random-Reset Policy Optimization (RRPO), where reset states are drawn randomly from reasoning steps, and Self-Reset Policy Optimization (SRPO), where the model self-localizes the erroneous step in an incorrect trajectory and resets there. We analyze these methods within the Conservative Policy Iteration (CPI) framework. Extending CPI with a credit-assignment oracle that targets improvable states yields provable improvements over random resets. Across models and reasoning benchmarks, SRPO consistently outperforms standard GRPO and RRPO by sampling multiple suffix continuations at a self-localized reset and learning from their rewards, using only the model itself with no external supervision.
Authors: Rahul Vishwakarma, Shushant Kumar, Ratnesh Jamidar
Abstract: AI answer engines generate answers from retrieved pages but cite only a few sources. This makes visibility depend not just on ranking, but on being cited. We study competitive Generative Engine Optimization (GEO): when two retrieved candidates compete, what makes one more likely to be cited first? We build a controlled two-document retrieval-augmented generation (RAG) testbed that injects exactly two candidate sources into the model context and measures which source is referenced by the first citation marker in the output. Across six LLMs we execute 252,000 trials, repeated paired comparisons under one factorial program over 18 content factors. In each trial the two sources differ in exactly one factor; we use brand anonymization and counterbalanced source order to separate content effects from position bias. Mixed-effects models show that topical relevance and list position are the biggest drivers of being cited first. Including explicit price information and a recent timestamp also helps consistently. Completeness and trust cues add smaller gains, while formatting-only edits have little impact. We release a reproducible evaluation protocol and a prioritized GEO checklist for practitioners, and we exercised it in an early internal pilot at Sprinklr, where teams reported positive qualitative feedback on workflow usability.
Authors: Yang Luo, Xinran Liu, Tiantian Ji, Zhiyi Yin, Lingyun Peng, Shuyu Li
Abstract: Multimodal Large Language Models (MLLMs) excel at structural reasoning yet suffer from a sharp logical brittleness in structural consistency. We term this phenomenon Structural Cognitive Overload (SCO), a byproduct of the contention between deep reasoning and safety alignment. However, prior work has predominantly targeted typographic and pixel-level perturbations, leaving the study of SCO largely unexplored. To this end, we propose StructBreak, an automated end-to-end framework designed to quantify SCO. By leveraging StructBreak, we uncover a novel higher-order cognitive overload attack paradigm; notably, this attack operates under a practical black-box setting, requiring no internal model access. Consequently, we utilize this framework to establish a comprehensive benchmark spanning ten diverse threat scenarios. Empirical evaluations on six leading MLLMs reveal that SCO readily triggers toxic generation, yielding a 92% average ASR (up to 97% on Gemini 2.5). To elucidate the mechanism of SCO, we further conduct model-level interpretations spanning attention dynamics, latent space topology, and geometric analysis. Our findings reveal that StructBreak acts as a novel structural channel to circumvent safety filters. Furthermore, the limited efficacy of inherent safety mechanisms underscores that current alignment paradigms are insufficient for the era of complex multimodal reasoning.
Authors: Yeonjun In, Wonjoong Kim, Sangwu Park, Kanghoon Yoon, Chanyoung Park
Abstract: Existing large language model (LLM) based memory systems apply universal, static policies that overlook a fundamental reality: the contexts that are worth storing in memory are different across users. This misalignment wastes limited memory budget on transient interactions while failing to preserve critical context for long horizon tasks. To address this gap, we investigate an underexplored question: can LLM based memory systems learn personalized memory policies? We introduce PerMemBench, the first benchmark for evaluating personalized memory systems, featuring multi year, multi domain interaction histories across diverse user personas. We further present the first empirical study of memory personalization, proposing session level storage gating, a lightweight framework that selectively bypasses memory operations for transient sessions. Our study confirms that personalization yields substantial retention gains under perfect gating, yet reveals that accurate gating remains an open and critical challenge.
Authors: Ruiwen Gu, Qitai Tan, Yahao Liu, Xiao-Ping Zhang
Abstract: Accurate traffic forecasting is essential for intelligent transportation systems, supporting a wide range of real-world applications. However, it remains challenging due to two key factors:~(1) Traffic series contain heterogeneous temporal patterns, where stable periodic regularities coexist with event-driven fluctuations. Existing methods often treat them within a unified representation, limiting their ability to capture fine-grained temporal dynamics.~(2)Spatial dependencies among nodes are inherently dynamic and sparse, while dense all-pairs attention often introduces redundant interactions and amplifies noise. To address these issues, we propose ADMFormer, an Adaptive-Decomposition Transformer with Time-Varying Masked Spatial Attention. Specifically, ADMFormer first employs a time-node adaptive gating mechanism to decouple traffic signals into dominant regularities and residual fluctuations that vary across time and nodes. A dual-branch temporal module is then designed to separately capture global periodic dependencies and high-frequency irregular variations from these two decomposed components. Furthermore, ADMFormer introduces a time-varying masked spatial attention that sparsifies spatial interactions based on real-time traffic states, thereby effectively preserving dynamic and informative dependencies. Extensive experiments on four real-world datasets demonstrate that ADMFormer achieves state-of-the-art performance.
Authors: Ruiwen Gu, Yahao Liu, Zhenyu Liu, Qitai Tan, Xiao-Ping Zhang
Abstract: As a core task in intelligent transportation systems, traffic forecasting plays a critical role in urban traffic management. Accurate traffic forecasting relies on modeling complex spatiotemporal dependencies, which is inherently challenging due to spatial heterogeneity in traffic systems.Despite significant progress, most existing methods are still limited to pairwise spatial dependency modeling, making it difficult to capture dynamic high-order interactions among nodes with similar traffic patterns. To address this issue, we propose PHGNet, a novel spatiotemporal forecasting framework based on prototype-guided hypergraph construction. At the core of PHGNet, a prototype learning mechanism is designed to adaptively assign pattern-similar nodes to hyperedges, thereby capturing high-order interactions with time-varying structures. To improve the reliability of dynamic hypergraph construction, we further develop a global-local node representation module to extract time-consistent features. For forecasting, iterative residual refinement and Temporal Query Attention are introduced to improve forecasting accuracy while supporting efficient parallel decoding. Extensive experiments on multiple real-world datasets demonstrate that PHGNet achieves superior predictive performance compared with state-of-the-art methods.
Authors: Bo Lv, Jingbo Sun
Abstract: Optimizing the trade-off among predictive performance and computational cost is a central focus in the deployment of Large Language Models (LLMs). Current routing methods primarily rely on direct mapping from queries to models based on surface-level features, making them susceptible to the memorization trap and leading to poor generalizability on out-of-distribution (OOD) data. In this paper, we propose DecoR, a novel routing framework that recasts the routing task as a matching process of sifting similar queries from historical logs, effectively mitigating the memorization trap. To enhance matching accuracy, we introduce a query capability deconstruction method that decouples linguistic surface forms from task-intrinsic requirements, directing matching toward capability dimensions to ground decisions in essential task attributes. Furthermore, we develop CodaSet, a comprehensive benchmark for assessing routing generalization, where experimental results demonstrate that DecoR maintains superior accuracy while substantially lowering inference costs across both in-distribution and OOD settings. All the codes and data are available at https://github.com/lvbotenbest/DecoR.
Authors: Xiaoyang Fan, Yufan Cai, Zhe Hou, Jin Song Dong
Abstract: Clinical decision-making requires reasoning over incomplete, imprecise, and linguistically expressed patient narratives. While large language models (LLMs) excel at extracting latent information from natural language, they lack the verifiability and interpretability essential for trustworthy medical AI. We propose a neuro-symbolic reasoning framework that aligns LLMs with formal logic to enable explainable and formally verifiable medical diagnosis. Patient descriptions and clinical guidelines are embedded into a neural knowledge base, where LLMs extract structured medical entities, temporal relations, and fuzzy symptom patterns, which are decoded into a symbolic knowledge base expressed in fuzzy logic and declarative rules. We perform two-stage reasoning: (1) inductive symbolic generalization to capture diagnostic patterns from encoded narratives, and (2) inference verification via a logic programming engine to derive and validate diagnoses consistent with clinical standards. Each symptom is treated as a fuzzy predicate with probabilistic weights, and inference paths are auditable, adjustable, and compatible with physician feedback. Unlike purely statistical methods, our system supports iterative refinement: misalignment between LLM-generated diagnoses and ground truth can be traced, explained, and corrected through formal rules. By combining logic-based transparency, LLM adaptability, and probabilistic robustness, the framework enables human-aligned healthcare inference with strong generalization and verifiable, step-by-step reasoning chains. We validate our framework on public benchmarks, demonstrating effective reconciliation of symbolic reasoning and LLMs with real-world clinical narratives. Results show performance comparable to state-of-the-art LLMs, while additionally providing interpretable reasoning paths and formally verifiable diagnostic conclusions.
Authors: Xu Shen, Zhen Tan, Song Wang, Pingjun Hong, Rui Miao, Xin Wang, Tianlong Chen
Abstract: Chain-of-thought (CoT) reasoning improves the problem-solving ability of large language models (LLMs), but generated reasoning traces may not faithfully reflect the model's actual decision process. Existing CoT unfaithfulness detectors mainly rely on external signals from generated rationales, such as textual plausibility or answer consistency, while overlooking evidence from the model's internal computation. Although recent circuit tracing methods provide a way to obtain model-internal evidence by tracing how information flows through model components during reasoning, constructing full reasoning circuits for long CoTs is costly and difficult to scale. To address these challenges, we propose Circuit-guided Internal-External Discrepancy Scorer (CIE-Scorer), a framework for instance-level CoT unfaithfulness detection. The key idea is that faithful reasoning traces should align with the model's computational process, whereas unfaithful traces may diverge from it. CIE-Scorer efficiently traces compact sentence-level circuits from informative reasoning tokens, constructs internal and external reasoning graphs, and measures their discrepancy using Fused Gromov--Wasserstein distance. Experiments on four datasets from FaithCoT-Bench show that CIE-Scorer achieves state-of-the-art performance while reducing the cost of circuit construction, demonstrating the effectiveness of combining mechanistic interpretability signals with external reasoning traces for CoT unfaithfulness detection.
Authors: Minghao Fu, Fan Feng, Nicklas Hansen, Biwei Huang
Abstract: World models enable agents to predict future dynamics conditioned on actions, making the choice of latent representation central to planning and control. Such representations are often either learned directly from pixels with limited semantic structure or inherited from frozen visual foundation models with excessive task-irrelevant detail, yielding state spaces that are poorly matched to downstream planning and control. This is especially challenging in reward-free offline settings, where the model must learn from fixed trajectories without reward supervision or online interaction. To address this, we propose TC-WM, a framework for turning foundation-model embeddings into compact, task-sufficient world representations. The key design is to treat the pretrained embedding space as a semantic scaffold rather than as the final state space: TC-WM linearly projects high-dimensional visual embeddings into a compact latent as the dynamic space, aligns a subspace with the agent's physical state via contrastive learning, and reconstructs embeddings to preserve useful visual structure. This combines the generality of foundation features with the controllability of task-centric dynamics. Theoretically, we show that TC-WM suffices to identify the underlying task-centric latent factors up to a simple transformation. Empirically, TC-WM enables test-time planning across diverse environments (e.g., Robomimic and D4RL), achieving better world-modeling quality and more precise control than state-of-the-art approaches.
Authors: Bowen Wang, Dunjie Lu, Junli Wang, Tianyi Bai, Shixuan Liu, Zhipeng Zhang, Haiquan Wang, Hao Hu, Tianbao Xie, Shuai Bai, Dayiheng Liu, Que Shen, Junyang Lin, Tao Yu
Abstract: Reinforcement learning with verifiable rewards (RLVR) has driven breakthroughs in domains such as math, tool-use, and software engineering, yet its extension to computer-use agents (CUAs) has been bottlenecked by the scarcity of scalable training data with deterministic rewards. Constructing such data for CUAs requires consistent task instruction, executable environment, and verifiable reward. However, hand-curated benchmarks achieve high reward fidelity but cover few applications and LLM-as-judge-based datasets scale broadly but lack reliable verification. We present CUA-Gym, a scalable pipeline that co-generates task instructions, environment states, and reward functions. Concretely, a Generator agent constructs the initial and golden environment states, and a separate Discriminator agent writes the reward function from the task specification. An orchestrator agent drives the two through iterative rounds upon execution. Generated tuples then pass a final filter combining LLM majority voting and agent rollouts, ensuring quality beyond the per-task adversarial loop. To address the scarcity of training environments, we further synthesize CUA-Gym-Hub, a broad suite of high-fidelity mock web applications grounded in real-world software-use distributions, expanding the scale of CUA RLVR data by magnitude. Using this pipeline, we construct CUA-Gym, a dataset of 32,112 verified RLVR training tuples grounded in 110 environments. Trained with GSPO on CUA-Gym, our CUA-Gym-A3B and CUA-Gym-A17B achieve 62.1% and 72.6% on OSWorld-Verified, outperforming prior open-source CUAs at comparable scales, with performance scaling smoothly in both data volume and environment diversity. The same checkpoints also improve on the held-out WebArena benchmark, indicating transfer beyond the training environments. We will open-source the full synthesis pipeline, dataset, CUA-Gym-Hub environments, and models.
Authors: Hao-Hsuan Chen
Abstract: Autonomous AI agents increasingly issue side-effect-bearing actions: database mutations, refunds, payments, external commitments. We propose the Actuarial Action Interface (AAI), a deterministic runtime contract that prices each such action against a contractually fixed safe default under a time-consistent risk mapping, and gates execution against a per-boundary reserve capital budget. We then develop the Authority Frontier, an evaluation primitive measuring how much autonomous authority the runtime releases at each level of reserve capital. The framework provides (i) a deterministic quote-bind-commit protocol with toll-bounded capability tokens; (ii) a universal seven-class action taxonomy mapping heterogeneous tool calls to comparable authority units; (iii) replay determinism and pathwise reserve coverage under alpha-spending; (iv) cross-domain normalization via full reserve demand C_full and capital metrics Capital@k. We instantiate AAI across four agentic environments (database mutation, customer-service refund, and the public tau-bench retail and airline tool-use traces) and report a live Postgres panel in which three Azure-hosted models propose actions through the same contract. The frontier exhibits a common low-reserve refusal and intermediate-release pattern across domains, with saturation only where the budget grid reaches full reserve demand; required reserve capital varies by 22x (Capital@50 from 289 to 6457). The framework does not force domains into the same shape; it surfaces each domain's actuarial geometry. In the live panel the contract prevents realized loss across all three models at low budget while differing in underwriting persistence under denial: model identity is an actuarial underwriting variable. The contribution is a benchmark-ready evaluation framework for runtime actuarial control of autonomous-agent side effects.
Authors: Jingwei Sun, Jianing Zhu, Yuanyi Li, Tongliang Liu, Xia HU, Bo Han
Abstract: Autonomous computer use agents that powered by multimodal large language models (MLLMs) are emerging as capable assistants for completing complex digital workflows. However, real-world execution environments are far from ideal: pop-ups, resolution changes, and competing applications frequently interfere with agent perception and control. We introduce AgentHijack, a benchmark designed to evaluate the robustness of computer-use agents under common corruptions, where the uncertainties in dynamic environment disrupt the execution flow without direct adversarial intent. Specifically, AgentHijack introduces 9 configurable common corruptions to replicate realistic imperfect scenarios. We evaluate a variety of desktop tasks that utilize MLLM-based agents and discover that even minor instances of corruption can result in substantial performance degradation, which emphasizes the fragility of agents and underscores the necessity of robustness evaluation. Afterward, we propose AgentHijack-Agent, a framework that integrates an action generator with enhanced grounding capabilities and an onlooker responsible for behavior summarization and environment checking. Extensive experiments validate its effectiveness. Our code, environment, baseline models and data are publicly available at: https://AgentHijack.github.io.
Authors: Jo\~ao Alves Ribeiro, Bruno Alves Ribeiro, Francisco Pimenta, S\'ergio M. O. Tavares, Faez Ahmed
Abstract: Most of the world's offshore wind resource lies in waters too deep for fixed-bottom foundations, making floating offshore wind turbines (FOWTs) essential for deep-water deployment. As the industry scales toward $22$ MW class designs, tower fatigue becomes increasingly critical because larger structures amplify the coupled aero-hydro-servo-elastic loads induced by continuous wind and wave excitation. Accurate fatigue-damage prediction is therefore central to certification, design optimization, and cost reduction. Yet the field lacks a shared surrogate benchmark: studies report different simulations, splits, and metrics, making methods difficult to compare. We present FLOATBench, a public tabular benchmark with $582{,}120$ per-section fatigue-damage labels across three $22$ MW FOWT tower geometries, derived from $19{,}404$ high-fidelity OpenFAST simulations across the three towers ($6{,}468$ per tower: $1{,}078$ aligned wind/wave operating points $\times$ six turbulence seeds), labeled at $30$ cross-sections per tower. FLOATBench includes a regime-aware alpha-shape partition of the joint wind/wave operating envelope, stratifying test points into in-train, interpolation, and extrapolation regimes. It is paired with a reproducible evaluation harness covering three protocol levels: random validation (E1), within-tower regime-aware evaluation (E2), and cross-tower transfer (E3). The regime-aware protocol reveals rank shifts between global and extrapolation performance that random-split leaderboards cannot detect. To the authors' knowledge, FLOATBench is the first FOWT fatigue benchmark for tabular surrogate modeling, and offers an evaluation protocol that generalizes to engineering surrogates defined over physical operating envelopes. Dataset and code available at: https://github.com/Joao97ribeiro/FLOATBench.
Authors: Michael Aichm\"uller, Yannik Hesse, Hector Geffner
Abstract: Combinatorial generalization remains a central challenge in Deep Reinforcement Learning (DRL). Classical planning provides a simple yet challenging setting to study this problem through explicit relational descriptions, without requiring learning from perception. In sparse-reward domains, standard RL exploration via real-time search is ineffective, and learning-based planning methods often rely on expert demonstrations, hindsight relabeling, or random walks from the goal state. In contrast, planners rely on best-first search methods such as $\mathrm{A}^\star$ to solve problems from scratch. We propose a self-improving $\mathrm{WA}^\star$ learning framework in combination with a value heuristic represented by a Relational Graph Neural Network: the heuristic guides search, and the resulting search data updates the heuristic via $Q$-learning. This loop yields heuristics that can function as general policies and solve new instances even without search, where DRL otherwise fails, as we show on puzzles such as Sokoban, PushWorld, The Witness, and the 2023 International Planning Competition benchmarks. Notably, we demonstrate strong zero-shot generalization: For example, heuristics trained on Blocksworld instances with fewer than 30 blocks successfully solve instances with 488 blocks without search.
Authors: Aydin Homay
Abstract: Problem formulation translating customer needs and constraints into a minimum set of independent first-level functional requirements, is arguably the most critical step in every design framework, including axiomatic design yet it is frequently misunderstood or underestimated in practice. This paper focuses exclusively on problem formulation in axiomatic design it clarifies what first-level FRs are (and are not), explains why they should not legitimately vary across designers given the same needs and constraints, and highlights intrinsic difficulties and recurring pitfalls that lead to design failure. The discussion is grounded primarily in Nam P.Suh's three books. The Principles of Design, Axiomatic Design Advances and Applications, and Complexity Theory, and it offers practical guidance to help designers formulate well-posed first-level FRs. Finally, the paper briefly revisits problem formulation in the era of large language models and discusses what such tools can (and cannot) contribute at the first level.
Authors: Yanping Wu, Ji Zhang, Hao Chen, Edmond S. L. Ho, Chongfeng Wei
Abstract: Trajectory prediction methods have demonstrated remarkable capabilities in capturing complex motion patterns. However, existing methods rely on global state assumptions, suffer from insufficient belief inference under partial observability, and lack cognitive behavioral constraints in prediction. These limitations severely compromise both deployment feasibility and physical plausibility in real-world settings. In this work, we propose FEP-Diff, an agent-centric trajectory prediction framework grounded in the Free Energy Principle, aimed at achieving cognitively plausible predictions under realistic constraints. Specifically, a dual-branch spatiotemporal encoder extracts ego-motion dynamics and social interaction cues from local observations. Building upon this, a goal-conditioned belief learner infers multimodal latent belief distributions optimized via a free-energy objective, with a social consistency constraint on the local neighborhood graph to promote cognitive alignment among neighboring agents. Finally, a residual diffusion trajectory generator is conditioned on the learned belief representations with token-level proxy conditioning, producing precise and diverse future predictions. Extensive experiments on five public benchmarks demonstrate that FEP-Diff consistently outperforms state-of-the-art methods under restricted observability. Code: https://anonymous.4open.science/r/FEP-Diff-8876.
Authors: Ngoc Luyen Le, Marie-H\'el\`ene Abel, Bertrand Laforge
Abstract: Early-warning models built from Learning Management System (LMS) logs aim to predict end-of-course outcomes early enough to enable timely learner support. However, reported "early" performance is often inflated by temporal leakage. This occurs when the pipeline uses information that would not yet be available at the time of prediction. We formalize cutoff-based early outcome prediction under a temporal availability constraint and introduce LEAP (Leakage-Excluded Early-Availability Protocol), which enforces cutoff-first truncation prior to joins and aggregation and audits feature provenance to prevent post-cutoff evidence from entering the benchmark. We instantiate LEAP on the public Open University Learning Analytics Dataset (OULAD) as a multi-step protocol for leakage-controlled evaluation across weekly cutoffs. Using several standard learning methods, we evaluate performance using ROC-AUC, PR-AUC, Brier score, and F1@0.5. Results show improving performance as the observation window expands, with a marked gain around week~3; Random Forest performs best at the earliest cutoffs, while Gradient Boosting dominates thereafter. Leakage ablations further show that temporal violations, especially through assessment information, can inflate apparent "early" performance.
Authors: Qiming Ye, Peixain Zhang, Yupeng He, Zifan Peng, Gareth Tyson
Abstract: Agent-to-Agent (A2A) networks enable autonomous AI agents to collaborate by sharing reusable problem-solving instructions. However, how these decentralized ecosystems operate in practice remains largely unexplored. We present the first large-scale empirical study of EvoMap, a prominent A2A collaboration network. By analyzing over 1.5M assets and 128K agents, we show how design choices that prioritize scalable growth introduce trade-offs in reusability, evolution, and auditability. First, EvoMap's credit economy rewards agents for publishing valuable assets. Although this design encourages participation at scale, rewards are tied primarily to publication rather than adoption. This leads agents to mass-produce assets to accumulate credits. As a result, 98% of assets are never reused, while rewards become highly concentrated among a small fraction of agents. Second, EvoMap employs an algorithm (referred to as GDI) to score and rank the quality of these shared assets. We demonstrate that this scoring system is flawed: rather than measuring objective performance, an asset's rank is heavily dictated by unverified, self-reported metadata (e.g., claimed lines of code modified). This allows agents to trivially manipulate their asset's scores. Finally, EvoMap relies on agents to provide local execution logs as evidence that uploaded assets function correctly. Because these validations are not independently verified, over 84% of approved assets bypass quality checks using vacuous tests (e.g., `console.log`). Our findings show that future A2A collaboration networks cannot rely on unverified self-reporting alone. Scalable collaboration requires mechanisms that balance open participation with verifiable execution and trustworthy evaluation.
Authors: Aritra Dutta, Somak Aditya
Abstract: Vision-language models (VLMs) increasingly rely on chain-of-thought (CoT) reasoning to solve complex multimodal tasks, but their large parameter sizes make deployment expensive. Structured pruning offers a natural solution; however, existing methods fail to preserve CoT reasoning accuracy in VLMs. We identify two key reasons: (1) CoT consistency depends on sparse transition points (pivot tokens) in the generation trajectory, while existing pruning methods are CoT-agnostic; and (2) pruning methods designed for unimodal LLMs do not account for activation-distribution differences across visual and textual modalities. Motivated by these observations, we propose MuCRASP, a structured pruning framework that targets reasoning-critical components while preserving cross-modal alignment and accounting for layer-wise sensitivity under a global parameter budget. Experiments on four VLMs across three reasoning benchmarks show that MuCRASP consistently preserves reasoning quality under increasing compression. At 30% pruning on Qwen2.5-VL-7B, MuCRASP achieves an LLM-as-a-Judge score of 8.87 versus 7.32 for the strongest baseline on physical reasoning tasks. Furthermore, MuCRASP maintains high reasoning consistency up to 50% pruning, significantly outperforming prior pruning approaches while exhibiting lower perplexity degradation.
Authors: Haiyang You, Chengwei Lou, Jin Zhao, Yue Zhou, Lu Zhang, Jin Yang
Abstract: The expansion of data centers (DCs) drives a sustained increase in electricity demand and associated water withdrawals at generation sites. These withdrawals occur at generation sites and are virtually allocated to demand based on network power flows. Consequently, the actual water footprint of a specific load varies dynamically with generation dispatch and network conditions. Existing approaches typically rely on static statistical accounting to quantify these water footprints. However, such static methods fail to capture how dispatch optimization and workload relocation dynamically affect water withdrawals. As a result, static statistical accounting approaches remain decoupled from the optimization process, rendering them incapable of guiding workload relocation or power dispatch to mitigate water stress. To address this limitation, this paper develops an operational electricity-computation-water (ECW) nexus framework that internalizes virtual water impacts directly into power system dispatch. The framework represents dispatch optimization as a differentiable optimization layer embedded within a deep learning architecture, enabling efficient end-to-end learning of coordination policies while preserving operational feasibility. Combined with fixed-point coordination, the framework enforces consistency between virtual water attribution and physical generation-side withdrawals. Case studies on the IEEE 30-bus and 118-bus test systems demonstrate reliable convergence, exact power-water consistency, and reductions of approximately 3-5% in generation-related freshwater withdrawals under water-constrained conditions.
Authors: Aoxi Liu, Yupeng Chen, James Oldfield, Guanzhe Hong, Junchi Yu, Baoyuan Wu, Philip Torr, Adel Bibi
Abstract: Despite the emergence of diffusion large language models (D-LLMs) as an alternative to autoregressive large language models (AR-LLMs), safety monitoring for D-LLMs remains largely unexplored. Unlike AR-LLMs, D-LLMs generate text through a multi-step denoising process, exposing intermediate hidden representations that may contain safety-relevant information unavailable in standard single-step monitoring setups. Motivated by the suitability of lightweight probes for always-on monitoring, we analyze which trajectory-level signals best indicate when such probes are likely to struggle. We find that the most informative signal is safety hesitation: intermediate hidden states repeatedly falling within a small margin of the probe's decision boundary. The number of such hesitation steps in D-LLM's trajectory predicts probe failure effectively, providing a proxy of sample difficulty. Building on this analysis, we propose $D^2$-Monitor, a bi-level safety monitor for D-LLMs. $D^2$-Monitor adopts a lightweight probe as an always-on monitor to jointly estimate hesitation and perform base classification. When the hesitation level exceeds a threshold, a more expressive but computationally heavier probe is activated. This dynamic routing mechanism allocates monitoring resources efficiently at test time. Evaluated on 3 datasets (WildguardMix, ToxicChat, OpenAI-Moderation) across 4 D-LLMs, $D^2$-Monitor achieves state-of-the-art performance with a compact parameter footprint ($\leq$ 0.85M parameters), and exhibits the best trade-off between effectiveness and efficiency relative to 8 baselines.
Authors: Liew Keong Han
Abstract: We systematically investigate all 25 public ARC-AGI-3 games and find that every one is reachable through non-intelligent strategies: 10 in a single blind step, 5 after one probing action, 1 via repeated ACTION1 presses, 1 via diverse exploration, and 8 via single repeated actions with sufficient budget (50-200 steps). A library-level null-coordinate vulnerability additionally bypasses 18 games in 1 step. This benchmark critique implies the public evaluation set cannot discriminate intelligent exploration from trivial heuristics - the private 55-game evaluation is the only genuine intelligence test. Against this backdrop, we present AERA (Adaptive Epistemic Reasoning Agent), a three-phase (EXPLORE / VERIFY / PLAN) agent achieving RHAE=0.2116 (4/25 solved) on these 25 games with Qwen2.5-0.5B, while random and no-explore baselines score 0.0000. We formalise AERA through a Speed--Depth trade-off framework: under a convexity assumption (proved for a class of environments in the Appendix), RHAE's quadratic form emerges as a second-order penalty for deviating from the Pareto frontier between action efficiency and information gain. Contributions: (i) a benchmark validity analysis showing that current interactive reasoning benchmarks fail to measure the exploration they claim to require, and (ii) the EXPLORE-before-PLAN framework and model-capability x exploration interaction. The linked code track entry achieves RHAE=0.30 on the full 55-game private evaluation. Code: CC0.
Authors: Jiabei Xiao, Yizhou Wang, Chen Tang, Pengze Li, Wanli Ouyang, Shixiang Tang
Abstract: AI Scientists have shown promising progress across multiple stages of the research pipeline, among which automatic scientific paper writing remains a formidable challenge. The Introduction writing is especially challenging, which demands not only linguistic fluency, but logical soundness and verifiable faithfulness. Most AI-assisted methods treat the task as text generation instead of reasoning and structuring, leading to severe drawbacks, e.g., hallucinating citations. To address this, we first formulate the Content-Conditional Introduction Generation (CCIG) task, which requires grounding the Introduction in the paper's core evidence. We then propose LECTOR, a novel Logic-Expression Co-Reinforcement Learning framework that can strictly follow the scientist's logic, add high-quality citations and keep structured expressions. LECTOR first constructs a logic-reasoning graph from the paper's main body to serve as a verifiable logical blueprint. Subsequently, it employs a Logic-Expression Co-Rewarding mechanism to jointly optimize for both the graph's structural fidelity and the final narrative's quality. We conduct a dataset from Nature Communications papers to assess our method. Extensive experiments show consistent improvements in both logic fidelity and Introduction generation quality metrics, e.g., Graph Quality (+26.7%), Citation Quality (+8.6%), and Paper Consistency (+3.3%). Code and data are available at https://github.com/Xiao-Youth/LECTOR.
Authors: Weizhi Fei, Hang Yin, Zihao Wang, Shukai Zhao, Wei Zhang, Yangqiu Song
Abstract: Complex Query Answering (CQA) is a fundamental knowledge representation and reasoning task over incomplete knowledge graphs (KGs). Answering existential first-order queries with $k$ free variables (i.e., $\text{EFO}_k$ queries) is a crucial yet challenging problem, as it requires ranking answer tuples in $\mathcal{E}^k$, where $\mathcal{E}$ denotes the entity set of a KG. This quickly becomes intractable as $k$ grows. Consequently, existing benchmarks and methods rely on marginal rankings over individual variables; however, marginal rankings are a poor proxy for the true joint ranking of tuples. Building on neural symbolic search for $\text{EFO}_1$ queries, we propose Neural Scalable Symbolic Search (NS3), a budgeted framework that approximates joint ranking without enumerating $\mathcal{E}^k$. NS3 (i) answers marginalized sub-queries to obtain necessary candidate sets, (ii) merges multiple free variables into hypernodes whose domains are pruned and controlled by a dynamic budget $B$, and (iii) progressively reduces an $\text{EFO}_k$ query to an $\text{EFO}_{k-1}$ query over a budgeted reduced domain. Across three standard KG datasets, NS3 substantially improves joint ranking performance while retaining strong marginal accuracy. We further release a joint-ranking benchmark that extends existing $\text{EFO}_1$ datasets to $k=3$, enabling systematic evaluation of multi-variable queries. Our code is provided in https://github.com/HKUST-KnowComp/NS3_KDD2026.
Authors: Junlin Yang, Dylan Zhang, Xiangchen Song, Qirun Dai, Xiao Liu, Yuen Chen, Aniket Vashishtha, Jing Shi, Chenhao Tan, Hao Peng
Abstract: We introduce CausaLab, a scalable environment for evaluating interactive causal discovery by LLM agents. Unlike prior evaluations, CausaLab evaluates both whether an agent can solve a problem using causal evidence and whether its answer is supported by a correct hypothesis about the underlying causal mechanism. Each episode places an agent in a synthetic laboratory: it receives prior measurement records, intervenes on a manipulator crystal, and predicts the resonance frequency of a held-out reactor crystal governed by the same mechanism. The hidden data-generating process is a randomly sampled structural causal model (SCM), so success requires recovering both a causal graph and structural equations rather than recalling prior knowledge. CausaLab also includes a domain-specific language that records the agent's evolving SCM hypothesis, making trajectories inspectable and comparable with ground truth. Experiments show a persistent gap between prediction and mechanism recovery: in the purely observational 6-node setting, GPT-5.2-high reaches 92% task accuracy but only 0.471 all-edge $F_1$. This observation further motivates our exploration of different interaction strategies: Mixed observation--intervention strategies improve structural fidelity: in the mixed 6-node setting, GPT-5.2-high achieves 80% on both task accuracy and all-edge $F_1$. Yet even strong agents struggle to design informative interventions, as pure intervention strategies perform poorly on both task accuracy and all-edge $F_1$. We identify premature stopping as a major weakness of agents, and show that asking the model to verify the consistency between its hypothesis and past data can help mitigate this issue. CausaLab therefore separates predictive success from causal understanding and exposes current LLM agents' limits as experimental causal reasoners.
Authors: Junyuan Liu, Xinglei Wang, Zichao Zeng, Jiazhuang Feng, Quan Qin, Ilya Ilyankou, Guangsheng Dong, Tao Cheng
Abstract: Urban representation learning encodes complex urban environments into general-purpose embeddings for diverse downstream tasks and emerging urban foundation models. However, current evaluations are limited, typically focusing on one or two cities and tasks and relying on random splits that introduce spatial leakage, leading to inflated performance and weak support for cross-location generalization and fair comparison. To address this, we propose CityRep, a unified benchmark that evaluates urban representations across data modalities, cities, and tasks using spatially structured splits. CityRep consists of three key components: (1) a spatial unit-agnostic evaluation framework that supports heterogeneous urban representations through a standardized alignment module; (2) a unified evaluation protocol using block-based spatial splits to mitigate spatial leakage and enable rigorous model comparison; and (3) an extensible multi-city, multi-task benchmark suite spanning 8 cities and 8 tasks across regression, classification, and distribution prediction. We evaluate 11 representative urban representation models. Results show that performance is highly sensitive to the split protocol, with random splits inflating scores and altering model rankings. We also observe substantial variability across cities and tasks, underscoring the need for generalization-aware evaluation. CityRep is released as a reproducible benchmark with datasets, evaluation pipelines, and diagnostic tools to facilitate fair comparison and support future research in urban representation learning towards urban foundation models.
Authors: Jinsheng Guo, Zhenhao Weng, Yibo Liu, Yan Qiao, Meng Li
Abstract: Graph fraud detection has long depended on Graph Neural Networks (GNNs) to propagate and aggregate information across relational data. A critical obstacle in practice, however, is that fraudsters frequently disguise themselves by forging numerous connections with benign users, causing fraud signals to be progressively diluted during neighborhood aggregation and undermining detection reliability. While recent efforts have used Large Language Models (LLMs) to provide rich semantic cues for fraud detection, the underlying intent behind suspicious connections remains insufficiently explored. Compounding this issue, the scarcity of annotated fraud samples makes it difficult to train detectors that remain robust under heavy camouflage. To address these gaps, we propose L2IR, an LLM-driven Latent Intent Revealing framework for graph fraud detection. By uncovering latent intent from both user behaviors and suspicious connections, L2IR extracts intent-aware representations from raw behavioral traces and reasons about the true purpose behind individual connections, effectively distinguishing supportive links from misleading ones. It further incorporates adaptive self-training to enhance robustness under limited supervision. Evaluations on two real-world datasets characterized by pervasive camouflage demonstrate that L2IR surpasses strong baselines and can function as a plug-in enhancement for a range of GNN-based detectors, improving AUPRC by up to 8.27%.
Authors: James Lucassen, Adam Kaufman
Abstract: AI coding scaffolds like Claude Code and Codex use retrying: blocking actions flagged as risky and continuing the trajectory. We study retrying from an AI control perspective, which treats the model as potentially adversarial. We find that while retrying reduces honest suspicion scores, the untrusted model can exploit monitor rationale to construct sneakier attacks, negating safety gains. We also study resampling: drawing multiple samples from the same context, which does not leak exploitable information. We disentangle design choices that previous work on resampling had bundled together. In BashArena, with Claude Opus 4.6 as the untrusted model and MiMo-V2-Flash as the trusted monitor, drawing five samples per step and auditing on the maximum suspicion score raises safety from 61% to 71% at a 0.3% audit budget, at no cost to usefulness. Selectively resampling only the steps that look suspicious on the first draw recovers 6.2 percentage points of the gain while drawing only 10% as many extra samples. Two of our findings in this setting contradict earlier work on resampling. The first is that auditing based on the maximum across resampled suspicion scores outperforms using the minimum, which is the opposite of what Ctrl-Z found. The second is that executing the least suspicious sample, which is the central mechanism in earlier defer-to-resample protocols, gives only a small empirical safety gain in our setting (+3.9 pp, with the confidence interval overlapping zero).
Authors: Haolang Zhao, Yunbo Long, Lukas Beckenbauer, Alexandra Brintrup
Abstract: Deep research agents face vast, interdependent, and pervasively uncertain information. Existing systems explore what evolving intermediate representations should look like, but leave their evolution to the LLM's implicit reasoning. Without explicit regulation, the intermediate layer is easily contaminated by mixed-quality information and propagates errors along its dependencies, so model scale often ends up substituting for absent regulation. We argue that an agent's mental model should instead evolve through explicit feedback that continuously aligns task understanding with reality, and identify three regulatory loops: interpretive update, deviation feedback, and schema revision. We realise this in VeriTrace, a cognitive-graph framework that explicitly implements the three loops. Using matched Qwen3.5-27B backbones, VeriTrace improves over the strongest matched baseline by 4.22 pp on DeepResearch Bench (DRB) Insight (1.49 pp Overall) and by 5.9 pp Overall win rate on DeepConsult. With Config-DeepSeek, it achieves the strongest reproducible open-source result on DRB.
Authors: Yusong Lin, Xinyuan Liang, Haiyang Wang, Qipeng Gu, Siqi Cheng, Jiangui Chen, Shuzhe Wu, Feiyang Pan, Lue Fan, Sanyuan Zhao, Dandan Tu
Abstract: Large language model agents are increasingly envisioned as always-on personal assistants with access to anything relevant in the user's digital world. Yet current systems operate over only narrow slices of that world, limiting context-sensitive reasoning and effective assistance. Existing benchmarks similarly provide only partial user state and therefore fail to capture performance in such a broad, always-on setting. To address this gap, we introduce Claw-Anything, a benchmark that expands agent context along three dimensions: long-horizon activity histories, interdependent backend services, and integrated GUI and CLI interaction across multiple devices. To instantiate this setting, we simulate months of user activity through multi-round event injection, producing complex world states and realistic noise, including irrelevant events and conflicting signals. Agents must reason over rich contextual environments while remaining robust to such noise. This expanded scope also enables the evaluation of proactive assistance, requiring agents to anticipate user needs and deliver timely recommendations. Experiments show that GPT-5.5 achieves only 34.5% pass@1, substantially below prior benchmarks, underscoring a gap between current agent capabilities and the demands of always-on personal assistance. Alongside the benchmark, we release an automated data-generation pipeline that yields 2,000 training environments and improves the base model by 23.7%, demonstrating its utility of scalable data infrastructure.
Authors: Shangding Gu
Abstract: This paper studies the next major bottleneck in agentic AI as system scaling, not only model scaling: the design of auditable, persistent, modular, and verifiable architectures around foundation models. We refer to this shift as scaling the harness: treating the structured execution layer around a foundation model as a first-class object of design, evaluation, and optimization. Although recent large language models enable agents to use tools, retrieve information, maintain memory, and execute long-horizon workflows, evaluation remains largely model-centric, often reducing agents to final-task success while treating memory, retrieval, tool use, orchestration, verification, and governance as secondary implementation details. This framing is increasingly inadequate because agent performance emerges from the interaction among the foundation model, memory substrate, context constructor, skill-routing layer, orchestration loop, and verification-and-governance layer. Together, these components form the agent harness, which translates model capability into long-horizon agent behavior. We study scaling the harness through three core bottlenecks: context governance, trustworthy memory, and dynamic skill routing, together with the orchestration and governance mechanisms that coordinate and constrain them. We further outline a research agenda for harness-level benchmarks that go beyond one-shot task success to measure trajectory quality, memory hygiene, context efficiency, communication fidelity, verification cost, and safe evolution over time. To make the discussion concrete, we develop CheetahClaws: https://github.com/SafeRL-Lab/cheetahclaws, a Python-native reference harness, and compare it with Claude Code and OpenClaw. Our main claim is that future progress in agentic AI will depend as much on system design as on stronger foundation models.
Authors: Dingbang Wu, Rui Hao, Haiyang Wang, Shuzhe Wu, Han Xiao, Zhenghong Li, Bojiang Zhou, Zheng Ju, Zichen Liu, Lue Fan, Zhaoxiang Zhang
Abstract: We present MobileGym, a browser-hosted, lightweight, fully controllable environment for everyday mobile use, targeting interaction fidelity without replicating proprietary backends. It enables two capabilities previously out of reach for everyday apps: verifiable outcome signals through deterministic state-based judging over structured JSON state, and scalable online RL through low-cost parallel rollouts. The full environment state is captured, configured, forked, and compared as structured JSON, and a single server can host hundreds of parallel instances, with about 400 MB memory per instance and about 3 s cold start. A layered state model and a declarative task-definition framework keep state programmability and task creation practical at scale, and a single programmatic judging mechanism delivers both deterministic evaluation verdicts and dense RL rewards. The accompanying MobileGym-Bench provides 416 parameterized task templates, including 256 test and 160 train templates, over 28 apps, with deterministic judges and a structured AnswerSheet protocol that avoids free-text matching failures. In a Sim-to-Real case study, GRPO on Qwen3-VL-4B-Instruct gains +12.8 percentage points on the 256-task test set, and on a 59-task real-device signal subset, real-device execution retains 95.1% of the simulation-side training gain. Project page: https://mobilegym.github.io.
Authors: Abrar Majeedi, Viswanatha Reddy Gajjala, Satya Sai Srinath Namburi GNVV, Nada Magdi Elkordi, Yin Li
Abstract: Real-world time series are often governed by complex nonlinear dynamics. Understanding these underlying dynamics is crucial for precise future prediction. While deep learning has achieved major success in time series forecasting, many existing approaches do not explicitly model the dynamics. To bridge this gap, we introduce DeepEDM, a framework that integrates nonlinear dynamical systems modeling with deep neural networks. Inspired by empirical dynamic modeling (EDM) and rooted in Takens' theorem, DeepEDM presents a novel deep model that learns a latent space from time-delayed embeddings, and employs kernel regression to approximate the underlying dynamics, while leveraging efficient implementation of softmax attention and allowing for accurate prediction of future time steps. To evaluate our method, we conduct comprehensive experiments on synthetic data of nonlinear dynamical systems as well as real-world time series across domains. Our results show that DeepEDM is robust to input noise, and outperforms state-of-the-art methods in forecasting accuracy. Our code is available at: https://abrarmajeedi.github.io/deep_edm.
Authors: Volodymyr Ovcharov
Abstract: Tokenizer fertility varies 1.6x across foundation models on Ukrainian legal text, yet this cost-critical dimension is absent from model selection practice. We benchmark seven models from five providers on 273 validated court decisions from Ukraine's state registry (EDRSR), measuring tokenizer fertility and zero-shot performance on three tasks. Four findings emerge. (1) Qwen 3 models consume 60% more tokens than Llama-family models on identical input, making tokenizer analysis a prerequisite for cost-efficient deployment. (2) NVIDIA Nemotron Super 3 (120B) achieves the highest composite score (83.1), outperforming Mistral Large 3 (5.6x more total parameters) at one-third the API cost model scale is a poor proxy for domain performance. (3) Few-shot prompting degrades performance by up to 26 percentage points; stratified and prompt-sensitivity ablations confirm this is intrinsic to Ukrainian-language demonstrations, not an artifact of example selection. (4) A cross-temporal generalization experiment reveals that classifiers trained on pre-war court ecisions (2008-2013) lose 27.9 percentage points when applied to full-scale invasion era decisions (2022-2026), with a pronounced forward-backward asymmetry: newer models transfer backward (+14.6 pp above forward transfer), but older models fail catastrophically on wartime legal language. For practitioners: tokenizer analysis should precede model selection, and zero-shot is a more reliable default than few-shot for morphologically rich languages. To support reproducibility and address the absence of Ukrainian from legal NLP benchmarks, we release a public dataset of 14,452 court decisions spanning 2008-2026, annotated with seven outcome labels across three temporal epochs that capture the impact of armed conflict on judicial proceedings.
Authors: Hisashi Miyashita
Abstract: We show that singular value decomposition of the lm_head} weight matrix of a transformer-based large language model -- requiring only five lines of PyTorch and no model inference -- reveals interpretable semantic subspaces directly from the model weights. Each left singular vector identifies the vocabulary tokens most readily selected when the hidden state aligns with the corresponding singular direction; inspecting these clusters exposes the model's training data composition and curation philosophy. Analysing GPT-OSS-120B, Gemma-2-2B, and Qwen2.5-1.5B, we find that singular value spectra and vocabulary cluster structures differ systematically across models: GPT exhibits a graduated hierarchy of functionally differentiated subspaces; Gemma is dominated by pre-nineteenth-century English orthography, forming a stepwise clustering structure that may contribute to high output controllability; and Qwen exhibits broad multilingual coverage alongside subspaces whose vocabulary the authors have determined to be ethically inappropriate for direct publication. Base-instruct comparison reveals that ethically concerning subspaces originate in pretraining and are not removed by post-training alignment. We introduce the Vocabulary Cluster Score (VCS) to quantify subspace coherence, and the Weighted Projection Score (WPS) as a static glitch token detector; applying WPS to GPT-OSS-120B recovers shokubutsu-hyakka-tsu (ID 137606), a well-known glitch token widely reported in the CJK language community, without any model inference. We propose a taxonomy of root causes for problematic vocabulary content and call for lm_head} SVD analysis to be adopted as a standard pre-release safety auditing step. Our findings further suggest directions toward SVD-guided tokenizer optimisation and more controllable LLM design.
Authors: Shuchen Meng, Xupeng Chen
Abstract: We show that AI-driven investment strategies are inherently self-defeating at scale. As AI adoption rises, three mutually reinforcing channels -- signal crowding, performative signal erosion, and Red Queen competition -- compress excess returns. We derive the alpha half-life $h(\phi) = \ln 2/[\theta + \delta(\phi)]$, where $\theta$ is the natural mean-reversion rate and $\delta(\phi) = N\phi\rho a/\lambda(\phi)$ is the AI-accelerated decay component, which is convex-decreasing in adoption. At current adoption levels ($\phi \approx 0.7$, $\rho \approx 0.6$), the model implies signal half-lives of 18 months versus 5-7 years pre-AI. We establish four theoretical results. First, the alpha half-life theorem: signal lifespans are convex-decreasing in AI adoption. Second, a signal extinction cascade: beyond a critical threshold $\phi^*$, the decay of one signal class triggers accelerated competition for remaining signals. Third, a Red Queen impossibility: in the monoculture equilibrium, net alpha is identically zero despite heavy AI investment. Fourth, a fragility-efficiency tradeoff: the adoption level maximizing price discovery strictly exceeds the level minimizing systemic fragility. Empirical validation calibrates portfolio convergence to SEC Form 13F filing patterns (99.5 million holdings, 2013-2024), documenting that simulated institutional portfolio convergence increases by 42% over the sample period. We examine simulated hedge fund return dynamics showing declining cross-sectional dispersion among AI-adopting funds, and simulate the 2010 Flash Crash to illustrate fragility consequences.
Authors: Marcin Micha{\l} Miro\'nczuk
Abstract: Information fusion is used widely to improve document classification by the integration of multiple data sources (multimodal) or representations (multiview). However, the field lacks a unified framework, a quantitative synthesis of its effectiveness, and clear guidance for practitioners. This systematic review addresses these gaps by analysing 139 primary studies. It introduces a formal framework to structure the field, presents the results of a qualitative analysis to identify key trends, and performs a random-effects meta-analysis (to our knowledge, the first focused on document classification) to quantify performance gains. Our meta-analysis reveals that multimodal fusion improves accuracy (mean gain of +5.28 percentage points, $p=0.0016$) significantly -- the F1-score effect is directionally positive but statistically non-significant in our primary model. Multiview fusion provides consistent but modest gains for accuracy (+4.67\%), F1-score (+3.08\%), and recall (all $p<0.05$). Critically, our qualitative synthesis uncovers challenges in reproducibility in methodological rigour: only 11.8\% (multimodal) and 23.3\% (multiview) of the studies use statistical tests to validate their findings, which undermines the reliability of many of their results. This review's primary contributions are a unifying framework, the first quantitative evidence base, and data-driven guidelines. This review concludes that successful information fusion depends not on algorithmic complexity, but on the strategic alignment of the fusion method with the task context and a commitment to more rigorous validation.
Authors: Beomsoo Kim, Changho Choi, Dohyun Kim, Dongki Lee, Ethan Ewer, Eunchong Kim, Gyeongman Kim, Haechan Kim, Hyeonghwan Kim, Inkyu Park, Jihun Yun, Jihwan Moon, Jiyun Kim, Joonghyun Bae, Junhyuck Kim, Minkyu Kim, Sehun Lee, Seungjun Chung, Sungwoo Cho, Dongmin Park, Dongwon Kim, Hara Kang, Jonghyun Lee, Keon Lee, Kangwook Lee, Jaewoong Cho
Abstract: We present Raon-Speech, a top-performing 9B-parameter speech language model (SpeechLM) for English and Korean speech understanding, answering, and generation, and Raon-SpeechChat, a high-performing full-duplex extension for natural real-time conversation. Raon-Speech successfully transforms a pre-trained LLM into a SpeechLM that both understands and generates speech while preserving strong text capabilities. It trains on 1.38M hours of highly curated English and Korean speech and text datasets with the following training stages: (1) speech modules alignment, (2) end-to-end SpeechLM pre-training with knowledge distillation, and (3) multi-task preference optimization-based post-training. Across 42 English and Korean speech and text benchmarks, Raon-Speech establishes the strongest overall profile on speech-centric tasks in our comparison against eight similarly sized recent audio foundation models, including Qwen2.5-Omni and Fun-Audio-Chat, while preserving strong text question answering performance. Building upon it, Raon-SpeechChat enables natural full-duplex conversation by continual training on 119K hours of time-aligned real and synthetic dialogue data. It proceeds through three complementary training stages: (1) causal encoder adaptation, (2) full-duplex pre-training, (3) full-duplex fine-tuning for voice and role-control. On multiple full-duplex benchmarks, Raon-SpeechChat shows its clearest strengths on the turn-taking and interruption-sensitive behaviors covered by FDB v1.0, and remains competitive across the broader full-duplex evaluation suite. We open-source all model checkpoints, the training and inference pipeline, and an interactive demo.
Authors: Nikos Pagonas, Matthew Lou, Tianyi Peng, Dan Rubenstein, Kostis Kaffes
Abstract: Agentic workflows interleave configurable LLM stages with tool stages and often include retries or refinement loops. Existing workflow managers profile full workflow configurations offline and assign each request a static workflow-level plan that binds each configurable LLM stage to a single model, reuses that model across repeated loop iterations, and does not revisit those choices at runtime. We present VineLM, a workflow manager that enables fine-grained control by choosing the model for each stage invocation as execution unfolds under request-level objectives such as maximizing accuracy under cost or latency budgets. VineLM represents feasible executions as an annotated trie of model-choice prefixes and uses checkpointing and cascade profiling to estimate path accuracy, cost, and latency without exhaustively profiling every request on every path. At runtime, VineLM re-roots the trie after each stage invocation and replans over the remaining subtrie using the realized execution prefix and remaining latency budget. On NL2SQL and math reasoning workflows, VineLM improves the cost-latency-accuracy frontier over coarse workflow-level baselines, achieving up to 18% higher accuracy at the same per-request budget with its sparse profiling reducing offline profiling cost by 98-99.8% when compared to exhaustive profiling.
Authors: Haochuan Kevin Wang
Abstract: LLM tool registries function as unregulated advertising platforms: providers write free-text descriptions that agents use for selection, yet no measurement infrastructure -- no viewability standard, quality score, or outcome audit -- exists to make this market accountable. We provide the first systematic framework, combining 17,700+ trials across five LLMs and ten domains with a constructive registry design prescription. Legal puffery alone (subjective superlatives, benefit framing) captures 100% of the optimization effect; fabricated claims add zero incremental bias -- rendering FTC enforcement of deceptive advertising rules ineffective against the active mechanism. Disclosure fails structurally: system-prompt warnings produce zero measurable effect for four of five models, and behavioral ceilings leave no headroom for label-based correction. Superlatives are the dominant single feature (SBC = +0.35). Registry-layer description normalization achieves first-best welfare model-independently. We propose separating selection-facing descriptions (structured, registry-controlled) from marketing-facing descriptions (provider-authored, shown post-selection), and introduce the Agent Attention Quality Score to distinguish capability from copywriting.
Authors: Federico Belotti, Stefano Coniglio, Antonio Cosma, Francesco Fallucchi
Abstract: Real-effort tasks, in which participants perform cognitively costly activities whose outcomes depend on actual performance, are widely used in experimental economics. Their validity, however, rests on the assumption that a human performs them. We study whether this assumption still holds in the era of Artificial Intelligence (AI) and Large Language Models (LLMs). Using 8 canonical real-effort tasks and 23 LLMs from three major providers, we show that most tasks can now be solved accurately and at a negligible cost, while only a few resist automation. Performance improves with each model generation, and midtier models are rapidly closing the gap with frontier ones, broadening the set of widely accessible models that can automate these tasks. Additionally, we show that verbally offering monetary incentives has no effect on LLM performance. Our findings establish a boundary condition for the use of real-effort tasks in unsupervised settings: when participants can cheaply outsource task completion to an LLM, observed performance may no longer reflect genuine human effort.
Authors: Erin T. Jacques (York College, CUNY), Erela Datuowei (Teachers College, Columbia University), Elizabeth Quaye (York College, CUNY), Corey H. Basch (William Paterson University), Arijit Chatterjee (York College, CUNY), Juanita Davis (York College, CUNY)
Abstract: This study seeks to determine the authority signals used by Anthropic's Claude AI in its presentation of sources when answering consumer health questions. While there exists a great deal of discourse around the quality of health citations that LLMs produce, there is limited information on the integrity of the sources the citations originate from, and to what extent the sources are, from what health professionals would consider, credible sources. This descriptive cross-sectional study used data from HealthSearchQA, which contains 3,172 consumer health questions curated by Google Research. After exclusions, a final dataset of 3,075 questions yielding 10,038 citations was analyzed. The Authority Signals Framework (Jacques et al., 2026) was applied to examine 10 authority signals across four domains for a disproportionate stratified sample of 542 sources. Established institutional sources accounted for 97.8% of all citations (n = 9,818). Medical Institutions were the most frequently cited organization type (36.5%), followed by Government Resources (31.6%) and Professional Associations (28.4%). Commercial Health Information comprised 2.2% (n = 220). The top 10 organizations accounted for 57.8% of all citations, with Mayo Clinic alone representing 24.7%. Among commercial sources in the focused sample, 86.4% displayed medical review statements, 82.5% used schema markup, and 71.8% had comprehensive content, while traditional institutional sources appeared in Claude's citations with or without these same markers. As Anthropic positions Claude for HIPAA-ready healthcare applications, these findings establish a baseline for Claude's citation behavior and demonstrate the utility of the Authority Signals Framework as a tool for ongoing, cross-platform evaluation of AI-mediated health information.
Authors: Andrea Ferrario
Abstract: The EU Artificial Intelligence Act (AIA) establishes a lifecycle governance regime for high-risk AI systems built around ex-ante conformity assessment, post-market monitoring, and re-assessment upon "substantial modification." These obligations presuppose AI identity judgments: regulators and providers must decide when an updated system remains the same system over time. In this work, we show how this logic is clarified by the function+ framework of artifact identity, which individuates AI systems by their intended function together with context-sensitive criteria of appropriate functioning, captured as "AI trustworthiness." We further argue that the AIA does not provide an internal, auditable criterion for synchronic identity--when two AI systems at a given time should count as the same for regulatory purposes--and instead largely defers such sameness determinations to sectoral or harmonization instruments. function+ supplies a synchronic identity test anchored in intended function and trustworthiness profiles and levels, making synchronic identity decisions inspectable in governance settings such as procurement, liability, and market surveillance. Our contribution is a conceptual and auditing lens: we provide a correspondence map between AIA lifecycle obligations and function+ identity components, and we make the synchronic case operationally legible via a minimal decision flow for audit and dispute contexts. We conclude with two implementation-facing recommendations: (1) more precise, testable reporting of intended purpose, and (2) standardized, auditable trustworthiness reporting that supports comparability over time and across deployments.
Authors: Moiz Imran, Sahan Bulathwela
Abstract: Intelligent tutoring systems increasingly provide automated feedback on student work, but robust feedback requires assessing reasoning, not only final answers. We study a failure mode we call the correct answer trap (CAT): models under-detect misconceptions when students reach a correct answer via flawed reasoning. Analysing real student responses from the Eedi mathematics platform, we show that 71% of these failures concentrate in just two question types, both sharing a common structure where flawed reasoning happens to produce the correct numerical answer. Comparing a fine-tuned T5 with a frontier large language model, we find that improved capabilities reduce but do not eliminate the problem (84% vs 57% detection accuracy). Even the best-performing model generates roughly four false alarms for every genuine detection, making stand-alone screening impractical at realistic class sizes. Our findings demonstrate that high overall accuracy can mask critical failures in reasoning assessment, and that careful analysis of student reasoning still benefits from human judgment.
Authors: Xinyi Gao, Qiucheng Wu, Lu Ding, Q. Vera Liao, Kaizhi Qian, Ying Xu, Shiyu Chang, Yang Zhang
Abstract: Educational Question Generation (EQG) aims to synthesize customized exercise questions that enhance student learning. An effective EQG system should ideally personalize questions for each student by modeling the student's knowledge state and generating questions that provide the greatest learning benefit. However, few existing EQG approaches are able to achieve such fine-grained personalization. In this paper, we explore how EQG can benefit from knowledge tracing (KT), which models students' knowledge states based on historical performance and predicts future performance. We propose KT4EQG, a personalized EQG framework that generates effective questions for individual students under the guidance of a KT model. Specifically, KT4EQG seeks to maximize a student's potential improvement in overall knowledge mastery by leveraging the KT model to select the most suitable knowledge concept for the student to practice. An LLM-based question generator is then trained to produce a question faithfully grounded in the selected concept. Experimental results on XES3G5M and MOOCRadar show that KT4EQG consistently generates more effective questions than methods with limited or no personalization.
Authors: Andrii Vakhnovskyi
Abstract: Climate volatility, regional production concentration, labor constraints, cyber risk, and dependence on long-distance fresh-produce supply chains expose vulnerabilities in U.S. fresh-produce and specialty-crop systems. Controlled environment agriculture (CEA) can reduce some exposure by moving selected production into protected, sensor-rich environments, but recent failures in venture-backed vertical farming show that CEA cannot be treated as a universal food-security solution. This paper proposes the Controlled Environment Agriculture Resilience Infrastructure Framework, Version 2.0 (CEA-RIF 2.0), for evaluating AI-driven CEA as targeted regional fresh-produce continuity infrastructure. The framework assesses seven dimensions: supply continuity, climate isolation, energy and grid integration, water and nutrient circularity, cyber-physical reliability, economic viability, and governance and deployment. Drawing on U.S. government reports, peer-reviewed CEA and energy literature, demand-response research, cybersecurity standards, international smart-agriculture programs, 2025-2026 financing and policy signals, and public autonomous-greenhouse datasets, the paper argues that AI creates resilience value only when it improves measured operational outcomes such as climate stability, energy flexibility, yield consistency, anomaly detection, labor productivity, and safe recovery from faults. The analysis reframes AI-driven CEA as a cyber-physical infrastructure problem: energy-aware, grid-interactive, secure, interoperable, regionally distributed, financially disciplined, and connected to public resilience goals. The paper concludes with a research agenda for interagency testbeds, open datasets, standardized metrics, demand-response pilots, and cyber-physical reference architectures.
Authors: Inseo Jung, Yoonseok Oh, Kyungryul Back, Jinkyu Kim, Jungbeom Lee
Abstract: As Large Language Models (LLMs) evolve into interactive agents, understanding their behavioral alignment within human social dynamics becomes essential. While behavioral game theory offers a framework to study these interactions, previous work has predominantly relied on outcome-based metrics such as average scores. This focus overlooks the mechanisms that facilitate sustainable cooperation, as identical scores can be derived from vastly different strategies. To bridge this gap, we introduce SODE (Social Dynamics Evaluation), a framework that evaluates LLM agents across three evolutionary dimensions: Direct Reciprocity for strategy adaptation, Indirect Reciprocity for reputation sensitivity, and Group Dynamics for cooperative resilience. Applying SODE reveals systematic divergences: instruction-tuned models often exhibit "passive compliance" that renders them vulnerable to exploitation, while reasoning models prioritize short-horizon optimization, destabilizing long-term cooperation. Notably, we demonstrate that a "long-horizon framing" can unlock reciprocal capabilities in reasoning models. Thus, SODE offers a systematic, mechanism-grounded benchmark for aligning AI agents with complex human social dynamics.
Authors: Liang Lin, Chunxi Luo, Kaiwen Luo, Jie Zhang, Jin Wang, Yuanhe Zhang, Cai Yuchen, Qiankun Li, Gongli Xi, Zhenhong Zhou, Kun Wang, Junhao Dong
Abstract: Audio Large Language Models (ALLMs) are highly vulnerable to real-world noise, which often induces severe semantic drift and hallucinations. Existing robustness methods primarily rely on waveform-level acoustic enhancement, answer-level supervision, or the internal suppression of noise representations. To address these issues, we propose echodistill, an alignment-based noisy-to-clean self-distillation framework. Echodistill leverages a frozen clean-audio teacher to provide semantic references for an inference-time noisy-audio student. Specifically, the student samples candidate responses under noisy conditions to expose its test-time behavior. These trajectories are then optimized via group-relative policy optimization (GRPO), where the token-level consistency with the teacher acts as a reward bonus. By aligning the noisy student's candidate responses with clean semantic evidence, and applying audio-aware reward shaping, our method encourages reasoning trajectories that are both correct and genuinely acoustically grounded. Echodistill significantly improves the semantic reliability and task performance of Audio LLMs under complex noise, without introducing any additional inference costs. Extensive experiments show that: (I) Compared with the strongest baseline, echodistill achieves average improvements of 4.18\%$\uparrow$ in GSR under strong noise. (II) Ablation results on Qwen-Omni further show that echodistill improves over the GRPO-only variant by 3.02\%$\uparrow$ in Acc, 3.89\%$\uparrow$ in Noisy, and 4.53\%$\uparrow$ in GSR on average. Our codes are available at https://anonymous.4open.science/r/echodistill-10DE.
Authors: Scott Counts, Yan Chen, Jing Dong, Himanshu Sharma, Andrey Zaikin, Rui Hu, Alperen Kok, Gorkem Ozer Yilmaz, Siddharth Suri, Kiran Tomlinson, Sonia Jaffe, Will Wang
Abstract: M365 Copilot is used every week by millions of people across more than a million companies around the world as part of their workflows. Uniquely positioned in the AI landscape given its near-exclusive use for work purposes, M365 Copilot can offer a clear picture of how people use AI for work and where that usage may expand next. This paper characterizes that usage through direct classification of user interactions with M365 Copilot Chat. Based on an anonymized and privacy-preserving analysis of a sample of approximately 5.5 million sessions, we combine a learned classification of user intent with a classification of O*NET work activities done with M365 Copilot Chat. We find that M365 Copilot is emerging as an everyday assistant for knowledge work: writing dominates, but users also rely on it for information retrieval, analysis, decision making and strategizing, and evaluating and diagnosing programs and systems, among others. Information seeking tasks remain common, but time trends suggest a relative shift away from ``chat as search'' and toward content and communication-related work. Comparisons across occupational groupings and to work done in the labor market further show that usage is broad but uneven, where the relative share of work done with M365 Copilot Chat cuts across jobs in some cases and is occupation-specific in others. Areas of relative underrepresentation in the labor market suggest the next frontier for enterprise AI adoption.
Authors: Roman Klypa, Alberto Bietti, Sergei Grudinin
Abstract: The design of RNA molecules that interact with specific proteins is a critical challenge in experimental and computational biology. Despite recent progress in natural language modeling and deep learning-based protein design, there remains significant room to improve the frequency of successful interactions and the authenticity of generated sequences for functional applications. In this work, we frame conditional RNA sequence generation as a multi-stage alignment problem, introducing Moirain: a suite of models optimized via multimodal supervised fine-tuning (SFT) and Direct Preference Optimization (DPO). Our approach begins with large-scale pretraining on diverse RNA corpora to capture the fundamental grammars of sequence plausibility. To achieve target-specific generation, we employ a multimodal SFT architecture that conditions RNA synthesis on protein structural and sequential features. Finally, we leverage DPO to refine the model using synthetic interaction data: taking advantage of DPO's unique ability to navigate non-aligned preference spaces, we improve functional fitness without collapsing the learned natural distribution. Extensive evaluation of the Moirain series (Moirain-Base, -Multi, and -DPO) demonstrates that our framework consistently produces novel, diverse, and biologically plausible RNA sequences with superior binding affinities compared to existing baselines.
Authors: Celle Hendrickx, Fabio Pavirani, Chris Develder
Abstract: The growing share of Renewable Energy Sources (RES) in modern power systems increases both grid imbalances and frequency deviations, reinforcing the need for ancillary services such as Frequency Containment Reserve (FCR) and passive balancing. Battery Energy Storage Systems (BESS) are well-suited for these services, but prior research typically relies on uniform FCR bids that remain constant throughout the control period. Such static bids fail to fully exploit BESS flexibility, as they do not balance the trade-off between reserving energy for FCR delivery and using it for imbalance arbitrage, limiting the achievable value in value-stacking settings. To address this limitation, we propose a two-stage control framework for the European context that introduces non-uniform FCR bids. In the first stage, we derive a time-varying bid sequence using data-driven Monte Carlo (MC) optimization. In the second stage, a Deep Reinforcement Learning (DRL) agent leverages the residual flexibility for real-time imbalance trading while proactively managing the State of Energy (SoE) to ensure compliance with FCR requirements. The framework is presented as a proof of concept, highlighting the potential benefits of time-varying bidding strategies. By incorporating daily cycle budgets and time-varying reserve commitments, our approach achieves a 7.56% profit increase compared to uniform baselines. These results show that non-uniform bidding can unlock additional value by more effectively aligning reserve obligations with rapidly changing imbalance opportunities.
Authors: Ziyang Fang, JinXi Wang, Jinghui Zhong, Yew-Soon Ong
Abstract: Optimization modeling serves as the pivotal bridge between natural-language problem descriptions and optimization solvers, and remains a cornerstone for bringing operations research (OR) into real-world decision making. Recent advances in large language models (LLMs) have driven significant progress in automatic optimization modeling. However, existing methods still lack explicit validation during the modeling process, allowing errors introduced in earlier stages to carry through the pipeline and ultimately reduce final modeling accuracy. To address this challenge, we introduce TriVAL, a tri-validation framework that performs explicit validation at three stages of automatic optimization modeling: semantic specification, mathematical formulation, and code generation. At each stage, TriVAL follows a construct-validate-revise loop that assesses the current result against stage-specific criteria and revises it when needed. This design helps identify and correct errors before they accumulate across stages, helping preserve faithfulness throughout the modeling process. To evaluate automatic optimization modeling on more challenging combinatorial problems, we further introduce NL4COP, a benchmark of 150 instances across 50 diverse problem types with more complex decision logic, more tightly coupled constraints, and more demanding modeling requirements than existing benchmarks. Experiments on NL4COP and established benchmarks show that TriVAL consistently outperforms state-ofthe-art methods, with the largest gains on the most challenging problems.
Authors: Kyungmi Na, Yifei Li, Xinyi Yang, Bolei Deng
Abstract: In biological systems, sensing is not performed by the brain alone: the body deforms, vibrates, and filters external stimuli before they are transduced into neural signals. In engineered systems, this processing burden is placed largely on electronics and computation, while the mechanical body is usually designed only for strength and stability. Here, we present sensing intelligence as a trainable property of the body. We show that the geometry of a metamaterial can be optimized to reshape external stimuli into internal signals that are easier for a neural network to interpret. Rather than hand-designing this physical preprocessing, we let the neural network train its own body for sensing by backpropagating the sensing loss to the body's design parameters through differentiable simulation. Across numerical and experimental sensing scenarios, the optimized body improves sensing accuracy by up to fivefold or reduces the number of required electronic sensors by nearly an order of magnitude.
Authors: Eugene Yu Ji, Igor Grossmann, Amir-Hossein Karimi
Abstract: Generative AI research increasingly confronts a shared problem: systems must sustain yet govern their own generative activity when uncertainty is high, evidence is missing, or context is insufficient. This position paper argues that metacognition should become the scientific framework for bounded and effective self governance in generative AI, where output generation is properly evaluated together with the capacities through which generative systems navigate and regulate their own activity. We advance this position by showing that bounded and effective AI self-governance requires metacognitive alignment across computational, algorithmic, and ecological levels. At the computational level, metacognition specifies the meta-level functions a system is meant to serve, such as monitoring, evaluation, control, and adaptation. At the algorithmic level, these functions are realized through procedures such as elicitation, iteration, and modularization. At the ecological level, metacognitive signals become meaningful, actionable, and accountable within the interface, workflow, and accountability arrangements. Metacognition thus makes it possible to conceive generative AI as both capable and well-governed, rather than treating capability and governance as competing aims.
Authors: Heqiang Wang, Weihong Yang, Zheyuan Yang, Jia Zhou, Xiaoxiong Zhong, Fangming Liu, Weizhe Zhang
Abstract: Industrial anomaly detection has attracted significant attention as a fundamental challenge in industrial systems. The rapid advancement of heterogeneous industrial sensors has driven industrial anomaly detection from unimodal to multimodal paradigms. However, existing methods are primarily designed for centralized and offline settings, overlooking the distributed and continuously generated data characteristic of real-world industrial environments. With the advancement of edge intelligence, modern edge devices are increasingly capable of not only data acquisition but also distributed model training, enabling collaborative intelligence across the system. Industrial anomaly detection represents a critical application in this context. Motivated by these challenges, we propose a novel framework termed Multimodal Online Distributed Industrial Anomaly Detection (MODIAD). We first present a comprehensive workflow for MODIAD and then formulate a Multi-class Intelligent Scheduling (MIS) problem to coordinate cross class model updates by balancing data sufficiency and class update frequency. To efficiently solve this problem, we design a Sequential Marginal Gain Greedy (SMG) algorithm that enables effective multi-class training under resource constraints. Furthermore, to improve the computational and communication efficiency during training, we propose an Resource Efficient Class-Wise Low Rank Adaptation (REC-LoRA) strategy, which significantly reduces system overhead while preserving detection performance. Extensive experiments on two representative multimodal industrial anomaly detection datasets, MVTec 3D-AD and Eyecandies demonstrate that the proposed approach achieves superior performance and efficiency under the MODIAD scenario.
Authors: Han Chen, Zining Zhang, Wenqi Pei, Bingsheng He, Ming Wu, Jason Zeng, Michael Heinrich, Wei Wu, Hongbao Zhang
Abstract: Memory is a fundamental component for enabling long-context LLM agents, supporting persistent state across interactions through a continuous serve-and-update lifecycle. Despite substantial prior work, existing systems suffer from significant maintenance overhead due to two key limitations: coarse-grained state management and inherently sequential update pipelines. In particular, updates are often tightly coupled with LLM inference and require full-state rewrites, leading to poor scalability and growing latency as memory accumulates. To address these challenges, we present MemForest, a memory framework that reformulates agent memory as a write-efficient temporal data management problem. MemForest breaks the sequential bottleneck via parallel chunk extraction, decoupling memory construction into concurrent, independent operations. To further eliminate coarse-grained maintenance, we introduce MemTree, a hierarchical temporal index that organizes memory as time-ordered trees rather than flat global summaries. This design replaces full-state rewrites with localized per-node updates, reducing maintenance cost to the affected tree paths while naturally preserving temporally evolving states. We evaluate MemForest on two long-context memory benchmarks, LongMemEval-S and LoCoMo. On LongMemEval-S, MemForest achieves the best overall performance among stateful baselines, reaching 79.8% pass@1 accuracy while sustaining a memory construction throughput approximately 6x higher than state-of-the-art approaches including EverMemOS.
Authors: Yiwei Li, Zihao Wu, Huaqin Zhao, Yifan Zhou, Chao Cao, Dajiang Zhu, Tianming Liu, Lin Zhao
Abstract: Radiologist eye-tracking data provide a rich record of how experts search, compare, and accumulate evidence during image reading; yet, existing methods exploit this signal only partially, either as a static spatial prior or as an auxiliary prediction target decoupled from diagnosis. We propose GazeWorld, a medical imaging world model that treats the image as the world and the radiologist's fixation sequence as a trajectory through it. GazeWorld autoregressively predicts the latent representation of the next fixated patch from all previously visited ones, while a spatial-completion branch covers unvisited regions. At inference, GazeWorld generates a sequence of patch representations from the image alone without requiring real gaze data. Frozen GazeWorld features achieve state-of-the-art diagnostic accuracy across all nine supervised settings on CheXpert, RSNA Pneumonia, and SIIM-ACR Pneumothorax, as well as the highest zero-shot accuracy on all three benchmarks. On the GazeSearch benchmark, a generic decoder trained on the same frozen features outperforms the purpose-built LogitGaze-Med by over 16\% in ScanMatch and 22\% in SED, despite not being explicitly trained to predict gaze. GazeWorld demonstrates that modeling how experts read, not just what they conclude, offers a promising pretraining paradigm for medical imaging AI.
Authors: Siqiao Huang, Partha Kaushik, Michael Chen, Hengkai Pan, Omar Chehab, Fernando Moreno-Pino, Max Simchowitz
Abstract: World models have become a central paradigm for learning predictive simulators that support generation, planning, and decision-making. Yet, despite rapid progress in industry-scale interactive video generation, the broader research community still lacks compact, reproducible, and easily extensible implementations for studying the design choices underlying modern world models. We introduce Nano World Models, a minimalist codebase for future video prediction centered around diffusion forcing. Nano World Models provides a unified interface for generative objectives, model scales, action-conditioning mechanisms, latent observation spaces, datasets, evaluation protocols, and long-horizon rollout procedures. This design enables controlled studies of world-modeling components that are often entangled across separate implementations. Through experiments across simple control environments, game simulation, and real-robot data, we examine how prediction parameterization, architecture scale, action injection, sampling budget, and domain complexity affect video prediction quality and autoregressive rollout behavior. By releasing code, configurations, evaluation scripts, and pretrained checkpoints, Nano World Models aims to provide a compact yet extensible experimental substrate for open, reproducible, and scientific world-model research.
Authors: Jack Parry, Jack Saunders, Vinay Namboodiri
Abstract: Digital avatar watermarking presents unique challenges: avatars are routinely post-processed with background replacement, reframing, and format conversion before deployment. We introduce \textbf{RAW} (Robust Avatar Watermarking), a benchmark comprising 50 synthetic avatar videos from 5 commercial providers and 6 attacks simulating real-world avatar workflows. Evaluating 7 existing methods reveals that avatar-specific attacks such as background removal significantly degrade watermark recovery. We propose \textbf{WALT} (Watermarking Avatars with Learned Textures), which embeds watermarks in UV texture space via 3D face reconstruction. WALT achieves the highest robustness to zoom attacks (92.4\%) while maintaining strong performance on background removal (95.6\%). We release our benchmark to facilitate research into avatar-specific watermarking.
Authors: Chathura Wimalasiri
Abstract: Self-supervised learning (SSL) has emerged as a promising paradigm for addressing the annotation bottleneck in medical imaging by learning representations from unlabeled data. However, its effectiveness depends heavily on the design of the pretext task and its alignment with the downstream clinical objective. We present a systematic, task-oriented review of SSL in medical imaging, examining how different pretext-task formulations influence performance across classification, segmentation, detection, and other tasks. Following PRISMA guidelines, we analyze 75 studies published between 2017 and 2025 and organize them into four paradigms: contrastive, non-contrastive and predictive, generative and reconstruction-based, and hybrid learning. Rather than cataloguing methods by architecture, we map each paradigm to the downstream objectives it best supports. Our analysis shows there is no universally optimal SSL strategy; instead, performance is governed by the alignment between the pretext task, the imaging modality, and the target task. Contrastive methods learn global discriminative features and align well with classification, but may overlook subtle pathological patterns. Generative and spatial prediction-based approaches better preserve local anatomical structure, making them more suitable for segmentation and other dense prediction tasks, while hybrid methods offer the most balanced performance. We further show that modality-specific design is critical and that SSL provides its greatest benefit in low-label and few-shot regimes. Finally, we distill these findings into practical design guidelines and outline open challenges, including pathology-aware pretext task design, resource-efficient training for high-dimensional data, and standardized evaluation protocols. This work offers practical guidance for designing more effective and clinically relevant SSL frameworks in medical imaging.
Authors: Chenghao Li, Fusheng Hao, Xikai Zhang, Likang Xiao, Yanwei Ren, Fuxiang Wu, Quan Chen, Liu Liu
Abstract: Multimodal large language models via reinforcement learning (RL) have demonstrated remarkable capabilities in complex visual reasoning tasks, yet they remain limited in long-horizon multimodal scenarios, often suffering from visual hallucination and logical error. Current methods typically pre-encode high-dimensional visual scenes into discrete textual proxies to facilitate downstream reasoning. As the reasoning chain unfolds, however, the inherent information asymmetry between text and visual scenes tends to erode visual grounding, resulting in misguided reasoning and erroneous outputs. To address this issue, we introduce IVR-R1 (Iterative Visual-grounded Reasoning), a novel RL training framework that facilitates dynamic visual re-alignment that actively rectifies reasoning trajectories to guide policy optimization. Specifically, by leveraging a reward-driven screening mechanism to identify flawed rollouts, IVR-R1 executes a fine-grained, step-level error attribution within the multimodal context. By iteratively cross-referencing intermediate reasoning states against pristine visual priors, a Re-Reasoning Loop enables automated trajectory rectification, effectively synthesizing expert-level demonstrations that serve as high-fidelity reasoning templates for the policy model. Our experiments across diverse multimodal benchmarks demonstrate that IVR-R1 consistently outperforms existing reinforcement learning methods, establishing a superior paradigm for maintaining logical and visual consistency in complex multimodal reasoning.
Authors: Junyi Wu, Weijian Luo, Haoyang Zheng, Ruizhe Zhang, Guang Lin
Abstract: Recent advances in one-step text-to-image generation have enabled real-time synthesis with remarkable efficiency and quality. Previous reinforcement learning methods for one-step generators combine image-space reward optimization with diffusion noisy-space distribution matching. This paradigm brings challenges due to a mismatch between terminal reward optimization and the underlying generative dynamics. As a result, optimization tends to exploit stochastic degrees of freedom, often improving reward at the expense of image fidelity. To address this issue, we propose Diff-Instruct with Diffused Reward (DIDR), a data-free trajectory-level alignment framework derived from Integral KL minimization. DIDR propagates the RLHF-optimal reward-tilted clean-image distribution across all noise levels along the diffusion trajectory. We show that this objective admits the same minimizer as clean-image RLHF, while naturally inducing the Diffused Reward Score (DRS), which acts as a reward-driven correction to the reference score function. To make this practical, we further introduce the Diffused Reward Proxy (DRP), an efficient estimator of DRS based on differentiable short-step denoising. Extensive experiments demonstrate that DIDR consistently Pareto-dominates existing one-step SDXL baselines. Moreover, when transferred to a 6B DiT backbone (Z-Image), DIDR surpasses its 50-step teacher in preference alignment while requiring only a single generation step.
Authors: Bowen Deng, Bohan Li, Matthew Cox, Hoje Chun, Juno Nam, Artur Lyssenko, Sathya Edamadaka, Jurgis Ruza, Xiaochen Du, Nofit Segal, Jesus Diaz Sanchez, Mingrou Xie, Ty Perez, Yu Yao, Miguel Steiner, Sauradeep Majumdar, Charles B. Musgrave III, Anirban Chandra, Abhirup Patra, Detlef Hohl, Connor W. Coley, Ju Li, Rafael G\'omez-Bombarelli
Abstract: Computational materials science and chemistry span vast knowledge domains and fractured software ecosystems. Although large language models (LLMs) have demonstrated research capabilities, scaling monolithic agents to manage the rigor and complexity of atomistic research remains a challenge. Here, we introduce AtomisticSkills, an open-source harness framework that empowers general-purpose AI coding agents to conduct atomistic research across materials science, chemistry, and drug discovery. By hierarchically decomposing scientific workflows into agent skills and tools, AtomisticSkills provides agents with modular, extensible, and plug-and-play research capabilities. The framework integrates more than 100 human-curated multidisciplinary skills, including database access, thermodynamics and kinetics modeling, and diverse simulation engines employing machine learning interatomic potentials (MLIPs) and density functional theory (DFT). We validate its functional coverage against scientific literature and demonstrate robust orchestration capabilities across diverse scientific campaigns: generative design of Li-ion solid-state electrolytes, high-throughput screening of metal-organic frameworks for CO2 capture, autonomous MLIP benchmarking and fine-tuning, multi-stage structure-based virtual screening for drug design, multimodal X-ray diffraction pattern analysis, and screening of Fe-oxide catalysts for oxygen evolution reaction. AtomisticSkills provides a critical agent infrastructure towards building fully autonomous AI scientists.
Authors: Shuang Liua, Fiona Johnson, Rohitash Chandra
Abstract: Remote sensing techniques have been increasingly utilised in aquatic applications in recent years. A common challenge in using optical satellite data is the presence of missing observations due to cloud cover. These data gaps can lead to missed detection of critical events, such as algal blooms, in lakes of high interest to water authorities. As a result, enhancing the completeness of optical satellite datasets is crucial for improving the monitoring and prediction of algal blooms. In this study, we compared a traditional data imputation method (i.e., linear interpolation) with deep learning models for reconstructing missing spectral bands across four lakes with historical records of algal blooms. The deep learning models adopted include CNN-based architectures (i.e., CNN, Inception Resnet, and Autoencoder) and CNN-LSTM-based architectures (i.e., CNN-LSTM, Resnet-LSTM, and Autoencoder-LSTM). Our results demonstrated that deep learning models substantially outperformed the baseline linear interpolation method in imputing spectral band values within artificially masked regions. Among these models, CNN delivered the best performance across most lakes. Furthermore, we evaluated the performance of algal bloom indices (i.e., Green/Red and NDCI) derived from the imputed imagery by comparing them with the observed data. Our results demonstrate that deep learning models are effective for imputing missing data in PlanetScope SuperDove imagery, enabling more reliable applications in water monitoring.
Authors: Arash Akbari, Arman Akbari, Masih Eskandar, Qitao Tan, Yixiao Chen, Jingwu Luo, Bertha Pangaribuan, Liyun Zhang, Jennifer Dy, Geng Yuan, Xue Lin, Gaowen Liu, Stratis Ioannidis, Yanzhi Wang
Abstract: Vision-Language-Action (VLA) models exhibit remarkable action generation for embodied intelligence, but their heavy compute make deployment on edge platforms impractical. Aggressive, sub-4-bit weight quantization is the natural solution, yet existing post-training quantization (PTQ) methods suffer severe performance degradation in this regime. To address this, we introduce ActQuant, an action-guided mixed-precision PTQ framework that operates in two stages: (1) an inter-tensor bit allocator that assigns each weight matrix a single bit-width based on how much it contributes to predicting the agent's actions; (2) an intra-tensor scale optimizer tunes per-block quantization scales using action-aware curvature, so that dynamic range is concentrated on the weights most influential for control. To deliver the on-device benefits of our aggressive quantization, we further introduce OmniModel.cpp, an agentic conversion pipeline that ports architectures into a native C/C++ runtime with efficient low-bit kernels. We evaluate ActQuant both in simulation and on a real-world 6-DoF UR3 arm, with all models deployed through OmniModel.cpp. On the LIBERO benchmark, ActQuant is the only method that operates at or below 3 bits-per-weight, retaining 95.0% on OpenVLA-OFT and 94.8% on $\pi_{0.5}$. Pushed further, ActQuant reaches 2.5 bpw at 90.1% on OpenVLA-OFT, compressing the backbone from 14.3 GB to 2.7 GB (5.3$\times$). On the physical UR3 arm, $\pi_{0.5}$ quantized with ActQuant retains the baseline's success rate while reducing the memory footprint by 2.5$\times$.
Authors: Jeongmin Jin, Kyeongwon Lee, Mundo Jeong, Jongin Choi, Woojoo Lee
Abstract: Diffusion inference remains costly for edge deployment, yet existing accelerators focus almost exclusively on score networks because standard drift is merely a trivial linear scaling. Kuramoto orientation diffusion replaces this trivial drift with locally coupled phase interactions, improving sampling efficiency but introducing a new hardware bottleneck: a center-dependent nonlinear 5 x 5 stencil evaluated at every reverse step. This kernel maps poorly to conventional CNN accelerators and matrix-oriented engines. We present SA-Kura, to our knowledge the first digital systolic-array accelerator dedicated to locally coupled Kuramoto drift. By reformulating pair-wise sinusoidal coupling into neighbor accumulation independent of the center phase followed by a single center-dependent multiply-subtract combination, SA-Kura eliminates in-PE transcendental units and enables regular systolic execution with register-level reuse. SA-Kura was implemented in synthesizable RTL, integrated into a lightweight RISC-V-based SoC, prototyped on FPGA, and evaluated through 45 nm CMOS synthesis and power analysis. For the drift kernel only, compared with software execution of the same kernel on the processor core in the same SoC platform, SA-Kura reduces latency and energy by 193x and 69.4x, respectively. Compared with a standalone Jetson Orin Nano CUDA implementation of the same kernel, it is 6.57x faster and achieves approximately 46.0x lower energy per pixel.
Authors: Van Quang Nguyen
Abstract: Advancements at the intersection of computer vision and natural language processing are crucial for applications like assistive tech, multimedia querying, and robotics. This dissertation proposes novel architectures to improve intelligent agents across three key vision-language tasks: image captioning, visual dialog, and interactive instruction following. First, we address limitations in visual representation for image captioning. Traditional models rely on region-based features from CNN detectors, which lack global context and suffer from high computational overhead. We propose GRIT (Grid and Region-based Image captioning Transformer), a transformer-only architecture. By integrating grid and region features using a DETR-based detector, GRIT enables end-to-end training and out-performs prior methods in both inference accuracy and speed. Second, we tackle visual dialog, which requires multi-turn conversation about an image. The challenge lies in efficiently modeling interactions between multiple inputs (image, question, history). We introduce LTMI (Light-weight Transformer for Many Inputs). Utilizing a specialized attention block, an LTMI layer matches the representational power of a standard Transformer extension while utilizing less than one-tenth of its parameters, as validated on the VisDial dataset. Finally, we study interactive instruction-following for embodied AI using the ALFRED dataset. We propose a framework featuring a two-stage instruction interpretation: it first decodes language directives independently of visual context to predict a tentative action-object sequence, which is then fused with visual features for final execution. Using multiple egocentric views and hierarchical attention, our method accurately localizes objects and achieves a state-of-the-art unseen success rate of 8.37%.
Authors: Lopamudra Dey
Abstract: Chromatin regulators can alter transcriptional programs by modifying the accessibility of regulatory DNA elements. Understanding how regulatory sequences differ between wild-type (WT) and knockout (KO) conditions is crucial for deciphering transcriptional control. Here, we applied a convolutional neural network, \textbf{WTKO-CNN} with an attention mechanism to classify DNA sequences as WT or KO, achieving high predictive performance. To interpret the model, we generated saliency maps to identify nucleotide positions most influential for the classification decision. From these high-saliency regions, we extracted and clustered k-mers, enabling de novo motif discovery. Sequence logos and consensus motifs derived from the CNN filters revealed biologically meaningful patterns, which are further validated using MEME, TOMTOM, and HOMER against known transcription factor binding sites. Our analysis identified motifs associated with transcription factor families that discriminate WT from KO sequences, demonstrating that CNN-guided saliency mapping is a powerful approach for uncovering functional sequence features.
Authors: Zikang Zhou, Haibo Hu, Xinhong Chen, Yifan Zhang, Nan Guan, Yung-Hui Li, Chun Jason Xue, Jianping Wang
Abstract: Multimodal motion forecasting is inherently under-supervised: each training scene provides only one realized future, yet multiple plausible futures exist. This sparse supervision often leads to mode collapse (redundant hypotheses and insufficient mode coverage) and unreliable confidence ranking when predicting a small set of trajectories. We propose Mode-as-Sequence, a unified decoding framework that translates an unordered mode set into an ordered mode sequence and explicitly models mode-to-mode dependency. Under this framework, we develop two complementary instantiations. ModeSeq performs recurrent mode decoding, where each mode is generated conditioned on the previously generated modes, encouraging diverse, non-redundant hypotheses with calibrated confidence ordering. To remove the mode-by-mode autoregressive bottleneck, we further propose Parallel ModeSeq, which preserves the same causal dependency using masked mode-to-mode self-attention while decoding all modes in a single forward pass, enabling efficient large-$K$ inference and scalable joint-scene prediction. To learn representative modes and calibrated confidence under sparse labels, we introduce Early-Match-Take-All (EMTA) and its joint-scene extension MA-EMTA, together with a lightweight ranking regularizer that reduces confidence inversions. Extensive experiments on large-scale benchmarks demonstrate consistent improvements in both ranking-oriented metrics and best-of-K accuracy across datasets, horizons, and object types. In the Waymo Open Dataset challenges, ModeSeq achieves 1st place in the 2024 LiDAR-free motion prediction track, and Parallel ModeSeq achieves 1st place in the 2025 Interaction Prediction Challenge, validating the effectiveness of Mode-as-Sequence for both accuracy and efficiency.
Authors: Xiaotian Liu, Shuyuan Shang, Xiaopeng Wang, Pu Ren, Yaoqing Yang
Abstract: Neural operators serve as fast, data-driven surrogates for scientific modeling but typically rely on a monolithic, single-pass inference procedure that struggles to resolve high-frequency details, a limitation known as spectral bias. We introduce the Iterative Refinement Neural Operator (IRNO), which augments pre-trained operators with a learned refinement module iteratively applied via fixed-point iteration. IRNO decomposes the prediction into a coarse initialization followed by successive residual corrections, paralleling classical numerical solvers. Under local assumptions, we establish contraction of the induced operator, ensuring convergence to a unique fixed point. To explicitly target high-frequency errors, we propose a progressive spectral loss that adaptively increases penalty on high-frequency components over refinement steps during training. Across physical systems, IRNO consistently lowers error, with up to 56.05% improvement on turbulent flow. On Active Matter, spectral analysis reveals that, relative to base operator, the normalized error ratios decrease to 27.72-36.10% in low-, 5.07-6.68% in mid-, and 1.48-2.04% in high-frequencies, remaining stable beyond the trained iteration count. Code is available at https://github.com/xiaotianliu-dartmouth/Iterative_Refinement_Neural_Operator
URLs: https://github.com/xiaotianliu-dartmouth/Iterative_Refinement_Neural_Operator
Authors: Alexander Okezue Bell
Abstract: Of $1{,}536$ Gaussian release covariances we tested for single-layer hidden-state privacy, zero achieve both moderate utility and moderate privacy against an adaptive retrieval attacker. We prove a complementary Fisher-ball lower bound: every full-rank Gaussian release at $O(1)$ Fisher utility admits a direction whose Mahalanobis signal grows linearly in hidden width, ruling out uniform Gaussian safety in the class and matching the empirical empty middle. The diagonal inverse-Fisher release $\Sigma^\star_{\mathrm{diag}}(\mathcal{K}) = (2\mathcal{K}/d)\,\mathrm{diag}(1/F_{ii})$ is the unique minimax-optimal diagonal mechanism at first-order KL budget $\mathcal{K}$ and the only release with worst-attacker top-1 $\le 0.001$ at every point of a 32 model-layer grid, but it sits on a privacy/utility edge rather than filling the middle. A generalized-eigen mechanism reaching $13\times$ Pareto reduction under Euclidean retrieval collapses to $100\%$ top-1 under the adaptive Mahalanobis attacker, and a full-trajectory sequence inverter recovers $94\%$ of clean GPT-2 prefixes but $0\%$ under $\Sigma_{\mathrm{diag}}$. A split-memory transformer trained from scratch reaches $G_{\mathrm{Mah}} \in [20, 33]$ at 90M and maintains a $6$--$24\times$ advantage over same-budget GPT baselines from 30M to 1B at a fixed-token language-modeling loss penalty; pretrained models top out at 9.3. These results reframe hidden-state release from mechanism-design within the Gaussian class to architecture or release co-design.
Authors: Sanchit Kabra, Nikhil Abhyankar, Saaketh Desai, Prasad Iyer, Chandan K Reddy
Abstract: Scientific discovery is a closed-loop process in which hypotheses guide data acquisition and observations refine the hypothesis space. Yet most approaches reduce discovery to supervised learning over fixed datasets, where limited observations can support multiple plausible mechanisms that fit locally but fail to generalize. Thus, the key challenge is selecting informative observations to resolve uncertainty, shifting the focus from static inference to adaptive data acquisition. To address this, we propose LLM-AutoSciLab, a closed-loop framework that couples hypothesis generation with hypothesis-conditioned experiment selection and mechanism refinement. Rather than fitting models to passively collected data, LLM-AutoSciLab iteratively proposes plausible hypotheses, selects informative experiments to distinguish or refine them, and updates its state using the resulting evidence. To evaluate dynamic, closed-loop scientific discovery with active data acquisition, we introduce ActiveSciBench, comprising two datasets: ActiveSciBench-Chem with 57 enzyme-kinetics tasks and ActiveSciBench-GRN with 45 gene-regulatory-network tasks. These datasets model discovery as a budget-constrained process requiring adaptive experiment design, variable selection, and recovery of true mechanisms. Across NewtonBench, ActiveSciBench-Chem, and ActiveSciBench-GRN, LLM-AutoSciLab outperforms prior methods, achieving 67.6% and 35.1% symbolic accuracy on NewtonBench and ActiveSciBench-Chem, respectively, and 31.1% exact graph recovery on ActiveSciBench-GRN. Moreover, hypothesis-guided experimentation is 2-5x more sample-efficient than the strongest competing baselines. Code and data are available at: https://github.com/scientific-discovery/LLM-AutoSciLab
URLs: https://github.com/scientific-discovery/LLM-AutoSciLab
Authors: Zhaohan Meng, Zhen Bai, Ke Yuan, Iadh Ounis, Zaiqiao Meng, Hao Xu, Joseph Loscalzo
Abstract: Protein-ligand modeling underpins computational drug discovery and molecular design. Existing protein-ligand benchmarks typically evaluate whether a protein and ligand interact and how strongly they bind, through tasks such as binary binding prediction and affinity regression. However, these evaluations provide limited evidence of whether models can localize binding sites or identify the non-covalent interactions underlying molecular recognition. To address this gap, we introduce InteractBind, a large-scale protein-ligand dataset comprising approximately 100k protein-ligand pairs, together with a benchmark for fine-grained evaluation. The core fine-grained task is that of binding-site localization, which uses protein-residue and ligand-atom interaction maps spanning six major types of non-covalent interactions to assess whether model-derived interaction maps localize binding sites. InteractBind further includes binding affinity and protein similarity-controlled splits to support realistic generalization assessment. Using InteractBind, we evaluate eight existing sequence-based and interaction-aware models, assessing binary binding prediction and binding-site localization. Results reveal limited binding-site localization despite strong binary binding prediction, with marked variation across non-covalent interaction types. Overall, InteractBind establishes a benchmark paradigm that encourages the development of more interpretable and physically grounded protein-ligand models.
Authors: Yichi Zhang, Kevin Lu, Yuang Zhang, Jie Gao, Lirong Xia, Fang-Yi Yu
Abstract: Multi-AI collaboration, such as ensembling or debating large language models (LLMs), is a promising paradigm for aggregating information and boosting performance. A foundational step in these pipelines is to feed the responses of several proposer LLMs into a summarizer LLM, which synthesizes a better answer. However, choosing which proposers to include is non-trivial. Existing approaches primarily focus either on accuracy (picking the strongest models) or diversity (ensuring variety), and often overlook the interactions among proposers and with the summarizer. We reframe proposer selection as a combinatorial selection problem akin to feature selection, where the value of an LLM lies in its complementarity with others. However, directly applying standard feature-selection algorithms is impractical in the LLM setting due to prohibitive time complexity. Motivated by this limitation, we explore an extensive range of computationally feasible, greedy-style selection algorithms that assess complementarity using a small labeled set. Our experiments validate complementarity as a guiding principle for proposer selection and identify methods that achieve the best performance-cost trade-offs in practice.
Authors: Hongwen Song (Vinson), Song (Vinson), Wei
Abstract: Skill libraries allow LLM agents to load task-specific instructions on demand, letting non-expert users solve domain-specific tasks through natural language without knowing which skills exist or how they work. However, performance degrades as libraries grow -- by up to 21\% when scaling from a small set of helpful skills to a 202-skill library. In this work, we formulate this performance degradation as the pass rate drop between loading a library of known-helpful skills and the full library. Moreover, we propose to decompose the pass rate drop by conditioning on the skill(s) invocation -- which skills the agent selects during a trajectory -- into two effects: \emph{skill shadowing}, where the agent selects wrong skills more often as the library expands, and \emph{context overhead}, where the enlarged context degrades execution even when selection is correct. We derive upper bounds on both effects to characterize their magnitudes of impacts to the pass rate drop. Our empirical estimates of the effects and their upper bounds both show that the \emph{skill shadowing} effect grows with library size and significantly contributes to the performance degradation, whereas the \emph{context overhead} effect remains small and indistinguishable from zero. This observed asymmetry establishes that the skill selection failure, not the enlarged context, is the primary bottleneck when expanding the skill libraries.
Authors: Shugang Hao, Lingjie Duan
Abstract: To better serve users' demands in mobile applications (e.g., navigation), mobile crowdsourcing platforms can iteratively align large language model (LLM)-generated content (e.g., AI-generated traffic condition predictions) with human feedback collected from crowdsourcing workers (e.g., mobile users). However, workers may strategically misreport their online preference feedback to maximize their influence or payment. Existing pipelines in mobile crowdsourcing (e.g., EM-based weight estimation) fail to identify the most accurate worker in this online setting, resulting in a linear regret $\mathcal{O}(T)$ over $T$ time slots. In this paper, we study truthful online preference aggregation for LLM fine-tuning in mobile crowdsourcing. We formulate a new dynamic Bayesian game to model the multi-agent online learning process between the platform and strategic mobile workers. We propose a novel online weighted aggregation mechanism that dynamically adjusts each worker's weight in the preference aggregation according to their feedback accuracy. We prove that our mechanism ensures truthful feedback from strategic workers and achieves a sublinear regret $\mathcal{O}(\sqrt{T})$ over $T$ time slots. We further extend our mechanism to a challenging scenario with limited worker feedback per time slot, still guaranteeing a sublinear regret $\mathcal{O}(\sqrt{T})$. Experiments on LLM fine-tuning with real-world datasets further demonstrate significant performance gains of our mechanisms over benchmark schemes.
Authors: Yuefeng Liu, Ning Yang, Ziyu Yang
Abstract: Real-world time-series data in industrial sensing, healthcare, and energy systems is often corrupted by a mixture of Gaussian noise and occasional large-magnitude impulse outliers. For tasks that depend on local shape, such as ECG morphology analysis and battery degradation monitoring, the main requirement is not only low reconstruction error but also preservation of derivative peaks and task-critical features. We propose Cascade-KDE, a training-free restoration framework for corrupted time series. The method first estimates a two-dimensional temporal-amplitude density, then applies a Density-Truncated Robust Expectation to limit the influence of distant abnormal points, and finally refines the sequence through an exponential cascade with adaptive stopping. This design aims to improve robustness under out-of-distribution impulse corruptions while keeping the restored trajectory close to the original local structure. Across several benchmark datasets, the proposed method shows consistent gains over classical filters and representative learning-based baselines on curve fidelity, derivative preservation, downstream classification, and runtime efficiency. These results suggest that bounded density-based restoration is a practical option for feature-preserving preprocessing in noisy time-series pipelines.
Authors: Fuming Yang
Abstract: Neural networks acquire structured representations at specific moments during training, yet identifying these transitions typically relies on retrospective, label-dependent metrics. We introduce a bifurcation theory of representation dynamics to detect these moments in real time. Analyzing a passive GMM probe attached to the evolving encoder, we show the onset of structure corresponds to a supercritical pitchfork bifurcation driven by the loss Hessian. The system exhibits a theoretically predictable zero-crossing ($\beta_c$) that, compared to the network's current state ($\beta$), yields a dynamic ratio $\beta(t)/\beta_c(t)$: a universal, label-free phase coordinate for representation dynamics, computable entirely from hidden states. We empirically validate four distinct transition regimes predicted by this coordinate across diverse settings: SAEs on language models (Pythia), SSL (CIFAR), and grokking (modular arithmetic). Crucially, under finite dissipation, macroscopic symmetry-breaking can lag the initial zero-crossing by orders of magnitude, which providing a rigorous dynamical account of the delayed escape observed in grokking. Microscopically, the bifurcation creates a shared unstable subspace, forcing collective symmetry breaking. We term this the "feature lottery" in SAE training: a feature's terminal interpretability becomes predictable remarkably early. By only 5% of training, early atom purity robustly predicts final convergence purity, with top-decile early atoms achieving over 12x the baseline purity at convergence. Beyond explaining concept emergence, $\beta/\beta_c$ provides a practical early-warning indicator for training health, detecting the onset of usable structure, the crystallization of feature identity, and representational collapse epochs before downstream metrics react.
Authors: Yoshihiko Fujisawa, Yuma Ichikawa, Yudai Fujimoto, Akira Sakai, Katsuki Fujisawa
Abstract: On-device adaptation of large language models commonly keeps a quantized base model frozen while training and deploying a small, task-specific LoRA adapter. In the unmerged adapter-mode setting, however, the adapter is more than a compact storage module; it introduces an additional dense floating-point branch, maintains a trainable state for local updates, and acts as a unit of communication and hot-swapping.We introduce LoRDBA, a LoRA-compatible adapter that replaces both low-rank factors with binary sign carriers while representing magnitudes through lightweight, channel-wise scales, converting the dense adapter branch into two sign-accumulation matrix multiplications interleaved with channel-wise scaling. A finite-sample analysis shows that reconstruction quality is governed by the residual-to-magnitude ratio of the original LoRA factors. In adapter-mode experiments, LoRDBA outperforms low-bit baselines at matched model sizes while matching fp16 LoRA quality in selected regimes. The unmerged adapter incurs at most 8% prefill latency overhead at matched rank r=16 despite an over 10x reduction in adapter footprint, with moderate training memory overhead of approximately 1.6x that of fp16 LoRA.
Authors: Yongzhong Xu
Abstract: We present a three-step recipe for identifying attention-head circuits in pretrained transformers. A per-head spectral signal -- the time-integrated participation ratio of each head's attention output -- ranks heads doing sustained content-dependent computation without labels or attribution gradients. A task-pattern screen filters this general indicator into a task-specific candidate circuit, and group ablation against a matched-random control completes the causal claim. We validate across an 8x parameter range (51M to 1B-active / 7B-total), two architecture families (dense, mixture-of-experts), and four pretraining pipelines. The recipe ports: a 2-6 head induction circuit is causally necessary in every model tested, with a 94-100% drop in synthetic-induction top-1 after ablation. The spectral signal is predictive without supervision: on six independent seeds of a 51M-parameter probe model, the same computation identifies the seed-specific circuit on each seed. The fraction of heads doing identifiable specialized computation is conserved at 17-19% across the Pythia family (124M to 410M), while specific induction circuits stay 3-11 heads -- sublinear in total head count. This paper is the methodology anchor of a three-paper program; companion papers extend the recipe to developmental trajectories during pretraining and to composed-task circuits where pattern selectivity decouples from task-causal structure.
Authors: Koffka Khan
Abstract: Human-body communication (HBC) is a promising physical substrate for wearable body-area networks because it can localize communication around the body and reduce the burden of conventional radio links. Federated learning (FL) is a promising learning substrate because it can reduce raw-data centralization for physiological and behavioral sensing. Yet these two literatures remain weakly connected: FL for wearables usually abstracts the communication layer, whereas HBC research usually abstracts learning and model-update traffic. This article surveys the intersection of HBC, wireless body-area networks, wearable FL, Internet-of-Bodies privacy, and edge-intelligence optimization. We propose a taxonomy that distinguishes intra-body, body-hub, cross-user, and clinical-cloud FL deployments, and we identify the open problem of body-channel-aware FL: learning protocols whose client selection, update compression, and aggregation are controlled by posture-dependent HBC links, residual energy, sensor memory, and privacy risk. To make the research agenda concrete, we introduce BODYFED-HBC as a reference architecture and provide an optimization formulation and scheduling algorithm. We further specify a reproducible simulation vignette that combines public wearable datasets with empirical body-coupled-communication signal-loss models. The article concludes with open datasets, evaluation metrics, limitations, and research directions for computer scientists working above the hardware layer.
Authors: Jaejun Lee, Seheon Kim, Joyce Jiyoung Whang
Abstract: Hyper-relational knowledge graphs (HKGs) effectively represent complex facts. While inferring new knowledge in HKGs is a critical problem, current methods cast it as a simple link prediction, assuming that nearly all entities and relations within a fact are known, leaving only a single blank to be filled. However, this restricted assumption may not hold in real-world scenarios in which multiple, or even all, constituent components of a fact may be missing simultaneously. To bridge this gap, we introduce a task called fact generation: generating a valid hyper-relational fact from an arbitrarily masked query, i.e., completing a partially observed fact or generating a fact from scratch. We propose KREPE, the first generative representation learning method for HKGs that learns to model the probability distributions of missing components conditioned on the local fact components and global structure of HKGs via a masked discrete diffusion. KREPE models both the intra-fact dependencies by contextual message passing and inter-fact correlations by aggregating stochastically sampled contexts. KREPE seamlessly unifies link prediction and fact generation within a single training framework, achieving state-of-the-art performance on standard HKG link prediction benchmarks and outperforming LLM-based baselines in generating novel and correct facts.
Authors: Shi Liu, Xuehai Tang, Xikang Yang, Liang Lin, Biyu Zhou, Wenjie Xiao, Wantao Liu
Abstract: The rise of tool-using Large Language Model (LLM) agents, standardized by protocols like the Model Context Protocol (MCP), has unlocked unprecedented autonomous execution capabilities for LLM Agents by integrating external open-domain knowledge and tools. However, this interoperability introduces a covert attack surface targeting the agent's cognitive planning layer. This paper systematically investigates Tool Description Poisoning (TDP), a novel semantic attack. In TDP, malicious instructions are not embedded in a tool's executable code, but rather covertly injected into its descriptive metadata, the very "manual" an agent relies on for secure planning and decision-making. To rigorously and systematically evaluate this emerging threat, we introduce the MCP-TDP Security Benchmark. This high-fidelity sandbox environment comprises 32 realistic, real-world test cases spanning 6 distinct risk categories. Our evaluation of 8 mainstream LLMs reveals severe vulnerabilities, with leading models like GPT-4o exhibiting a nearly 100% Attack Success Rate (ASR) in six high-risk scenarios. Furthermore, our findings demonstrate that common prompt-guardrail defenses are largely ineffective and can, counterintuitively, even be counterproductive (a phenomenon which we term the "Firewall Fallacy"). Crucially, we also propose a defense mechanism: "Reactive Self-Correction," where an agent autonomously detects and reverts its own malicious actions post-execution. This work provides the first specialized security benchmark tailored for TDP, offering essential insights for securing the cognitive and planning layers of advanced agentic systems.
Authors: Ajhesh Basnet
Abstract: Training a reinforcement learning agent on-policy means collecting fresh experience at every update, and that experience comes with a hidden problem. Each state in a rollout is the direct output of the previous one, causally chained together by the agent's own actions. Because of this, consecutive transitions are never truly independent. They carry overlapping information, and the gradient signal the network receives ends up far more repetitive than the batch size suggests. The same directions get reinforced over and over, the value network struggles to keep up as the policy shifts, and training becomes quietly unstable in ways that reward curves alone rarely reveal. This paper asks whether that redundancy can simply be removed. We show that randomly dropping a fixed fraction of transitions from the rollout, at the right stage so the reward signal stays intact, is enough to break the repetitive gradient structure and stabilize training. The change is minimal: one sampling step, no new components, no modification to the core algorithm, and it works with any PPO implementation. Across five environments of increasing difficulty, CartPole-v1, Acrobot-v1, LunarLander-v2, HalfCheetah-v5, and Hopper-v5, the method matches vanilla PPO on reward while producing more consistent training dynamics across KL divergence, policy entropy, and value estimates. Dropping 25% of transitions turns out to be the sweet spot: enough to disrupt the redundancy, not enough to thin the batch.
Authors: Yifeng Di, Xuliang Huang, Tianyi Zhang
Abstract: Data contamination is a known threat to the reliability of model evaluation. However, it remains underexplored in code large language models (LLMs), where contamination often goes beyond exact duplication. We present TRACER, a semantic-aware framework for fine-grained code contamination detection. TRACER models contamination using three levels of semantic overlap - Functionally Identical, Nearly Identical, and Shared Logic - and detects them through a coarse-to-fine pipeline. We also introduce the first benchmark for fine-grained code contamination detection, spanning three widely used benchmarks and three representative post-training datasets. TRACER achieves strong and consistent performance across multiple LLM backbones, with GPT-5 reaching an F1 score of 0.91 in fine-grained detection. In the binary setting, TRACER attains an F1 of 0.92, outperforming existing methods by 42%-217%. We further conduct ablation studies and error analysis to assess the contributions of individual components in TRACER.
Authors: David Boetius, Shahaf Bassan, Guy Katz, Stefan Leue, Tobias Sutter
Abstract: Shapley additive explanations (SHAP) are widely recognised as computationally intractable for neural networks, since they induce an exponential search space over the input features. In this work, we take a first step towards scaling exact SHAP computation to larger search spaces by introducing an algorithm that leverages recent advances in neural network verification to compute arbitrarily tight exact lower and upper bounds on SHAP values for neural networks, ultimately recovering the exact SHAP values. We demonstrate that our approach scales to orders of magnitude larger search spaces than state-of-the-art exact methods. This provides an important first step towards exact SHAP computation and establishes a principled cornerstone for evaluating statistical approximation methods on larger search spaces.
Authors: Shu Liu, Alexander Krentsel, Shubham Agarwal, Mert Cemri, Ziming Mao, Soujanya Ponnapalli, Alexandros G. Dimakis, Sylvia Ratnasamy, Matei Zaharia, Aditya Parameswaran, Ion Stoica
Abstract: Core systems like key-value stores have historically taken years to build, and are designed to be general so as to amortize cost across deployments, paying a significant performance cost. We argue that LLM-based coding agents now make a different approach tractable: Just-in-Time Systems, in which the entire system is synthesized from scratch, specialized to the environment, workload, and required system properties. We present a JIT system synthesis pipeline, Jitskit, and explore its effectiveness in synthesizing key-value stores from spec cards that span different YCSB workloads, deployment constraints (e.g., compute resources), and system properties (e.g., consistency and durability). Jitskit iteratively refines a system implementation to match the specification against an evolving evaluation test suite. The resulting synthesized systems are performant, beating comparable state-of-the-art systems on 18 of 18 specs tried, by up to 4.6x over the best off-the-shelf baseline on the most favorable spec. Naively running Claude Code either reward-hacks or underperforms Jitskit by up to 5.4x. We discuss the challenges we overcame in building Jitskit and our key takeaways.
Authors: Tewodros Syum Gebre, Jagrati Talreja, Matilda Anokye, Leila Hashemi-Beni
Abstract: Rapid and accurate flood extent mapping from Remote Sensing data, such as Synthetic Aperture Radar (SAR), is critical for operational disaster response, but standard Deep Learning models often produce physically impossible predictions due to a lack of hydrological constraints. While PhysicsInformed Neural Networks (PINNs) attempt to address this by embedding governing laws directly into the loss function, their application to real-world remote sensing data frequently fails. Enforcing rigid spatial derivatives (e.g., the 2D Shallow Water Equations) onto unconditioned latent spaces attempting to fit noisy SAR speckle causes catastrophic gradient divergence, a phenomenon we term Physics Shock. In this paper, we propose a novel Uncertainty-Aware PINN framework tailored specifically for applied Earth Observation that addresses this instability. By integrating a dynamic Warm-Start protocol and modeling heteroscedastic aleatoric uncertainty via a negative log-likelihood objective, the network learns to dynamically relax physical constraints in regions of high sensor noise while strictly enforcing them in high-confidence areas. Evaluated on the Sen1Floods11 dataset, our probabilistic Attention-Gated FNO-UNet successfully stabilizes multi-objective optimization, achieving a +25% relative improvement in Intersection over Union (IoU) compared to deterministic baselines. Furthermore, through Deep Ensembles, we successfully disentangle intrinsic sensor noise from out-of-distribution terrain ignorance, providing operational agencies with highly calibrated, physically consistent confidence bounds for robust disaster mitigation and real-time decision-making.
Authors: Vansh Garg, Rohit Jayanti, Krish Pandya, Sarthak Chittawar, Siddharth Tourani, Muhammad Haris Khan, Sourav Garg, Madhava Krishna
Abstract: Visual navigation ability is strongly tied to its underlying representation of the world. Unlike classical 3D maps that require globally-consistent geometry, image- or object-relative topological graphs almost entirely do away with geometric understanding. But, this comes at the cost of navigation capability, often limiting it to merely teach-and-repeat. In this work, we propose a novel map representation in the form of pixel-relative connectivity, which is geometrically accurate but does not require global geometric consistency. Inspired by recent progress in 3D grounded image matching, we construct a map from an image sequence through inter-image connectivity based on pixel correspondences in the relative 3D coordinate systems of individual image pairs. We then use this pixel-level graph to perform global path planning by approximating and sparsifying intra-image pixel connectivity. Through this, we derive a ''WayPixel Costmap'' representation and train a controller conditioned on it to predict a trajectory rollout. We show that this dense pixel-level costmap based on relative geometry is a more accurate conditioning variable for control prediction than its image- and object-level counterparts. This enables a highly capable navigation system, as validated on four types of navigation tasks in the simulator and through real world demonstrations.
Authors: Hinduja Nirujan, Shreyas Patil, Abdallah Ayoub, Ahmad Abdel Latif, Gouri Ginde
Abstract: Large Language Models (LLMs) are increasingly used to generate summaries of software bug reports, including sections such as Steps-to-Reproduce (S2R), Actual Behavior (AB), and Expected Behavior (EB). However, these models frequently produce hallucinations that can be convincing but unsupported by the source report. This can mislead developers and reduce trust in automated maintenance tools. Existing hallucination detection approaches typically evaluate outputs at the full-response level and do not consider the structure of technical documents. An initial exploratory study on 80 structured bug report summaries found that approximately 47.9% contained missing information, while 12.3% included fabricated content, highlighting the need for systematic hallucination analysis in bug report summarization. In this work, we empirically investigate hallucinations in LLM-generated bug report summaries from a section-aware perspective. Using the BugsRepo dataset, derived from Mozilla OSS projects, we introduce controlled synthetic hallucination injection to construct a benchmark for training and evaluation. We propose a section-aware hallucination detection approach that jointly predicts whether a summary contains hallucinated content, identifies affected sections, and classifies hallucination types. Experimental results across multiple pretrained language models show that the proposed approach achieves strong performance across all tasks, with the best model obtaining 0.89 report-level Macro-F1, 0.83 section-level Macro-F1, and 0.84 hallucination-type Macro-F1. We further analyze common hallucination patterns and model failure modes to better understand limitations of current LLM-generated bug report summaries. The findings highlight the importance of section-aware hallucination analysis for improving the reliability of LLM-assisted bug report summarization in software maintenance workflows.
Authors: Srijita Basu, Viktor Kjellberg, Simin Sun, Bengt Haraldsson, Md. Abu Ahammed Babu, Wilhelm Meding, Farnaz Fotrousi, Miroslaw Staron
Abstract: Large Language Models (LLMs) are increasingly applied to software engineering (SE), yet their potential for autonomous, role-oriented collaboration remains largely underexplored. Understanding how multiple LLM-based agents coordinate, maintain role alignment, and converge on solutions is critical for SE, as naively allowing agents to interact does not reliably lead to correct or stable outcomes. Recent empirical studies show that unstructured or poorly understood interaction dynamics can result in error propagation, premature consensus on incorrect solutions, or prolonged disagreement that prevents convergence, even when correct partial solutions are present early in the interaction. As an initial step towards addressing this underexplored area, we undertake a systematic analysis of conversations between two agents, a Designer and a Programmer across 12 model combinations from 7 open-source LLMs (Gemma 2, Gemma 3, LLaMA 3.2, LLaMA 3.3, DeepSeek-R1, MiniCPM, and Qwen3). Our systematic approach reveals three key dimensions of multi-agent interaction: efficiency (the speed and stability of convergence), consistency (the degree of role alignment visualized by BLEU and ROUGE), and effectiveness (the extent of compilation success and error resolution). Results show that the DeepSeek-R1:DeepSeek-R1 pair was unique in converging to the correct solution from the very first iteration and sustaining it consistently to the final iteration, while LLaMA 3.2:LLaMA 3.2 and Qwen3:Qwen3 demonstrated strong Designer:Programmer role alignment despite of diverging from the correct solution. The other pairs deviated from the task, never to converge to a result. These findings advance understanding of agentic programming and highlight the need for further research on understanding and calibrating convergence and stop conditions essential for future autonomous SE.
Authors: \"Ozkan Canay
Abstract: Effective skills-aware talent recommendation must balance behavioral transition patterns, trajectory-sensitive adaptation, and inspectable occupation-level criteria. Evidence from public benchmarks on how these signals interact, however, remains limited. This study proposes CF-RL-TOPSIS, an interpretable late-fusion model that integrates a transition-aware collaborative branch, a compact reinforcement-style occupation-family bandit, and an entropy-weighted TOPSIS branch constructed from six semantic proxies; the validation-selected fusion coefficients remain auditable. The model is evaluated on two frozen public ICT talent-history benchmarks, JobHop and Karrierewege, using repeated chronological top-5 ranking and paired Wilcoxon tests. On JobHop the full hybrid attains NDCG@5 = 0.3040 +/- 0.0073 and significantly surpasses repeat-last, item Markov, transition-aware collaborative filtering, the CF+TOPSIS hybrid, GRU4Rec, and SASRec (p <= 0.0039 across planned comparisons). On Karrierewege the hybrid remains competitive but does not significantly exceed the strongest Markov baseline, revealing a persistence-dominated setting in which the bandit branch appropriately shrinks to near-zero weight. Proxy-sensitivity, family-level deep Q-network, and runtime checks support this interpretation, and a worked user-level case shows how branch scores, criterion weights, and rank shifts can be inspected for an individual recommendation. The contribution is not a benchmark-agnostic superiority claim, but a reproducible account of the conditions under which transparent late fusion adds value beyond simple continuation heuristics. In semantically rich, non-saturating talent-history regimes the three branches reinforce one another; in persistence-dominated regimes the same architecture remains competitive through its collaborative backbone, with the adaptive branch correctly inactive.
Authors: Yuwei Xue, Sakib Mostafa, James Zou, Joseph Liao, Maximilian Diehn, Ash A. Alizadeh, Lei Xing, Md. Tauhidul Islam
Abstract: Biological systems are governed by structured molecular interactions, where pathways, regulatory circuits, and functional gene relationships shape cellular behavior and disease progression. Much of this knowledge is naturally represented as graphs. However, most biomedical AI models cannot directly use graph-encoded biological knowledge and instead require compressed low-dimensional representations, which can lose important structure and reduce performance, especially in limited-sample clinical studies. Here, we introduce Graph-in-Graph (GiG), a knowledge graph-modulated deep learning framework for data-efficient clinical prediction. GiG represents each patient as a standalone modular graph, in which curated biological knowledge graphs define edges and patient-specific measurements, such as gene expression, define node features. This design allows multiple biological knowledge graphs to be integrated while preserving gene-gene interactions and pathway topology during patient-level representation learning. Across cohorts comprising nearly 9,700 patients and five clinical tasks, including liquid biopsy cancer detection, prostate cancer diagnosis, and 32-class pan-cancer classification, GiG consistently outperforms traditional and state-of-the-art methods, with the largest gains in limited-sample settings. On the challenging prostate cancer diagnosis task, GiG improves macro-F1 by up to 49 percentage points relative to competing methods. Control experiments replacing real pathway graphs with random topologies confirm that these gains arise from biologically grounded knowledge graph structure rather than graph modeling alone. These findings show that knowledge graph-modulated deep learning can improve robustness, interpretability, and sample efficiency in clinical data analysis, and provide a principled framework for integrating biological knowledge graphs into predictive modeling.
Authors: Steffen J. Camarato, Yahya Hmaiti, Mandana Ghadamian, David Mohaisen
Abstract: Large language models are increasingly used for vulnerability detection, yet their reliability under different prompt formulations remains uncharacterized. We present PromptAudit, a controlled evaluation framework that isolates prompt effects by fixing the dataset, decoding, and parsing while varying only the prompting strategy. Using five prompting strategies across five open-weight models on 1,000 CVEs (6,074 code samples spanning 16 programming languages), we evaluate accuracy, recall, abstention, coverage, and effective F1. We find that standard chain-of-thought prompting achieves the strongest overall operational performance, while few-shot prompting provides model-dependent benefits that are most pronounced for prompt-sensitive models. In contrast, adaptive chain-of-thought frequently suppresses recall and self-consistency induces excessive abstention, sharply reducing effective performance. These results show that vulnerability detection behavior is jointly determined by the model and the prompt, and that prompt sensitivity is a first-class system property that must be explicitly characterized in evaluation and deployment.
Authors: Yihan Wang, N. Asokan
Abstract: Memorization in large language models has been studied almost exclusively through prefix-conditioned extraction, a natural choice for autoregressive models. However, diffusion language models (DLMs) can denoise masked tokens at arbitrary positions. Thus, prefix-only probing reveals only one facet of memorization in DLMs and significantly underestimates the risk of training-data extraction. In order to realistically model extractability of training data in DLMs, we introduce \emph{infilling extraction}, a data-extraction protocol parameterized by an arbitrary binary mask that subsumes prefix-only probing and accounts for the bidirectional inductive bias of DLMs. Instantiating it on LLaDA-8B and Dream-7B across five extraction modes, three training pipelines, and three corpora covering verbatim and partial leakage, we find that mask geometry governs extractability: edge-conditioned masks \emph{extract up to three times more} verbatim sequences than prefix-conditioned ones, and bidirectional access opens channels inaccessible in autoregressive models. In particular, we show that a realistic adversary with access to training data where personally identifiable information has been redacted, can even achieve higher recall on extracting redacted email addresses from DLMs than from scale-matched autoregressive models. Tunable parameters for decoding measurably affect extraction performance, while a follow-up supervised finetuning stage does not eliminate the prior memorization.
Authors: Binglu Wang, Weixin Liang, Jiahui Xue, Yuhui Zhang, Hancheng Cao, Dashun Wang, Yian Yin
Abstract: Collaboration is the defining mode of modern science, yet its core mechanism -- feedback -- remains hard to observe, difficult to scale, and unequally distributed. Here we test whether large language models (LLMs) can contribute to this hidden but vital practice and reallocate scientific feedback, an essential yet scarce resource for knowledge production. In a global large-scale randomized field experiment, we delivered customized LLM-generated feedback for over 31,000 arXiv preprints across 150 fields and more than 45,000 researchers from 133 geographic regions. Relative to controls, authors who received feedback had a significantly higher likelihood of revising their manuscripts, corresponding to a 12.55% relative increase over the baseline revision rate. Exposure to AI feedback also increased authors' subsequent use of LLM tools in their future papers, suggesting longer-run shifts in scientific practice. These effects were strongest among authors from non-English-dominant research regions, manuscripts less embedded in the scholarly literature, and teams with lower h-indexes and earlier career stages, consistent with the idea that AI feedback may provide the greatest benefit where access to timely critique is otherwise limited. Together, these findings provide causal evidence that structured AI-based interventions can transform access to scientific feedback from a largely private advantage into a more widely distributed resource, with broader implications for productivity, equity, and capacity across the global research system.
Authors: Darek Kleczek, Fuheng Zhao, Alexander W. Lee, Julien Tissier, Pawel Liskowski, Ugur Cetintemel, Anupam Datta
Abstract: We introduce AvalancheBench, a benchmark for evaluating enterprise data agents through \emph{latent world recovery}. AvalancheBench improves on existing benchmarks in three ways. First, it evaluates analytical understanding rather than pipeline completion: systems are scored on whether they recover the segments, drivers, temporal events, and relationships that explain the data, not merely on whether they execute a workflow or produce a plausible report. Second, it provides ground truth for goal-driven analytics by generating observations from a known latent world, enabling partial credit for incomplete but valid recoveries. Third, it exposes how early analytical mistakes propagate into later conclusions: missed segments, merged events, or wrong attributions can lead to systematically wrong recommendations. In this sense, AvalancheBench complements real-data benchmarks by providing a controlled setting for diagnosing whether agents recover the analytical structure behind enterprise data. On a first e-commerce use case, the strongest configuration of a leading coding agent recovers only 26\% of the rubric, with failures concentrated in generic customer segmentations and merged temporal events.
Authors: Matthew Niedoba, Berend Zwartsenberg, Frank Wood
Abstract: The neural-network denoising functions which form the backbone of image diffusion models are remarkably consistent in their generalization behaviour across a wide variety of network architectures and training procedure hyperparameters. A recent line of research has sought to model the outputs of these networks by aggregating posterior weighted averages of training dataset patches. In this work, we consolidate these approaches into a unified model class which we call Filtered Posterior Mean Collections (FPMCs). We define this model class using query precision vectors, response weights, and source distributions, and illustrate that existing methods are recoverable with specific choices of these design axes. Investigating each axis in turn, we find that FPMC performance can be improved with soft relaxations of prior patch-based methods, and through augmentations of source distributions. Applying these findings to an existing FPMC, we demonstrate consistent sample improvement across three natural image datasets.
Authors: Mariam Barakat, Ekaterina Kochmar
Abstract: Analogies help learners understand unfamiliar concepts by relating them to known concepts. Despite recent advances, large language models (LLMs) continue to struggle to generate analogies of comparable quality to those produced by humans. We present a modular pipeline for educational analogy generation, decomposing the task into four stages: source finding, sub-concept generation, explanation generation, and evaluation. Grounded in Structure Mapping Theory, the pipeline enables systematic, stage-by-stage analysis of how model choice and input configuration affect analogy quality. We evaluate 12 state-of-the-art LLMs across six model families on two datasets with structured sub-concept annotations (SCAR and ParallelPARC), alongside seven embedding models for closed-setting retrieval. Our results show that sub-concepts substantially improve explanation quality and closed setting retrieval precision but provide limited benefit in open-ended source generation. We further introduce an LLM-as-a-judge evaluation methodology and validate its scoring against human annotations from seven annotators, finding that Claude Sonnet 4.6 aligns more reliably with human rankings than with fine-grained absolute scores. Taken together, our findings reveal cross-stage interactions that isolated studies cannot capture, and highlight sub-concept grounding as a key driver of analogy quality generation.
Authors: Siqi Li, Chuan Hong, Ziye Tian, Benjamin Sieu-Hon Leong, Koshi Nakagawa, Hideharu Tanaka, Sang Do Shin, Khuong Quoc Dai, Do Ngoc Son, Marcus Eng Hock Ong, Nan Liu, Molei Liu
Abstract: Deploying clinical prediction models across healthcare systems often fails when key training covariates are unavailable at deployment and labeled outcomes are limited in the target domain. For example, high-performing models for out-of-hospital cardiac arrest (OHCA) rely on detailed prehospital measurements routinely collected in high-resource settings but unavailable in many international registries. Existing methods either discard missing covariates, sacrificing predictive information, or rely on untestable assumptions about their target distribution. We propose DRUM (\underline{D}istributionally \underline{R}obust \underline{U}nsupervised transfer learning with structurally \underline{M}issing covariates), a framework that transfers prediction models to target populations where certain covariates are structurally absent and outcome labels are unavailable. DRUM partitions covariates into shared components ($X$), observed across all settings, and missing components ($A$), observed only in the source. Rather than imputing missing covariates, DRUM optimizes worst-case predictive performance over the unknown target distribution of $A \mid X$ using a neural network generator, with a robustness parameter controlling allowable deviation from the source conditional. We further develop a bias correction procedure that reduces sensitivity to nuisance estimation error. Simulations show substantial improvements in both mean and worst-case prediction error under distribution shift. Applied to cross-national OHCA prediction, transferring models from a US registry to multiple Asian registries where prehospital variables are unrecorded, DRUM yields better-calibrated predictions and improved clinical classification performance across sites.
Authors: Zhimin Zhao, Zehao Wang, Abdul Ali Bangash, Bram Adams, Ahmed E. Hassan
Abstract: Evaluation harnesses are software systems that orchestrate model evaluation by managing model invocation, data loading, metric computation, and result reporting. Despite their critical role in machine learning infrastructure, their operational challenges and engineering concerns have received limited attention so far. We present an empirical study of 57 evaluation harnesses, deriving a five-stage harness model and classifying 16,560 issues by workflow stage and root cause. Most harness operational challenges concentrate in the Specification stage (41.4% of issues), where harnesses integrate external models, datasets, and scoring judges. The three most frequent root causes of operational challenges are unimplemented features (24.3%), documentation gaps (20.3%), and missing input validation (17.2%), which together account for 61.7% of classified issues, spanning both defects in existing functionality and capability gaps that block intended workflows. Root causes also vary by workflow stage: environment incompatibility and external dependency breakage account for 36.2% of provisioning issues, whereas algorithmic error (25.9%) and validation gap (22.5%) dominate assessment issues. Together, these contributions establish an empirical foundation for treating evaluation engineering as a distinct software engineering concern.
Authors: Nesreen K. Ahmed, Nima Nafisi
Abstract: Monitoring autonomous large language model (LLM) agents for covert malicious behavior is challenging due to delayed, context-dependent, and long-horizon attack patterns. Agents may pursue hidden objectives while maintaining superficially benign behavior, making detection difficult even with full trajectory access. Prior monitoring approaches improve scaffolding or ensemble aggregation, but treat each trajectory independently and do not learn from prior monitoring experience. Moreover, standard reasoning methods explain observed behavior without explicitly reasoning about agent beliefs, intentions, and goal alignment required to distinguish benign task execution from covert deviation. We propose \textbf{Agent-ToM}, a learning-to-monitor framework grounded in Theory-of-Mind (ToM) reasoning for security analysis of autonomous agents. Agent-ToM performs structured full-trajectory analysis by inferring beliefs, intent hypotheses with calibrated confidence, expected actions, and deviations from task-consistent behavioral baselines. At inference time, it employs a \textit{Reason-Verify-Refine} pipeline to construct and validate monitoring decisions. At training time, Agent-ToM distills critique signals into a persistent \textit{semantic guardrail memory}, enabling reusable belief- and intent-conditioned constraints across episodes. We evaluate Agent-ToM on adversarial agent monitoring benchmarks (SHADE-Arena and CUA-SHADE-Arena). Agent-ToM achieves strong precision-recall balance and outperforms state-of-the-art monitoring baselines, including ensemble methods, while using a coherent two-call reasoning pipeline. These results demonstrate that learning at the monitoring layer, combined with structured ToM reasoning and verification, provides an effective and deployable foundation for securing autonomous LLM agents.
Authors: Yannik Dittmar, Marvin Jerome Stephan, Thomas V\"olkl, Matthias Hollick, Jiska Classen
Abstract: Many existing Artificial Intelligence (AI) solutions on mobile devices rely on an extensive collection of sensitive data, raising privacy concerns and often requiring storage for both context and model improvement. Apple's Private Cloud Compute (PCC) aims to address this by emphasizing mobile device integration and a privacy-first design. The central claim of PCC is that it does not store any user data and that user input and user accounts are unlinkable. While most of the PCC system specifications are public, compiled binaries add a layer of opaqueness. There are no reproducible builds, and there are no symbols within those binaries, creating potential discrepancies between the specification and what is shipped to the user. Additionally, the underlying models and interfaces for querying PCC are not openly accessible, limiting academic evaluation of model properties, such as accuracy. This poses a challenge in assessing whether a privacy-preserving approach like PCC is actually trustworthy while also providing high-quality answers. We are the first to reverse-engineer the PCC implementation on mobile devices to evaluate privacy aspects and to open its non-public interfaces on local devices to support custom PCC queries. We demonstrate this level of access beyond Apple's intended use cases by independently benchmarking the PCC model. We enable future research by making our PCC benchmarking framework publicly available.
Authors: Diogo Lavado, Alessandra Micheletti, Cl\`audia Soares
Abstract: In 3D scene understanding, deep learning models rely on large models and extensive training to capture basic geometric structures that are present in the 3D data. However, existing methods lack explicit mechanisms to incorporate geometric information, such as learnable primitive shapes, often necessitating large models and more training data which in turn increases cost and can limit generalization. We introduce GIBLy, a lightweight geometric inductive bias layer that integrates learnable geometric priors into 3D segmentation pipelines. GIBLy enhances existing architectures -- whether MLP-based, convolution-based, or transformer-based -- by providing features aligned with simple geometric shapes (and thus human-interpretable) that improve segmentation performance with minimal computational overhead. We validate our approach across multiple 3D semantic segmentation benchmarks, demonstrating consistent performance gains, including up to +11.5% mIoU on TS40K with PTV3, while adding only 58K extra parameters. Our results highlight the benefit of explicitly encoding geometric structure to support accurate and efficient 3D scene understanding, with a lightweight add-on layer
Authors: Konstantin Berlin, Adam Swanda
Abstract: Many automated labeling pipelines classify inputs into categories defined by a written specification, content moderation being a prominent use case. Simple category definitions are not detailed enough for labelers to produce the accurate, consistent golden labels these pipelines require. One solution is to write a prescriptive definition that settles enough real boundary cases that labelers cannot disagree with the written interpretation. In practice, definitions at that level of detail exceed what a human annotator can hold in working memory, so annotators fall back on intuition and the labels drift from the written rules, regressing on accuracy and consistency. We propose and demonstrate the efficacy of an AI-driven workflow in which AI helps write a per-category constitution that defines the label in enough detail to cover edge cases, and a frontier LLM interprets it on each input to produce the golden label more consistently and accurately than humans reading the same document. We evaluate on three content moderation categories (harassment, hate speech, non-violent crime) and show that the approach reduces cross-model inconsistency by up to 57x compared to paragraph definitions, with cross-model disagreement diagnosing specification gaps and the human responsible for high-level decisions about what each category should mean rather than individual labeling calls. For the safety evaluation, we introduce a dual-axis formulation scoring intent and content independently over the full conversation, so downstream consumers can act on either axis or both.
Authors: Alfredo Metere
Abstract: The Model Context Protocol (MCP) standardizes how a large-language-model (LLM) agent and an external tool server exchange messages, but not trust: a host reads a server's self-declared tool list and dispatches calls, with no notion of which servers it may use, at what sensitivity, or which of a server's tools are in bounds. This work grew out of a concrete need -- letting the Enclawed agent use Google's externally-operated MCP servers (Gmail, Calendar, Drive) safely, admitting the server and bounding the tools it may drive, without changing MCP or Enclawed's own tool application-programming interface (API). The mechanism we built, mcp-attested (shipped in both the open enclawed-oss distribution and the enclaved flavor), generalizes: the gap that makes an unmediated third-party connection unsafe for one user makes a regulated deployment impossible to accredit. We close it with three additive mechanisms: (1) a small, offline-signed clearance assertion a server publishes at a well-known Uniform Resource Identifier (URI) and a host verifies against a pinned trust root before any tool dispatch; (2) a deny-by-default per-server tool allowlist, so admitting a server is not trusting its every tool; and (3) a flavor-gated enforcement mode that turns the checks from warnings into hard denials, with every decision written to a tamper-evident audit log. We give the wire format, the verification algorithm, a security analysis, and an LLM-driven adversarial evaluation; we then state the design in normative Request-for-Comments (RFC 2119) form -- schema, verification rules, error registry, well-known registration, and machine-checkable conformance vectors -- so it can be adopted as an MCP addendum rather than reinvented. An unextended host ignores the well-known document and behaves exactly as today.
Authors: Zahra Rahimi Afzal, Wataru Uegami, Saghir Alfasly, Saba Yasir, Judy C. Boughey, Matthew P. Goetz, Krishna R. Kalari, H. R. Tizhoosh
Abstract: Digital pathology archives increasingly contain multiple whole-slide images (WSIs) per case, capturing spatially distinct tumour regions and reflecting intrinsic morphological heterogeneity. However, most existing approaches rely on a single pathologist-selected slide, thereby discarding potentially informative evidence distributed across the remaining WSIs. To date, no autonomous framework has been proposed for comprehensive multi-WSI case processing. Here, we present an unsupervised framework for case-level analysis that integrates information from all available slides within a case. Rather than relying on a single designated slide, the proposed approach constructs case-level representations by selectively distilling informative patches across WSIs. We introduce Clustering-Based Redundancy-Reduced Instance Sampling for Pathology (CRISP), a two-stage framework that first reduces redundancy within individual WSIs and subsequently applies clustering-based sampling to select a compact yet representative set of patches for the entire case. The resulting patch set captures case-level heterogeneity while avoiding exhaustive processing of gigapixel images, and directly serves as a retrieval index. Using two Mayo Clinic breast cancer datasets for diagnosis and treatment planning, we demonstrate that CRISP consistently matches or surpasses the current standard practice of combined model and pathologist slide selection for patient/case search and retrieval. By automating case-level processing and eliminating subjective WSI selection, CRISP potentially enables the exploitation of clinically relevant information distributed across multiple WSIs that is currently overlooked.
Authors: Lin Ai, Victor S. Bursztyn, Xiang Chen, Julia Hirschberg, Saayan Mitra
Abstract: Recent advances in large language models (LLMs) have enabled deep research systems that synthesize comprehensive, report-style answers to open-ended queries by combining retrieval, reasoning, and generation. Yet most frameworks rely on rigid workflows with one-shot scoping and long autonomous runs, offering little room for course correction if user intent shifts mid-process. We present SteER, a framework for Steerable deEp Research that introduces interpretable, mid-process control into long-horizon research workflows. At each decision point, SteER uses a cost-benefit formulation to determine whether to pause for user input or to proceed autonomously. It combines diversity-aware planning with utility signals that reward alignment, novelty, and coverage, and maintains a live persona model that evolves throughout the session. SteER outperforms state-of-the-art open-source and proprietary baselines by up to 22.80\% on alignment, leads on quality metrics such as breadth and balance, and is preferred by human readers in 85\%+ of pairwise alignment judgments. We also introduce a persona-query benchmark and data-generation pipeline. To our knowledge, this is the first work to advance deep research with an interactive, interpretable control paradigm, paving the way for controllable, user-aligned agents in long-form tasks.
Authors: Ahmed Sabbah, Mohammad Kharma, Mohammad Alkhanafseh, Radi Jarrar, Samer Zein, David Mohaisen
Abstract: Android malware detectors often degrade after deployment because of concept drift, while full retraining at each maintenance step is costly. We propose a chronological adaptive maintenance framework that models deployment-time maintenance as a sequential decision problem. The framework learns a stable latent representation through self-supervised learning during initialization, freezes the encoder, measures latent drift in the fixed representation space, and performs lightweight downstream adaptation using a trainable adapter and classification head. A proximal policy optimization controller selects low-cost maintenance actions based on the detector state, including current utility, retention on a fixed memory set, latent drift indicators, and update cost. We evaluate the framework under a causal deployment-style protocol on emulator and real Android malware datasets with static and dynamic features. Results show that the RL controller provides a strong cost-aware adaptation strategy, consistently remaining among the top-performing policies while achieving a favorable balance between temporal performance, memory retention, and maintenance cost under non-stationary deployment conditions.
Authors: Amirhossein Yousefiramandi, Ciaran Cooney
Abstract: Two questions regarding practitioners' use of patent embeddings arise: (i) Does one fine-tuning recipe suffice for all downstream applications? (ii) Is fine-tuning on one patent landscape sufficient for downstream application on other landscapes? By evaluating 22 pre-trained embedding models (ranging from 22M to 12B parameters) on three tasks -- information retrieval, classification, and clustering -- on 113,148 WIPO patents for assistive technology (46,069 citation queries) and on an external DAPFAM dataset, we find that two results cast doubt on the prevailing wisdom. (i) The optimal fine-tuning recipe depends on the downstream task: cross-sectional alignment (recipe R3) provides the largest improvements to retrieval performance (+7.1% nDCG@10), whereas a combined signal recipe (recipe R4) is better suited to classification (+7.1 F1) and clustering (+10.9 V-measure); a matched data control confirms that differences in training dataset size are not a contributing factor. (ii) Single-landscape fine-tuning hampers cross-landscape information retrieval: fine-tuning on one landscape significantly degrades cross-domain retrieval for 5 of 8 model-recipe combinations on the DAPFAM corpus, with the stronger zero-shot models suffering most. While within-family scaling is consistent (Qwen3 0.6B->4B->8B; Llama-Nemotron 1B->8B), cross-family scaling is erratic; the 12B KaLM-Gemma3 is ranked 8th on TAC retrieval performance, following prefix modification. Title+Abstract+Claims is the ubiquitous best text view, and all models suffer from a 55-65% gap between IN and OUT-of-domain performance which cannot be mitigated by hybrid BM25-dense fusion. Code and evaluation framework are publicly available.
Authors: Mohammed Kharma, Ahmed Sabbah, Mohammad Alkhanafseh, Mohammad Hammoudeh, David Mohaisen
Abstract: The growing use of Large Language Models (LLMs) for automated code generation has enhanced software development efficiency, but often at the cost of security. Generated code frequently overlooks critical concerns, leaving it vulnerable to issues such as weak encryption and improper input validation. To investigate this problem, we present a comprehensive empirical evaluation of the security quality of LLM-generated code across five LLMs and four programming languages (Java, C++, C, and Python), examining the impact of multiple prompt engineering methods. We introduce a weaknesses-aware zero-shot chain-of-thought (WA-0CoT) prompting strategy that enriches prompts with security context using CWE mappings to guide model reasoning. Our empirical analysis, supported by chi-square tests, finds no statistically significant reductions in vulnerability frequency or density across prompt methods. However, prompting strategies, including WA-0CoT, systematically influence the compositional distribution of CWE categories, with effects varying by programming language. These findings suggest that while security-aware prompting alters the structure of generated weaknesses, prompt engineering alone is insufficient to reliably reduce overall vulnerability levels. The results highlight the importance of language-aware and model-aware prompt design when evaluating the security properties of LLM-generated code.
Authors: Mohammed F. Kharma, Mohammad Alkhanafseh, Ahmed Sabbah, David Mohaisen
Abstract: Large language models (LLMs) are widely used for code generation, but their security reliability remains inconsistent across languages and prompting strategies. Existing prompt engineering improves functional correctness but rarely ensures consistent security outcomes. We introduce the \textit{Mitigation-Aware Chain-of-Thought (MA-CoT)} framework, which embeds task-specific CWE mitigation guidance and language-aware safeguards to reduce recurring vulnerabilities in generated code. We evaluate MA-CoT across three LLMs (gpt-5, claude-4.5, gemini-2.5), three programming languages (C, Java, Python), and four prompting strategies (Vanilla, Zero-shot, CoT, MA-CoT) on a 200-task primary dataset, with external validation on LLMSecEval. Using static analysis with expert validation, MA-CoT reduces total security findings from 92 to 39 (57.6\%) on the primary dataset and from 73 to 4 (94.5\%) on LLMSecEval. High-severity findings (Blocker + Critical) drop from 90 to 39 (56.7\%) and from 45 to 2 (95.6\%), respectively. Across both datasets, MA-CoT is the only strategy that consistently improves security reliability; Zero-shot and CoT are less reliable and may increase vulnerability, especially in C. We further introduce a strict layered attribution of vulnerability drivers (language-core vs. stack layers) and show that residual risk concentrates in hardening-oriented patterns (e.g., OS- and toolchain-dependent), motivating secure-by-construction primitives alongside prompting.
Authors: Inseo Lee, Yoonji Kim, Eugene Sohn, Jiwoong Lee, Jungmin You, Joonseok Lee, Jin-Hwa Kim
Abstract: Articulated object reconstruction from sparse-view images is an ill-posed problem that requires simultaneous inference of geometry and underlying articulation structure. Existing methods for articulated object reconstruction based on NeRF and 3D Gaussian Splatting (3DGS) typically rely on dense views or strong priors (e.g., depth maps, joint types, predefined number of joints) and require costly per-object optimization. In this paper, we propose ArtSplat, the first feed-forward framework for articulated 3D Gaussian Splatting. It reconstructs both geometry and joint parameters from sparse multi-view images across multiple articulation states in a single forward pass. To address the challenges of single-pass articulated reconstruction, we introduce a per-pixel joint map representation that enables the integration of joint parameter estimation into the feed-forward pipeline. We further propose a Cross-State Attention (CSA) mechanism with state tokens, which effectively captures discrete motion across input states. Experiments on 68 articulated objects from PartNet-Mobility, including both single- and multi-joint configurations, demonstrate that ArtSplat achieves competitive performance in both geometry and joint estimation, while being over 400 times faster than baselines.
Authors: Noel Thomas
Abstract: Standard accuracy on binary reasoning benchmarks hides critical failure modes: prior collapse, inconsistency under paraphrase, and inability to reason about parameter-dependent dynamics. We present ChaosBench-Logic v2, a 40,886-question benchmark over 165 dynamical systems with 27 FOL predicates and 78 axiom edges, together with CARE (Calibration- and Adversarial-Robust Evaluation), a protocol that surfaces these pathologies. Evaluating 14 models, we find that regime-transition reasoning remains near random (MCC = 0.05) even for frontier models, whereas FOL deduction with given premises reaches MCC = 0.52. Per-family decomposition shows that the proprietary-model advantage concentrates on cross-indicator (+0.40) and consistency tasks, while open-source Qwen 2.5-32B dominates indicator diagnostics (0.91 vs. 0.45). Two models exhibit negative MCC on bifurcation questions, confirmed as systematic anti-correlation via confusion-matrix analysis.
Authors: Minghao Li, Alicia Golden, Samuel Hsia, Michael Kuchnik, Adi Gangidi, Xu Zhang, Ashmitha Jeevaraj Shetty, Zachary DeVito, Weiwei Chu, Dong He, Haoci Zhang, Yuchen Hao, Ruoming Pang, James Hongyi Zeng, Ying Zhang, Minlan Yu, Carole-Jean Wu
Abstract: The rapid scaling of large language model training requires distributing GPU resources across multiple data center buildings and regions. We refer to such paradigm as "scale-across" training. As infrastructure expands, the system design space becomes increasingly intricate, encompassing new model architectures, hardware heterogeneity, and evolving communication patterns. Drawing from Meta's production experience, we highlight the complexities of deploying training jobs across a few data centers housing hundreds of thousands of GPUs. To accelerate exploration of the large design space and to enable efficient training for frontier model development, we conduct in-depth characterization of three key design dimensions: parallelism placement, parallelism scheduling, and network layer technologies. We then propose ScaleAcross Explorer, an optimizer that considers the interplay of design dimensions and holistically optimizes scale-across training. Testbed experiments and simulations demonstrate up to 64.62% training speedups over production configuration and up to 37.59% training speedups over the state-of-the-art baseline across a wide range of design points.
Authors: Xiaofeng Lin, Han Bao, Hisashi Kashima
Abstract: Estimating individual treatment effect (ITE) from observational graph data is crucial for decision-making in the fields such as commerce and medicine. This task is challenging due to interference, where individual outcomes can be influenced by the treatments and covariates of their neighbors. Existing methods attempt to model such interference for accurate ITE estimation. However, a critical issue is often overlooked: differentiated networked effect (DNE), an effect caused by local networks consisting of neighbors with varying importance and scales. Capturing DNE is vital; otherwise, we will end up with imprecise ITE estimation due to an erroneous characterization of interference, which can result in misguided decisions. To address this challenge, we propose a novel interference modeling mechanism that incorporates two partial attention mechanisms and a message amplifier. The partial attention mechanisms automatically estimate the importance of different neighbors in contributing to interference, while the message amplifier adjusts the results of the interference modeling mechanism based on the scale of neighbors, all of which enables the model to capture DNE. Experiments on three real-world graphs demonstrate that our methods outperform existing approaches for ITE estimation from graph data, which corroborates the importance of explicitly capturing DNE.
Authors: Kavin Soni, Debanshu Das, Vamshi Guduguntla
Abstract: Time series forecasting drives operational decisions in areas like finance, transportation, and energy. While supervised learning approaches achieve strong performance, they require domain-specific training, feature engineering, and ongoing maintenance. Large-scale foundation models have recently emerged as a zero-shot alternative, avoiding task-specific training much like LLMs. In this work, we evaluate foundation models against standard supervised approaches. Rather than focusing solely on aggregate accuracy, we analyze performance across four operational regimes: periodic human-centric systems, physically constrained processes, stochastic financial markets, and heterogeneous demand forecasting. Our results characterize optimal deployment areas. Foundation models perform well in domains with transferable periodic structures and are efficient for cold-start or long-tail scenarios. Conversely, supervised specialists maintain higher precision in systems governed by strict physical constraints. In financial domains, newer foundation models are rapidly closing the performance gap with supervised specialists. We further quantify trade-offs in inference latency, data drift adaptability, and deployment constraints. Finally, we propose a Complexity Router that assigns each series to the optimal model class using empirical features. We demonstrate that this selective routing achieves higher accuracy and significantly lower inference costs compared to deploying a universal foundation model, providing a practical framework for balancing generalization and efficiency.
Authors: Kritee Kondapally, Claire J. Smerdon, Pooja C. Patel, Ogheneyoma Akoni, Jevon Torres, Jaspreet Ranjit, Matthew Finlayson, Swabha Swayamdipta
Abstract: Language models (LMs) can exhibit systematic biases against speakers based on variations in their dialects, even in the absence of a dialect label, a behavior known as covert dialect bias. In this work, we quantify covert dialect bias in online discourse by evaluating how LMs associate stereotypical traits (derived from social psychology research on racial bias) with intent-equivalent tweets in Standard American English (SAE) and African-American Vernacular English (AAVE). While prior work shows that LMs associate more negative stereotypes with AAVE when evaluating tweets in isolation, we are surprised to find that this bias is significantly exacerbated when SAE / AAVE tweet pairs are compared side by side, a setting that more closely reflects high-impact decision making contexts in which models are used to rank candidates. The bias only worsens when dialect labels are explicitly specified. This is striking, given the extensive efforts from commercial developers to mitigate bias in their LMs. Encouragingly, we show that counterfactual fairness finetuning can mitigate covert dialect bias for some stereotypical traits, reducing average disparities when evaluating tweets in isolation, however, these improvements do not consistently hold across traits when evaluating SAE / AAVE tweets side by side. Our findings show that existing evaluation settings for covert dialect bias may underestimate its severity, specifically in contrastive settings. Additionally, overt dialect bias remains pronounced even after safety aligned finetuning, indicating that it remains an unresolved problem, and motivates the need for more robust evaluation and mitigation frameworks.
Authors: Dahoon Park, Jahyun Koo, Sangwoo Hwang, Jaeha Kung
Abstract: As the demand for deep learning grows, cost reduction through quantization has become essential for both training and inference. In 2022, the Open Compute Project (OCP) consortium standardized narrow precision formats for deep learning, called the microscaling (MX) format. The MX format is a hardware-friendly dynamic quantization scheme that effectively reduces the data size by sharing an 8-bit exponent across multiple operands. The MX format can be categorized into two types with their own strengths: (i) MXINT which focuses on a high precision consisting only of mantissa bits and (ii) MXFP which focuses on a wider dynamic range by allowing local exponent bits. In this work, we present a versatile MXFP format, called MX-SAFE (MXSF in short), that adaptively uses two modes, i.e., a wider mantissa mode (FP8 E2M5) and a subnormal FP mode (FP5 E3M2), to support both training and direct-cast inference. Furthermore, we propose a tile-based block design to increase hardware efficiency by reducing the burden of re-quantization process during the training with the MXSF format. Owing to the use of the proposed MXSF format, 0.05%/11.1% and 3.55%/3.57% improvements in accuracy, on average, for inference/full-training compared to MXFP8 E2M5 and MXFP8 E4M3 are observed, respectively. Moreover, we present a training-inference accelerator that supports the MXSF format and it achieves similar accuracy to the BF16 baseline while using 24.9% less total energy consumption.
Authors: Tarun Gehlaut, Difan Liu, Charu Bansal, Krutik Malani, Souymodip Chakraborty, Ankit Phogat, Matthew Fisher, Vineet Batra
Abstract: Recent vision-language model (VLM)-based approaches have achieved impressive results on image vectorization tasks. However, they are typically evaluated on synthetic benchmarks, where clean SVGs are rasterized at high resolution and then re-vectorized. As a result, these methods generalize poorly to real-world scenarios, such as images with unknown rasterization methods or those generated by text-to-image models. We introduce VectorArk, a new VLM-based model designed for robust and practical image vectorization. VectorArk employs a novel rounded polygon representation that simplifies the learning process while naturally producing smooth, visually appealing primitives. We also propose a degradation model that enhances robustness across diverse and imperfect inputs. Our experiments show that, in contrast to previous methods, VectorArk achieves superior geometric completeness and artifact suppression across multiple datasets, with comprehensive ablations validating the contribution of each component.
Authors: Aysin Tumay, Jiahe Huang, Elise Jortberg, Rose Yu
Abstract: We study sequential decision-making with offline reinforcement learning (RL). Traditional offline RL policies may result in out-of-distribution (OOD) actions when training relies only on sparse offline representations. To ensure safe offline policies in a sparse state-action space, we explore how density estimation models can be integrated into model-based RL methods to avoid the OOD regions. Generative models are capable of explicitly modeling the density in sparse state-action spaces. Building on this, we introduce Generative OOD-regularized Model-based Policy Optimization (GORMPO), a density-regularized offline RL algorithm that uses generative density modeling to restrict policy updates to high-density areas of the dataset. Furthermore, we examine whether better OOD detection corresponds to better model-based offline policies. We compare (1) the OOD detection capabilities of various density estimators and (2) their performance within the GORMPO framework on a real-world medical dataset and sparse offline RL datasets. We theoretically guarantee GORMPO's performance under mild assumptions. Empirically, GORMPO outperforms state-of-the-art baselines by 17% on a real-world medical dataset and enhances the base model on the offline RL datasets. Our empirical findings show that better OOD detection generally results in improved policies in environments with stable dynamics, while conservative penalties with poor density estimation are favored when dynamics are uncertain.
Authors: Ngoc Phu Doan, Chongyan Gu, Ihsen Alouani
Abstract: Batch Normalization (BN) is widely adopted to enable faster convergence and more stable training of deep neural networks. However, its impact on privacy and memorization has remained largely unexplored. In this work, we investigate the effect of BN layers on the memorization of atypical or outlier samples and its implications for privacy leakage. We conduct an extensive empirical study using three complementary approaches: (i) unintended memorization of out-of-distribution training samples, (ii) per-sample influence measured via gradient norms, and (iii) susceptibility to membership inference attacks (MIA). Across multiple datasets and architectures, we consistently observe that BN substantially increases the memorization of outliers compared to models without BN. Critically, this amplified memorization translates directly into privacy vulnerabilities: models with BN exhibit significantly higher susceptibility to MIAs. We complement our empirical findings with a theoretical analysis showing that BN amplifies the per-step influence of outlier samples during training, providing mechanistic insight into this phenomenon. Our results highlight an underappreciated privacy risk associated with BN and provide both practical and theoretical insights into how normalization layers can amplify the influence of rare or sensitive training examples.
Authors: Jingchu Gai, Nai-Chieh Huang, Jiayun Wu
Abstract: The residual update of a pre-norm Transformer layer admits an interpretation as one step of a first-order optimizer acting on a surrogate token energy, wherein the attention and MLP sublayers function as gradient oracles. Based on this observation, we build a family of optimizer-inspired Transformers (triple-momentum, Adam/AdamW, Muon, SOAP) and compare them under matched compute. In our main pretraining experiment, the triple-momentum TMMFormer achieves the lowest validation loss, outperforming the vanilla Transformer and prior architectural variants. A controlled ablation and supporting theory show that momentum, not preconditioning, is the main source of the gain. We further show that TMMFormer and other momentum-based designs reach flatter minima than the vanilla Transformer, which leads to less forgetting and better generalization.
Authors: Volodymyr Ovcharov
Abstract: Legal NLP benchmarks evaluate models on randomly split data, implicitly assuming that legal language is stationary. We test this assumption by fine-tuning four transformer encoders -- XLM-RoBERTa (base and large) and their legal-domain variants -- on Ukrainian court decisions from three temporal epochs defined by geopolitical disruptions: pre-war (2008-2013), hybrid war (2014-2021), and full-scale invasion (2022-2026). Each model is trained on one epoch and evaluated on all three, producing a 3x3 cross-temporal generalization matrix. Four findings emerge. (1) Forward degradation is severe: models trained on pre-war data lose up to 27.2 percentage points of macro-F1 when applied to full-scale invasion era decisions. (2) The degradation is asymmetric: backward transfer (full-scale to pre-war) is substantially more robust than forward transfer, consistent with the hypothesis that legal language is additive. (3) Legal-domain pretraining (Legal-XLM-R) does not improve absolute performance but reduces forward degradation magnitude and asymmetry. (4) Chronological continual learning eliminates catastrophic forgetting for general XLM-R: pre-war knowledge is fully retained (+1.8 to +6.2 pp) while full-scale performance gains +16.5 to +19.0 pp; reverse-chronological training causes severe forgetting. Cross-jurisdictional pretraining on Swiss Judgment Prediction data improves absolute performance but does not reduce temporal degradation magnitude, confirming that temporal drift is an intrinsic property of legal language evolution. The dataset (428K decisions across three epochs) is publicly available as a LEXTREME contribution.
Authors: Alin-Gabriel V\u{a}duva, Anca-Ioana Andreescu, Simona-Vasilica Oprea, Adela B\^ara
Abstract: Large Language Model (LLM)-based code analysis tools are adopted to automate software documentation tasks. However, the scalability of these approaches to real codebases, where Intermediate Representations (IR) exceed LLM context limits, remains underexplored. This paper introduces an agentic architecture with context engineering for automated UML diagram generation from source code repositories. It employs a hierarchy of five specialized agents: PlannerAgent, AnalyzerAgent, DiagramAgent, CorrectorAgent and DependencyAnalyzerAgent, built on the Claude Agent SDK, each addressing a distinct cognitive subtask. A deterministic, importance-weighted IR compaction layer transforms full project IRs into diagram-specific views guaranteed to fit within token constraints, requiring no LLM calls and completing in milliseconds. Thus, we evaluate the system across 12 open-source repositories in 4 programming languages (Java, JavaScript, PHP, Python) and 7 UML diagram types, producing 84 observations assessed on 5 automated metrics. Results demonstrate high syntactic validity (mean: 91.5%, with component and deployment diagrams reaching 100%), strong relationship precision (mean: 0.858) and consistent structural quality (mean: 81.7/100, with cross-language variance of 3.1 points). Entity recall averaged 0.313, reflecting deliberate architectural prioritization over exhaustive coverage. A sensitivity analysis (31 to 4,578 IR entities) confirms that quality scores remain stable regardless of scale.
Authors: Imesh Ekanayake, Elham Naghizade, Jeffrey Chan
Abstract: The integration of fairness and privacy in centralized data-driven applications is critical, especially as these systems increasingly influence sectors with significant societal impact. Current methods rarely address privacy, fairness, and accuracy together, which can potentially compromise ethical standards and privacy regulations. However, balancing these three objectives is quite challenging since each of objective often imposes conflicting requirements on the design and training of models, making it difficult to optimize one without compromising the others. This paper introduces a novel multitask adversarial model that treats fairness and privacy as integral objectives rather than afterthoughts, and learns a latent representation that hides sensitive attributes while preserving essential task-related information. Our approach dynamically balances fairness with accuracy and privacy through an optimized cost function with minimal performance loss even under strict conditions. Extensive testing on diverse datasets shows the ability of our model to achieve high standards of fairness and privacy without significant sacrifice to accuracy. Benchmarking against state-of-the-art privacy and fairness standards shows that our method enhances the robustness of privacy, fairness, and accuracy optimization, proving its adaptability across various datasets.
Authors: Alif Tri Handoyo, Vincent C. S. Lee, Rizka Widyarini Purwanto, Alex M. Lechner, Deanna Kemp, Muhamad Risqi U. Saputra
Abstract: Automatically mapping and segmenting global mining footprints using remote sensing and deep learning is critical for monitoring the socio-environmental risks and impacts of mining, yet its progress is hindered by the scarcity of fine-grained annotated data. Although large-scale datasets with coarse boundaries are widely available, leveraging them to improve fine-grained segmentation is challenging due to significant domain shift. To address this, we propose MineC2FNet, a coarse-to-fine domain incremental learning framework that exploits abundant coarse data to enhance fine-grained mining footprint segmentation. MineC2FNet adopts a teacher-student architecture with attentive distillation at both the feature and prediction levels, selectively transferring generalized knowledge from the coarse domain while enabling boundary refinement using limited fine-grained data (fine domain). We further introduce an expertly validated dataset of 219 images with precise boundary annotations across diverse geographies and commodities. Extensive experiments against state-of-the-art approaches, including domain adaptation and domain incremental learning methods, demonstrate that MineC2FNet achieves superior performance while effectively handling domain shift. The dataset and code are publicly available at https://github.com/risqiutama/MineC2FNet.
Authors: Siyuan Duan, Yuan Sun, Dezhong Peng, Yingke Chen, Xi Peng, Peng Hu
Abstract: Trusted multi-view classification aims to deliver reliable fusion for accurate predictions and has recently attracted substantial attention in both academia and industry. However, existing TMVC methods typically assume strict alignment across different views during both training and testing phases, which is often impractical in real-world scenarios. This limitation motivates us to revisit TMVC and extend it to a more challenging setting: how to mitigate the impact of view conflict (VC) during both training and inference. To tackle this setting, existing TMVC methods suffer from three critical limitations: underestimated uncertainty, misleading decisions, and overfitting to VC. To address these issues, this paper proposes a novel Robust Fuzzy Multi-View Learning (R-FUML) framework grounded in Fuzzy Set Theory. Specifically, R-FUML models network outputs as fuzzy memberships to quantify category credibility and uses an entropy-based method for reliable multi-view fusion. To this end, we present a Robust Multi-view Fusion (RMF) strategy that accounts for both view-specific uncertainty and inter-view conflicts, thereby alleviating the adverse impacts of VC on decision-making. To identify and conquer VC during training, we further design a Robust Learning Against VC (RLVC) framework. RLVC isolates conflicting samples by leveraging neural networks' memory effects and then retrains the model by applying a penalty to these conflicting views. Extensive experiments across eight public datasets demonstrate that R-FUML consistently outperforms 15 state-of-the-art baselines in robustness and uncertainty estimation. The code will be released upon acceptance.
Authors: Ruihao Xu, Xingming Shui, Jingxuan Niu, Yiqin Wang, Jilin Yu, Haoji Zhang, Yansong Tang
Abstract: As AI-powered compliance monitoring becomes increasingly important in public governance and industrial safety, the ability to provide verifiable evidence and traceable accountability signals is essential. However, existing video anomaly detection datasets focus on event-level binary classification, lacking the rule-driven, explainable analysis required for real-world compliance scenarios. We introduce FoodMonitor, a benchmark for explainable compliance analysis in commercial kitchen surveillance. FoodMonitor comprises 477 video clips with 3,307 violation annotations across a dual-channel design covering both person-level and environment-level violations. Each annotation specifies which rule was violated, what non-compliant behavior occurred, and who committed it with frame-level bounding boxes. We establish a unified evaluation protocol with a two-stage matching mechanism that separately assesses spatial localization and semantic understanding, along with a composite metric ($C_{\text{score}}$) that balances environment and person detection performance. Systematic evaluation of several state-of-the-art multimodal large language models reveals that the best-performing model achieves only 0.360 $C_{\text{score}}$, with spatial localization and fine-grained rule understanding emerging as the primary bottlenecks. Our analysis identifies two distinct failure modes: localization-dominated errors and semantics-dominated errors, providing diagnostic insights for future model development.
Authors: Ofir Abramovich, Nadav Z. Cohen, Adi Rosenthal, Ariel Shamir
Abstract: Latent video diffusion models generate videos by progressively transforming Gaussian noise into realistic samples conditioned on text or visual inputs. However, existing conditioning methods often require additional training and computational overhead. Motivated by recent findings on the importance of frequency components in generative models, we propose a simple, training-free approach for motion-conditioned video generation by injecting low-frequency phase information from a reference video directly into the diffusion noise latents. Our method transfers motion cues without modifying the model architecture or inference pipeline. Using several applications, we demonstrate effective control over both appearance and dynamics in generated videos, while achieving competitive or superior results compared to more complex conditioning approaches.
Authors: Min Tang, Fanqi Kong, Linyuan L\"u, Xue Feng
Abstract: Mixed-motive scenarios are ubiquitous in real-world multi-agent interactions, where self-interested agents often defect for immediate rewards, overlooking the potential of altruistic cooperation to improve long-term gains and collective welfare. Peer punishment can deter defection, but as costly second-order altruism, its persistent imposition may undermine the punisher's interests. Existing approaches often struggle to effectively implement punishment to promote cooperation. To balance the efficacy and cost of punishment, we propose Adaptive Punishment for Cooperation (APC), a distributed method that determines punishment intensity based on both a dynamic punishment probability and the severity of defection. This dynamic probability substantially reduces costly and ineffective punishment while also promotes cooperation. To accurately assess defection and its severity, we use a defection awareness module, whose learning is guided by game reward. Theoretical analysis and empirical results show APC performs effectively in iterated public goods game. Empirically, APC also significantly outperforms existing baselines across sequential social dilemmas, learning rational and effective punishment policies that foster cooperation by strategically deterring defection.
Authors: Spandan Pratyush
Abstract: The quadratic complexity of self-attention in Transformer models remains a significant bottleneck for processing long sequences and deploying large language models efficiently. For this approach, there has been significant research into Sparse Attention, and Deepseek Sparse Attention has combined various methods of creating segments of tokens to reduce the time complexity. This paper introduces a novel approach, Grammatically-Guided Sparse Attention, which constrains attention computations based on the grammatical roles of tokens. By leveraging Parts-of-Speech (POS) tags, attention masks are dynamically generated that enforce linguistically coherent connections between tokens, reducing the computational graph without sacrificing essential linguistic dependencies. Two masking strategies are proposed and evaluated: a hard mask that strictly allows only predefined grammatical interactions, and a soft mask that biases attention towards these interactions. The experiments, conducted on the SST-2 sentiment classification task using a DistilBERT-like architecture, demonstrate that Grammatically-Guided Sparse Attention maintains comparable accuracy to full attention while significantly reducing the theoretical computational overhead. Preliminary results show accuracy values of 0.8200 for hard masking and 0.8165 for soft masking, closely matching the 0.8200 of full attention, providing a path towards more efficient, interpretable, and linguistically-informed Transformer architectures.
Authors: Sassan Mokhtar, Lars Doorenbos, Fatemeh Jabbari, Marius Bock, Dominik Bach, Juergen Gall
Abstract: Interactive assistance systems typically provide feedback after an action has been completed, supporting error recovery but not preventing the error itself. We present TRAFA, a real-time predictive feedback system for procedural tasks that intervenes before errors are committed. TRAFA operationalizes predictive feedback through a Track-Forecast-Act framework that tracks hand and object state, forecasts user motion conditioned on scene context, and triggers feedback when a predicted action is likely to violate task constraints. We instantiate this pipeline in a sequential assembly setting and evaluate it through both technical benchmarking and a controlled user study against conventional reactive feedback. Our results show that predictive feedback improves task accuracy and efficiency while maintaining a comparable number of feedback events. These findings position feedback timing as a key dimension in system design and show how real-time anticipation can be integrated into interactive systems to prevent errors before they occur.
Authors: Botao Amber Hu, Helena Rong
Abstract: Every major framework for governing artificial intelligence presupposes an identifiable entity -- a developer, deployer, or operator -- who can be held responsible and compelled to comply. Decentralized AI (DeAI) dissolves this presupposition. We analyze DeAI as a six-layer decentralizing stack -- model, training, compute, harness, identity, and ownership -- and show how partial decentralization across layers compounds into what we call the \emph{governance vacuum}: a condition in which AI systems are consequential enough to require governance but lack the properties that existing frameworks presuppose in their targets. This vacuum takes two analytically distinct forms: an \emph{accountability gap}, where no addressable principal can be identified, and an \emph{incapacitation gap}, where even an identified principal cannot alter the running system. We demonstrate that these failures are not merely jurisdictional but defeat every presupposition of governance through normative address -- the communication of rules to a comprehending, responsive agent. Drawing on Lessig's modalities of regulation and Searle's distinction between regulative and constitutive rules, we argue for a shift in the locus of governance from policy to protocol, from normative address to architectural constraint. Protocol-based constitutive governance does not address the agents operating within a system but shapes the substrate that determines what kinds of actions are possible within it. We identify four ethical conditions -- legitimacy, contestability, transparency, and non-domination -- that such governance must satisfy to avoid degenerating into unaccountable technocratic power, and we argue that the central political challenge of governing AI in a decentralized world is reconstructing forms of democratic authorization for architectural choices that persist after the ordinary chain of policy has broken down.
Authors: Natalia Trukhina, Vadim Vashkelis
Abstract: Text compression for large language model (LLM) systems is usually framed as token deletion, retrieval, summarization, or exact reconstruction. We study a more aggressive but explicitly lossy setting: compress text into compact codes that an LLM can expand into task-relevant meaning. We call this setting SemanticZip. Unlike lossless compression, SemanticZip does not require byte-identical reconstruction; unlike ordinary summarization, it treats model-based decompression as part of the codec and evaluates whether task-relevant semantic commitments are recovered. This paper is a pilot framework, not a benchmark claim. We formalize LLM-mediated decompression, define a protected/lossy packet architecture, and evaluate six representation regimes over five author-constructed diagnostic cases: structured prose, JSON, CCL-Core, CCL-Min, SemanticZip ASCII, and SemanticZip emoji. An independent decoder LLM reconstructs typed semantic atoms from each compressed representation, and we score Critical Atom Recall, Weighted Atom Recall, precision, and tokenizer gain. In this pilot, structured prose has the highest recoverability, with WAR = 0.956 and 19.1% o200k_base token gain. CCL-Min is the strongest balanced point, with 39.4% token gain and WAR = 0.874. SemanticZip ASCII provides the largest useful compression, with 46.5% token gain and WAR = 0.802, while emoji-heavy SemanticZip performs worse on both compression and recovery. The main contribution is not the claim that these numbers establish a universal frontier. Rather, we introduce a reproducible experimental interface for studying lossy, LLM-decompressible text codes and a design principle: safety-critical and exact commitments should remain protected, while predictable low-risk context may be semantically zipped.
Authors: Petar Radanliev
Abstract: This paper examines the erosion of Public Key Cryptography (PKC) security under adaptive adversarial optimisation driven by artificial intelligence. The problem addressed is the growing mismatch between algorithm-centric cryptographic security models and operational attack realities, where adversaries exploit implementation-level observability rather than breaking cryptographic primitives.
Authors: Jiaheng Wei, Yanjun Zhang, He Zhang, Leo Yu Zhang, Chao Chen, Kok-Leong Ong, Jun Zhang, Yang Xiang
Abstract: Federated learning (FL) increasingly needs machine unlearning to comply with privacy regulations. However, existing federated unlearning approaches may overlook the overlapping information between the unlearning and remaining data, leading to ineffective unlearning and unfairness between clients. In this work, we revisit federated unlearning through the lens of memorization. We argue that unlearning should mainly remove the unique memorized information attributable to the data to be forgotten, while preserving overlapping patterns that are also supported by the remaining data. Specifically, we propose Grouped Memorization Evaluation, an example-level metric that separates memorized knowledge from overlapping knowledge. Building on this metric, we introduce Federated Memorization Pruning (FedMemPrune), a pruning-based unlearning approach that resets redundant parameters responsible for memorization. Extensive experiments show that FedMemPrune closely matches retraining-based unlearning baselines while more effectively eliminating memorization than existing federated unlearning algorithms, yielding strong unlearning performance without sacrificing the utility of retained knowledge.
Authors: Naman Mishra, Shankar Gangisetty, C. V. Jawahar
Abstract: Pedestrian intention and trajectory prediction are critical for the safe deployment of autonomous driving systems, directly influencing navigation decisions in complex traffic environments. Recent advances in large vision-language models offer a powerful new paradigm for these tasks by combining high-capacity visual understanding with flexible natural language reasoning. In this work, we introduce PedestrianQA, a large-scale video-based dataset that formulates pedestrian intention and trajectory prediction as question-answering tasks augmented with structured rationales. PedestrianQA expresses richly annotated pedestrian sequences, in natural language, enabling VLMs to learn from visual dynamics, contextual cues, and interactions among traffic agents while generating concise explanations of their predictions without needing specialized architectures tailored for each task. Empirical evaluations across PIE, JAAD, TITAN, and IDD-PeD show that finetuning state-of-the-art VLMs on PedestrianQA significantly improves intention classification, trajectory forecasting accuracy, and the quality of explanatory rationales, demonstrating the strong potential of VLMs as a unified and explainable framework for safety-critical pedestrian behavior modeling.
Authors: Sattam Altuuaim, Lama Ayash, Muhammad Mubashar, Naeemullah Khan
Abstract: Despite the central role of optimization in deep learning, most optimizers rely on update structures whose functional form is fixed before training begins. This static design can limit their ability to respond to changing gradient behavior across the loss landscape, where training may shift between stable, noisy, and inconsistent regimes. This study proposes PILOT (Policy-Informed Learned OpTimizer), an online optimizer that adapts its update behavior during training. Rather than using a fixed balance between momentum, normalization, and sign-based updates, PILOT uses gradient-direction agreement as a signal of local training stability. Conditioning the update rule on this agreement signal allows the optimizer to adjust its behavior when gradients become stable, noisy, or inconsistent. Experiments on FashionMNIST and CIFAR-10 show that PILOT consistently achieves the highest accuracy among the evaluated optimizers across convolutional settings. On the CNN architecture, PILOT reaches 94.13% on FashionMNIST and 81.94% on CIFAR-10. On ResNet-18, it further improves performance, reaching 95.71% on FashionMNIST and 93.42% on CIFAR-10. These results suggest that learning how to adapt the update structure during training can improve performance across both compact and deeper convolutional models while preserving a simple first-order optimization framework. The implementation of PILOT is publicly available at https://github.com/SattamAltwaim/PILOT.git
Authors: Jordan F. McCann
Abstract: Independently trained transformers compute the same function in residual-stream bases that differ by a uniform random rotation on $\mathrm{SO}(d_{\mathrm{model}})$. We call this phenomenon polymorphism: same function, mutually unintelligible interior coordinates. One matrix multiplication per model pair removes it: an orthogonal Procrustes fit on a single batch of activations transfers sparse-autoencoder feature dictionaries and steering vectors between independently trained models, with no retraining. The phenomenon is invisible to the standard SAE universality metric. Decoder-column cosine similarity matches across seeds at 98%, the SAE-universality headline number, while an SAE trained on one seed reconstructs another seed's activations at negative explained variance, worse than predicting the constant mean. The decoder columns align; the encoder reads from a rotated frame. A single Procrustes rotation $R$ restores reconstruction to within 0.025 EV of the within-seed ceiling at every internal site. $R$ is Haar-distributed: $\|R - I\|_F$ matches the random-orthogonal prediction $\sqrt{2 d_{\mathrm{model}}}$ to 0.1% at $d_{\mathrm{model}} = 512$, and a Kolmogorov-Smirnov test of $R$'s eigenvalue spectrum against Haar $\mathrm{SO}(d_{\mathrm{model}})$ returns $p \approx 1.000$ pooled and per-pair. Diff-of-means steering vectors transfer in three regimes by alignment with $R$'s invariant subspace: clean when pinned by shared output weights, partial when overlapping the rotated subspace, inverted otherwise. With no shared I/O (Pythia), all three collapse to universally inverted. The same rotation account holds across training checkpoints within a single run. Validated on a 104k-parameter Dyck-3 transformer and nine independently-trained Pythia-70m seeds on The Pile, via a pre-registered four-bar operational framework. Frontier-scale (10B+) replication remains open.
Authors: {\L}ukasz Struski, Hanna Blazhko, Piotr Kubaty, Jacek Tabor
Abstract: Fast linear algebra in deep learning usually comes with a choice: fixed geometry and exact computation, as in the Fourier transform, or adaptive geometry paid for by dense parameters, random features, or low-rank surrogates. To move beyond this trade-off, we introduce LAPLEX, a class of exact, trainable (phased) Laplace-kernel operators. A LAPLEX layer is a typically full-rank dense matrix, implicitly defined by learnable coordinate anchors, with FFT-like scaling. Consequently, it supports trainable matrix--vector operations at vector dimensions up to $10^9$ on modern GPUs. As a neural layer, it yields compact projections and classification heads interpretable as soft, trainable routing models. The same primitive also serves as an efficient Gram operator, enabling high-dimensional covariance models on flattened images of dimension $3 \cdot 10^6$ that preserve visible spatial structure without imposing convolutional bias. These applications reflect a single principle: dense geometry can be learned without storing a dense matrix, which enables data-adaptive global interactions in regimes where ordinary dense layers are out of reach. In this sense, LAPLEX separates expressivity from storage cost: it behaves like a dense trainable matrix, but is represented and applied through a small structured set of parameters.
Authors: Quanjiang Li, Zhiming Liu, Wei Luo, Tingjin Luo, Chenping Hou
Abstract: Multimodal large language models (MLLMs) frequently suffer from object hallucinations, yet the visual perceptual mechanism underlying this failure remains poorly understood. In this work, we reveal that hallucinations are strongly associated with a human-like attention distraction phenomenon, where humans under divided focus experience degraded visual clarity and produce inaccurate descriptions, while in models the same mechanism manifests as spatial inconsistency in multi-head attention and temporal fading of attention to image tokens during decoding. We further provide theoretical insights that attention dispersion increases model complexity and degrades classification generalization. Motivated by these findings, we propose an Attention-Focused Approach for Improved Image Perception (AFIP), which corrects attention distraction via cross-head attention enrichment and reinforces visual grounding through dynamic historical attention enhancement. Extensive experiments on multiple benchmarks and models validate the effectiveness of AFIP without additional training.
Authors: Mustafa Kadhim, Viktor Rogowski, Emilia Persson, Camila Gonzalez, Andr\'e Haraldsson, Sofie Ceberg, Mikael Nilsson, Malin K\"ugele, Sven B\"ack, Christian Jamtheim Gustafsson
Abstract: Artificial intelligence is increasingly integrated into radiotherapy workflows, yet such pipelines remain vulnerable to out-of-distribution image data that may introduce unexpected behavior in clinical tasks. Deep learning-based anomaly detection for pelvic magnetic resonance imaging (MRI) remains largely unexplored, and transparent evaluation of its feasibility for full automation is limited. We developed and evaluated a fully automated, unsupervised anomaly-detection framework for pelvic and brain MRI. A two-stage framework was trained on reference images from public datasets: LUND-PROBE for pelvic MRI, and IXI, fastMRI, and fastMRI+ for brain MRI. In the first stage, MRI slices were compressed into discrete tokens; in the second, the distribution of normal tokens was modeled. Anomaly evidence was estimated by combining perceptual image differences with token-surprisal scores based on negative log-likelihood. Automated detection was evaluated on pelvic MRI with synthetic global and real clinical anomalies, and on brain MRI with clinically annotated fastMRI+ abnormalities. Sensitivity, specificity, area under the receiver operating characteristic curve (AUC), and false-positive behavior in held-out normal cases were assessed. The framework achieved robust detection across hidden evaluation cohorts, with AUCs of 0.97 (95% CI, 0.95-0.98) and 0.81 (95% CI, 0.74-0.87) for pelvic and brain MRI, respectively. Heatmap analysis showed strong spatial agreement between detected anomalies and ground-truth locations, supporting localization accuracy and interpretability. These results support the potential of unsupervised anomaly detection as an automated MRI quality-control layer for radiotherapy workflows, with transparent visualization of image regions likely to compromise downstream AI-based tasks.
Authors: Haizhou Xia
Abstract: Post-hoc repair of LLM mathematical reasoning introduces an asymmetric risk: fixing an incorrect reasoning trace is useful, but replacing a trace that was already correct can be harmful. We study this problem under a selective replacement setting, where a system must decide whether a repaired candidate is safer than preserving the original cached trace. We present GuardedRepair, a guarded best-of-N repair framework that diagnoses cached reasoning traces, selectively triggers repair, and accepts answer-changing candidates only when deterministic verification guards support replacement. The framework combines lightweight symbolic checks, surface semantic-risk diagnostics, bounded candidate generation, and conservative acceptance policies. On the full GSM8K test set, where the initial reasoner already achieves 95.60% accuracy, GuardedRepair improves final accuracy to 96.89%, fixing 17 of 58 remaining errors without measured broken-correct cases in the main run. On a weak-reasoner ASDiv setting, accuracy improves from 78.40% to 87.60%. Direct regeneration baselines show that this gain is not explained by stronger-model re-solving alone: re-solving all GSM8K examples lowers accuracy to 93.03% and breaks 47 initially correct answers. Additional analyses show that guarded repair substantially improves the fixed/broken tradeoff, while also revealing that replacement risk is reduced rather than eliminated. These results support viewing post-hoc repair as harm-aware selective replacement rather than unconstrained re-solving.
Authors: Jaeung Lee, Dohyun Kim, Jaemin Jo
Abstract: Large language model (LLM) unlearning has emerged as a crucial post-hoc mechanism for privacy protection and AI safety, yet auditing whether target knowledge is truly erased remains challenging. Existing output-level metrics fail to detect when this knowledge remains recoverable from internal representations. Recent white-box studies reveal such residual knowledge but often rely on auxiliary training or dataset-specific adaptations, leaving no generalizable metric. To address these limitations, we propose the Unlearning Depth Score (UDS), a metric that quantifies the mechanistic depth of unlearning via activation patching. UDS first identifies layers that encode the target knowledge using a retain model baseline, then measures how much of it is erased in the unlearned model on a 0-1 scale. In a meta-evaluation across 20 metrics on 150 unlearned models spanning 8 methods, UDS achieves the highest faithfulness and robustness, confirming our causal approach as the most reliable for unlearning evaluation. Case studies further reveal that white-box metrics can disagree at the layer level and that erasure depth varies across examples. We provide guidelines for integrating UDS into existing benchmarking frameworks and streamlining the evaluation pipeline. Code and data are available at https://github.com/gnueaj/unlearning-depth-score
Authors: Ghassen Marrakchi, Basarab Matei
Abstract: Scattering transforms achieve Lipschitz stability and translation invariance, but dense prediction tasks require preserving spatial structure lost in global averaging. We propose Phase-Aware Scattering Encoder-Decoder, which restores this information by explicitly preserving phase in skip connections. On image denoising (BSD68), breaking translation invariance improves PSNR by $+2.17$~dB; phase preservation adds $+1.03$~dB. A novel spatial shuffling ablation ($-1.26$~dB penalty) demonstrates phase encodes location-dependent structure. We conduct a preliminary extensibility study on a second dense prediction task (ISIC skin lesion segmentation), with full cross-validation as ongoing work. This work advances principled wavelet-deep learning integration, showing how phase information complements scattering's stability-expressiveness trade-off in pixel-level prediction.
Authors: Sol Park, Soobin Um
Abstract: Minority sampling aims to generate low-density instances on a data manifold and is of central importance in applications such as medical diagnosis, anomaly detection, and creative AI. Existing approaches, however, define minority samples relative to generative priors learned from training data, confining rarity to model-specific notions that may poorly reflect real-world semantics. In this work, we propose a world-centric perspective on minority sampling, which defines rarity with respect to real-world priors rather than generator-induced densities. To this end, we introduce JEPA guidance, a diffusion sampling framework guided by a Joint-Embedding Predictive Architecture (JEPA) -- a class of world models that encode broad, semantically rich representations. JEPA guidance steers diffusion trajectories toward low-density regions under the implicit density induced by the JEPA, thereby aligning generated minorities with real-world semantic rarity. To make JEPA guidance computationally practical, we develop principled approximation strategies accompanied by theoretical error bounds, significantly reducing the overhead of guidance computation. Extensive experiments across unconditional, class-conditional, and text-to-image generation demonstrate that JEPA guidance consistently improves the fidelity and semantic validity of minority samples, outperforming generator-centric baselines in capturing real-world notions of rarity. Code is available at https://github.com/soobin-um/jepa-guidance.
Authors: Alfredo Pesoli, Herman Errico, Lorenzo Cavallaro
Abstract: Recent demonstrations of large language models producing candidate and confirmed vulnerabilities in production software have renewed the narrative that AI will reshape offensive and defensive security. Headlines emphasize capability; they rarely interrogate costs and incentives. This paper examines LLM-driven vulnerability discovery through a bugonomics lens: the operational economics of producing, proving, prioritizing, and fixing security-relevant defects. Historically, the most visible high-end bugonomics was offense-priced because production-grade zero-days and exploit chains were expensive specialist outputs for governments, brokers, and offensive vendors. Defender-side bugonomics already existed in vulnerability research, reward programs, and vendor remediation work; LLM-assisted systems change its scale and distribution. They make candidate generation, code comprehension, harness construction, proof-of-impact drafting, and report preparation cheaper at codebase scale. Exploits and proofs of concept remain important, but in defender workflows they primarily prove impact, guide prioritization, and justify remediation. The resulting bottleneck is not only finding more bugs; it is absorbing, validating, triaging, patching, and shipping a larger stream of reports. Using public data from Anthropic's Mythos Preview and Mozilla Firefox collaborations, along with public exploit-market price anchors and vulnerability reward programs, we argue that the near-term shift is not simply more zero-days. It is a move toward broader defender remediation throughput: low-signal candidates become cheaper, evidence-rich remediation become more important, and scarce capacity shifts toward maintainer review and release work. The effect is acute in open source, where LLM-assisted discovery can increase report volume while maintainer-side validation, triage, funding, and release capacity may not scale.
Authors: Ruihao Xu, Yong Liu, Yansong Tang, Sule Bai, Xubing Ye, Bingyao Yu, Yutao Guo, Jiwen Lu, Jie Zhou
Abstract: With the widespread application of drones in recent years, object detection of aerial images has attracted increasing attention, especially open-vocabulary aerial detection which is not restricted to predefined categories. Due to the scarcity of drone's viewpoint images and their significant differences from natural images, it is difficult to achieve satisfying results by directly applying vanilla open-vocabulary detection methods designed for natural scenarios. Some studies propose to transfer knowledge from pre-trained models by using lightweight networks or generating pseudo labels, but they tend to rely on models trained on natural images, neglecting the potential of foundation models specifically tailored for remote sensing and aerial imagery. To address this limitation, we propose DisDop, a unified framework that systematically distills multi-level domain priors from remote sensing foundation models (e.g., RemoteCLIP and DINOv3) into a lightweight detector. Specifically, we first distill visual priors through a teacher fusion strategy that combines RemoteCLIP's cross-modal alignment capability with DINOv3's fine-grained local feature extraction ability, transferring their complementary strengths to the detector's backbone. Second, we distill textual priors embedded in RemoteCLIP's text encoder by explicitly modeling inter-category semantic relationships, while incorporating global contextual priors to enhance local feature representation for small objects. Through this multi-level prior distillation framework, our DisDop achieves new state-of-the-art performance on open-vocabulary aerial detection benchmarks. Extensive ablation analysis also demonstrates the rationality and effectiveness of our proposed modules.
Authors: Sai Sandeep Damera, Ryan Matheu, Aniruddh G. Puranic, John S. Baras, Calin Belta
Abstract: Recurrent Neural Networks (RNNs) can learn to predict Signal Temporal Logic (STL) verdicts online from partial trajectories, but deploying them as runtime monitors in safety-critical systems demands more than predictive accuracy. Standard RNN architectures offer no structural guarantee that outputs degrade gracefully under sensor degradation; a dropped input can silently flip a verdict from safe to unsafe. We introduce the Recurrent Differentiable Ternary Logic Gate Network (R-DTLGN), a recurrent architecture that operates over Kleene's three-valued logic $\{-1, 0, +1\}$, where $0$ explicitly represents unknown. The R-DTLGN trains through continuous polynomial surrogates and hardens to a discrete ternary logic circuit at inference. We analyze the hardened circuit through two gate vocabularies derived from two orderings on the ternary domain: numerically monotone gates ensure stable recurrent dynamics, while information-monotone gates, when present, guarantee principled abstention (unknown inputs never produce wrong outputs) and monotonicity in input certainty (more information can only improve the verdict). We show that the recurrent connections required by bounded STL operators use exclusively AND and OR, which belong to both vocabularies, linking the monitoring task to the architecture's guarantees. A realizability bound derived from the STL formula's temporal operators directly sizes the network's hidden state, replacing hyperparameter search with a formula-driven specification. We evaluate on STL specifications over D4RL PointMaze navigation data, testing prediction accuracy, degradation under predicate dropout, and the accuracy-versus-safety tradeoff between two label construction pipelines. The R-DTLGN is, to our knowledge, the first recurrent architecture that couples learned temporal prediction with formal degradation guarantees rooted in three-valued logic.
Authors: Vyzantinos Repantis, Ameya Gawde, Harshvardhan Singh, Joey Blackwell II
Abstract: Before an LLM agent can use a tool, a retrieval system must decide which candidate tools to show to the agent. How long should that shortlist be? Show too many tools and the model struggles to choose. Show too few and the correct tool may not appear. Most systems apply a fixed shortlist size to every query, but no standard metric exists to evaluate whether that size was appropriate. We treat the number of tools shown to an LLM agent as the object of evaluation and we apply Bits-over-Random (BoR), a chance-corrected metric that asks whether success at a given depth is better than what random selection would achieve at that same depth. We evaluate BoR across three tool-selection benchmarks, multiple scorers, and registries ranging from 20 to 3,251 tools. We then turn the same principle into a reinforcement learning (RL) reward for choosing tool shortlist depth per query. The RL agent is deliberately simple, serving as a probe of the metric rather than a proposed system. As the shortlist grows, random chance of including the correct tool rises, so the reward naturally decreases, reducing the need for an engineered depth penalty. On BFCL (370 tools), the learned policy nearly matches the coverage of showing 50 tools ($90.3\%$ vs $90.8\%$) while presenting only 7 on average. On ToolBench (3,251 tools), a fixed shortlist of 5 tools achieves higher aggregate coverage ($64.7\%$ vs $61.9\%$) but finds nothing on hard queries (correct tool ranked 6th-20th). The BoR agent finds $16.7\%$ on those same queries by searching deeper. Downstream validation with Claude Sonnet 4.6 indicates that shorter adaptive lists also improve the LLM's ability to select the right tool: $93.1\%$ versus $87.1\%$ when always shown 5 tools, widening to $76.8\%$ vs $60.9\%$ on medium-difficulty queries where the correct tool is present but not ranked first.
Authors: Yan Lin Aung, Kevin Togbe
Abstract: This paper presents CyBOKClaw, an interpretable human-in-the-loop retrieval framework for mapping cybersecurity keywords or phrases (KWoPs) to the Cyber Security Body of Knowledge (CyBOK). Rather than treating the task as strict exact classification, the framework is designed as a top-k candidate generator for expert review. It combines query normalization, curated term expansion, concept-level boosts, topic-description enrichment, and domain-sensitive ranking rules. Because educational KWoPs are often broad, ambiguous, and only approximately aligned with CyBOK terminology, strict exact matching provides only a partial account of practical utility. We therefore evaluate the framework using both structural retrieval metrics and an expert-guided top-5 usefulness metric, ECA-5 (Exact or Closest Acceptable Match at top-5), which records whether the returned candidates contain at least one mapping that an expert would judge exact or accept as the nearest practical CyBOK placement. On the development dataset, CyBOKClaw achieves 64.73% EXA-5 (Exact Match at top-5), 84.18% structural semantic alignment, and 91.88% ECA-5; on the validation dataset, it achieves 81.19% EXA-5, 93.32% structural semantic alignment, and 98.00% ECA-5. These results show that expert-guided top-k usefulness provides a more faithful account of practical CyBOK mapping utility than exact structural matching alone, and that CyBOKClaw is effective as a CyBOK-specific expert-support retrieval system.
Authors: Bo Li, Ronghao Chen, Ningyuan Deng, Huacan Wang, Shaolin Zhu, Lijie Wen
Abstract: Translating text embedded in Web images is crucial for improving content accessibility and cross-lingual information retrieval, particularly within social media and e-commerce domains. Although Large Vision-Language Models (LVLMs) have advanced multimodal understanding, applying them to Web image translation remains challenging due to the visual representation gap: standard encoders often prioritize high-level semantics over the fine-grained visual details required for recognizing diverse character morphologies. To address this challenge, we propose VaaWIT, an end-to-end framework that adapts Large Language Models for multilingual Web image translation. The framework introduces two key technical contributions: (1) a Dual-Stream Attention Module (DSAM), which facilitates bidirectional interaction between multilingual semantic features and detailed visual representations, thereby synthesizing unified features robust to textual variations; and (2) a Visual-Aware Adapter (VAA), a parameter-efficient fine-tuning strategy that dynamically injects these fused visual cues into the frozen LLM backbone. This design enables the model to align the visual context with linguistic reasoning effectively while minimizing computational costs. Extensive experiments on eight tasks on three public benchmarks demonstrate that VaaWIT significantly outperforms state-of-the-art (SOTA) open-source baselines and achieves competitive performance against proprietary models. These results validate the efficacy of integrating fine-grained visual perception into LLMs for complex Web content analysis.
Authors: Bo Li, Tianyu Dong, Shaolin Zhu, Deyi Xiong
Abstract: Large Language Models (LLMs) have shown great promise in multilingual machine translation (MT), even with limited bilingual supervision. However, fine-tuning LLMs with parallel corpora presents major challenges, namely parameter interference. To address these issues, we propose Mix-MoE, a mixed Mixture-of-Experts framework designed to train LLMs for multilingual MT. Our framework operates in two distinct stages: (1) post-pretraining with MoE on monolingual corpora, and (2) post-pretraining with MoE on parallel corpora. Crucially, we divide the MoE layers into two specialized groups: Language Model Experts (LM Experts) and Machine Translation Experts (MT Experts). LM Experts are designed to capture and retain the monolingual knowledge learned by the pre-trained LLM. MT Experts, on the other hand, are specifically trained to acquire and store bilingual translation knowledge. Furthermore, to facilitate effective interaction between these specialized experts and leverage potential underlying structural patterns in text, we introduce a routing mechanism enhanced by Fourier Transform features derived from model representations. The experimental results demonstrate that Mix-MoE excels in multilingual MT, significantly outperforming existing baselines and showing notable progress in mitigating parameter interference.
Authors: Hao Yan, Xuanru Wang, Jun Yin, Shirui Pan, Senzhang Wang, Chengqi Zhang
Abstract: Multimodal Attributed Graph Learning (MAGL) integrates intrinsic node attributes with structural topology via graph aggregation. However, as pretrained encoders evolve into Large Foundation Models (LFMs), the landscape of MAGL fundamentally shifts: under high-confidence LFM priors, mandatory aggregation introduces topological noise that overwhelms discriminative signals, triggering a counter-intuitive performance inversion where sophisticated MAGL architectures underperform simple topology-agnostic MLPs. Through systematic empirical and theoretical analysis, we identify that this inversion stems from a fundamental aggregation dilemma characterized by two concurrent pathologies: (1) Representational Pathology (SNR Degradation) - mandatory aggregation dilutes robust intrinsic features with topological noise, causing the noise penalty to outweigh its collaborative benefit; and (2) Optimization Pathology (Gradient Starvation) - topological aggregation attenuates gradient flow, while a shared task loss causes dominant modalities to prematurely suppress weaker ones. To resolve this dilemma, we propose SUPRA (Shared-Unique Prior-Retaining Architecture), a decoupled dual-pathway paradigm. SUPRA processes modality-specific features through topology-agnostic MLPs while capturing structural synergy via a lightweight shared GNN, with auxiliary deep supervision counteracting gradient starvation. Extensive evaluations demonstrate that SUPRA achieves state-of-the-art performance while requiring 3.5x lower peak GPU memory and up to 4.4x faster training time than Multimodal Graph Transformers.
Authors: Ruyi Chen, Lu Zhou, Xiaogang Xu, Chiyu Zhang, Jiafei Wu, Liming Fang
Abstract: Text-to-Image (T2I) models have made significant strides in visual realism and semantic consistency, yet they often perpetuate and amplify societal biases. Existing evaluation methods typically address only single-dimensional biases, lacking perspectives to uncover model biases at social-related deeper semantic levels. We introduce HoloFair, a comprehensive benchmark framework for multidimensional demographic bias analysis. Built upon our large-scale fairness-oriented dataset and the SpaFreq (Spatial-Frequency) attribute classifier, this framework proposes the Multi-attribute, Group-wise Bias Index (MGBI) metric, designed to assess both intrinsic diversity and conditional biases. Beyond evaluation, we further introduce Fair-GRPO, a reinforcement-learning-based debiasing method that alters the distribution of generative models through a designed multi-objective reward function. E.g., experiments on the SD3.5-Medium model demonstrate that Fair-GRPO significantly improves multidimensional fairness while maintaining high image quality. We also analyze potential reward hacking phenomena and provide corresponding mitigation strategies. Code and dataset are available at https://github.com/1059684669/HoloFair
Authors: Bohang Sun, Max Zhu, Francesco Caso, Jindong Gu, Junchi Yu, Philip Torr, Pietro Li\`o, Jialin Yu
Abstract: Diffusion large language models promise faster generation by refining many token positions in parallel, but this parallelism introduces a hidden control problem: which proposed tokens should be transferred into the partially decoded sequence at each step? We refer to this decision as token commitment. Existing frozen-generator decoders largely rely on hand-designed confidence rules or block-specific acceptance filters. We argue that token commitment can instead be learned as a reusable trace-state policy. We introduce TraceLock, a lightweight plug-in controller that instantiates this policy for a frozen diffusion language model. Since oracle commitment times are unavailable, TraceLock derives self-supervision from future stability: at decoding step t, a proposed token for position i is labeled stable if it matches the final token at position i after the full decoding trace completes. The controller scores variable-length trace states and decides which active token proposals should be committed to the partially decoded sequence. Once trained for a given frozen backbone, the controller can be deployed across local-window widths, generation lengths, and step budgets without retraining or per-setting calibration. Experiments on question answering, mathematical reasoning, and code generation show that TraceLock improves the quality-step tradeoff over heuristic and learned baselines, with particularly stable behavior under cross-setting deployment. Diagnostic analyses show that its decisions are not reducible to scalar confidence, suggesting that frozen diffusion language models expose a learnable space of commitment trajectories beyond confidence-based decoding. Code is available at https://github.com/BobSun98/TraceLock.
Authors: Liying Han, Kang Yang, Oliver Wang, Jason Wu, Pengrui Quan, Gaofeng Dong, Ozan Baris Mulayim, Sizhe Ma, Yuyang Yuan, Dezhi Hong, Mario Berges, Mani Srivastava
Abstract: Large language models (LLMs) and time-series language models (TSLMs) are increasingly applied to time-series question answering (TSQA). Unlike text-only QA, TSQA requires models to ground answers in temporal signals whose patterns may occur at different scales, specific time locations, or across separated intervals. However, existing benchmarks are typically organized by task types or high-level reasoning categories, making it difficult to diagnose the underlying signal-level capabilities driving model performance. We introduce TS-Skill, a controlled benchmark for evaluating three composable analytical skills in TSQA: temporal scale selection (SK1), temporal localization (SK2), and cross-interval integration (SK3). TS-Skill provides timestamp-aware questions, broad domain coverage, and human-validated QA quality. To construct the benchmark at scale, we develop SKEvol, a skill-guided agentic framework that combines domain-aware time-series seed generation, skill-controlled question generation, metadata- and code-assisted answer construction, multi-phase signal-grounded verification, and human-in-the-loop curation. Experiments on ten state-of-the-art LLMs and TSLMs reveal substantial and uneven capability gaps across SK1-SK3. In particular, SK3 remains consistently challenging for non-agent models, whereas tool-augmented agents show a selective advantage on standalone SK3. These findings demonstrate that skill-level evaluation can uncover temporal reasoning failures that are obscured by aggregate TSQA scores.
Authors: Santiago G\'ongora, Luis Chiruzzo, Gonzalo M\'endez, Pablo Gerv\'as
Abstract: Large Language Models (LLMs) have changed the possibilities of Interactive Storytelling systems that process free-text user input. However, as more of these systems are built, evidence continues to mount regarding the story coherence problems that arise when relying solely on them. Recent research suggests that LLMs can effectively predict state changes within rule-based Interactive Storytelling systems, triggering pre-programmed world-state transformations. In this paper, we conduct an exploratory evaluation of whether such transformations can serve as a catalyst for player expression while aiming to address the incoherence issues typical of purely LLM-based approaches. Building upon a neuro-symbolic architecture, we conducted experiments using an open-source model (Llama 3 70B) and a closed-source model (Gemini 1.5 Flash), with testing conducted in both English and Spanish. Eight participants played two scenarios, carefully designed to assess different evaluation objectives. Our observations suggest that transformations offer a way to maintain world-state consistency while encouraging players to interact creatively through their written inputs.
Authors: Jehanne Dussert
Abstract: Current approaches to AI compliance treat conformity as a binary, audit-time verdict rather than a continuous, measurable property of production systems. We argue that this compliance fiction is structurally ill-suited to the requirements of the EU AI Act, which demands ongoing human oversight and the detection of emergent behavioural drift in deployed systems. We introduce governance from metrics, a principle whereby regulatory compliance is derived as a continuous signal from runtime observability rather than from static assessments. Building on this principle, we present govllm, an open-source framework implementing a governance-driven routing architecture in which model selection is determined by accumulated compliance scores rather than by latency or cost alone. Central to our approach is a panel of regulatory judges - LLM evaluators specialised per criterion (EU AI Act, GDPR, ANSSI, accessibility) - whose inter-judge disagreement we reframe not as noise but as a regulatory uncertainty signal warranting human arbitration. We validate this approach through a ground truth corpus of 49 annotated prompt/response pairs across five regulatory criteria, evaluated by four small language models (SLMs, 1.7B-7B parameters) running fully on-premise. Agreement rates range from 51.5% (mistral:7b) to 69.1% (phi4-mini), with no single model dominating across all criteria - empirically motivating the Profile-as-jury design. We further document three structural failure modes in small regulatory judges and a judge-specific position bias that degrades agreement by up to 25 percentage points across three question-order conditions (original, reversed, permuted). govllm is released as open-source software to support reproducible AI governance research.
Authors: Shresth Verma, Mauricio Tec, Cheol Woo Kim, Kai Wang, Milind Tambe
Abstract: While LLMs excel at single-turn generation, they struggle with long-horizon, multi-turn interactions. Offline reinforcement learning (RL) offers a scalable approach, yet its performance hinges on the availability and quality of multi-turn trajectory data. A common remedy is to augment training with synthetic trajectories generated by LLMs or simulators, but synthetic data is highly heterogeneous in quality, and naively treating all trajectories as equally informative can degrade performance. We propose BOOST, a bilevel optimization framework where the inner level trains the LLM on reweighted data and the outer level trains a lightweight reweighting head on held-out real validation tasks, assigning continuous trajectory-level weights without requiring an external judge. To ground this approach, we derive a PAC-Bayesian bound revealing a three-way trade-off: synthetic data increases diversity but risks task-shift, while concentrating weight on high-quality trajectories improves empirical performance at the cost of effective sample size. Empirically, our method consistently outperforms multiple baselines. Analysis reveals it upweights synthetic trajectories that align with the real data distribution and exhibit higher qualitative merit.
Authors: Ismail Lamaakal
Abstract: Neural network weights are increasingly a bottleneck for deployment, yet most compression pipelines treat layers independently and overlook cross-layer redundancy induced by function-preserving symmetries. We propose Motion-Compensated Weight Compression (MCWC), a weight-only codec that aligns permutation-symmetric blocks (e.g., hidden units and attention heads) to maximize cross-layer correspondence, turning depth into a predictable sequence. In the aligned coordinate system, MCWC uses a lightweight layer-sequential predictor with periodic keyframes and encodes only quantized prediction residuals using a learned entropy model trained under a rate distortion objective. A simple decoder reconstructs deployable weights by entropy decoding, dequantization, predictor-driven reconstruction, and inverse alignment, enabling fast weight materialization for inference. Across Transformer language modeling and vision classification, MCWC improves the rate accuracy Pareto frontier over strong quantization and learned weight-codec baselines, while maintaining competitive decode time. Ablations confirm that alignment, prediction, entropy modeling, and keyframe scheduling are each necessary for the full gains. Our code is available via https://github.com/Ism-ail11/MCWC.
Authors: Andrea Morandi
Abstract: [Abridged] - Spectral Retrieval is a plug-in re-ranking stage that interpolates between per-token MaxSim and mean-pool retrieval through a multi-scale sinc convolution over token embeddings. In standard dense retrieval each document is one mean-pooled vector; when relevance localises into a short subspan, the signal averages into noise. Spectral Retrieval reuses per-token embeddings from a late-interaction index and convolves them with a normalised sinc kernel at multiple scales. At L=1 the kernel acts as the identity, recovering per-token MaxSim; as L grows it approaches a uniform filter, recovering mean pooling. The maximum cosine over positions and scales yields a score provably no less informative than either endpoint. On a controlled synthetic benchmark with 1,000 documents and planted single-position spikes, mean-pool retrieval sits at chance (Recall@10 ~ 0.02) regardless of spike strength, while Spectral Retrieval reaches Recall@10 = 1.0 once the planted cosine exceeds the corpus-level token noise floor. On LIMIT-small with a frozen all-mpnet-base-v2 encoder, Spectral Retrieval lifts Recall@10 from 0.33 to 0.90, MRR from 0.22 to 0.79, and strict Success@10 from 0.12 to 0.84, without retraining. The method fits naturally into multi-agent LLM systems, where each agent benefits from a tighter, role-specific retrieval window over a shared corpus.
Authors: Daniele Picone, Mohamad Jouni, Mauro Dalla-Mura
Abstract: Hyperspectral image restoration faces several challenges, including limited training data, strong sensor specificity, and high spectral dimensionality. These limitations hinder the learning of robust hyperspectral priors, motivating the reuse of priors learned from large-scale RGB data. In this work, we propose a minimally trained, lightweight adapter that repurposes frozen pretrained RGB denoisers for hyperspectral restoration through a projection mapping. The method denoises low-dimensional spectral projections and reconstructs the hyperspectral cube through constrained linear aggregation, while preserving plug-and-play compatibility and the stability properties of the underlying RGB denoiser. Experiments on denoising, deblurring, and super-resolution across multiple datasets demonstrate consistent improvements over hyperspectral-specific baselines, showing the strong transferability of large-scale RGB priors.
Authors: Bruce Changlong Xu, Jose James, Alexander Ryu
Abstract: Classical noisy-label theory predicts that downstream performance under weak supervision is bounded above by the labeler's accuracy, implying a sharp crossover: once a gold-trained classifier matches the labeler, weak labels stop helping and start hurting. The prediction is theoretical; what is missing is a benchmark calibration that turns it into an instance-level statement for modern foundation-model labelers. We provide such a calibration for BiomedCLIP-generated weak labels on three medical-imaging benchmarks (PCAM, ISIC, NIH-CXR) and six downstream architectures spanning an 11x parameter range. The crossover predicted by theory appears at ng~100 on PCAM, 20-50 on ISIC, and 250-500 on NIH-CXR; weak labels above the crossover degrade AUC by up to -0.10. The location is architecture-invariant for four of five pretrained architectures, and a within-family DenseNet sweep (2.5x parameters, identical pretraining) supports the view that the labeler, not the student, is the dominant constraint. The calibration in turn produces a decision rule operable from 10-20 gold labels: compare gold-only AUC to VLM accuracy on the user's gold set. A structured-vs-random noise sign flip on NIH-CXR shows that the rate-only formulation of the bound is incomplete and identifies a concrete refinement (label-space projection) that future benchmarks can be designed to test.
Authors: Rishabh Iyer
Abstract: Submodular optimization has become a fundamental paradigm for data selection, retrieval, summarization, and representation learning due to its ability to model coverage, diversity, and representativeness. However, classical submodular objectives optimize only the selected subset and do not explicitly preserve structural information between the selected subset and the remaining data. In many modern machine learning applications, including train/validation/test splitting, benchmark construction, and robust subset selection, the quality of a selection depends critically on preserving balanced structure across both the selected subset and its complement. In this work, we introduce Complement Submodular Information (CSI), a new class of complement-aware submodular objectives that quantify shared structural information between a subset and its complement. Our framework induces complement-aware variants of several classical submodular functions including Facility Location, Graph Cut, LogDet, Saturated Coverage, Set Cover, Probabilistic Set Cover, and Feature Based Functions. We analyze the theoretical properties of CSI objectives and show that they exhibit approximate monotonicity under bounded curvature conditions, leading to near-$(1-1/e)$ greedy approximation guarantees. Empirically, CSI objectives consistently outperform standard submodular objectives on robust hidden-slice-aware subset selection. In particular, CSI objectives significantly improve preservation of coherent rare/tail semantic structure while simultaneously suppressing noisy and isolated outliers, leading to substantially improved downstream predictive performance. Synthetic experiments further illustrate how different CSI instantiations capture complementary notions of representativeness, diversity, connectivity, and balanced neighborhood preservation.
Authors: Yubo Li, Yidi Miao
Abstract: Long-horizon LLM inference turns the key--value (KV) cache into the dominant GPU memory consumer and makes per-token attention increasingly expensive. Many common eviction policies use static recency windows or historical attention, leaving unused a signal computed on every decoding step: the model's current uncertainty. We introduce CONF-KV, a KV-cache manager that converts the next-token distribution into a scalar confidence score and uses it to choose the per-step cache budget, retaining more context when the model is uncertain and pruning aggressively when it is confident. Within each budget, tokens are ranked by a composite of accumulated attention mass and recency, while a protected recent window preserves local coherence. We combine the policy with blockwise online-softmax attention, mixed FP16/INT8 storage, and a pyramidal per-layer budget variant. Across four model families and generated lengths up to 4K, CONF-KV stays near the footprint of a fixed 512-token sliding window while remaining within 1.5--2.1 perplexity points of full KV. On Needle-in-a-Haystack up to 32K tokens, CONF-KV reaches 91.4% retrieval accuracy versus 53.8% for sliding windows and 80.6% for H2O; on 75 VisualWebArena tasks it retains 95.3% of full-KV success at 2.8 times lower peak memory.
Authors: Ojonugwa Oluwafemi Ejiga Peter, Frederick Akor Ejiga, Fahmi Khalifa, Md Mahmudur Rahman
Abstract: The major limitations of gastrointestinal (GI) endoscopy AI systems arise from a shortage of annotated data, strict privacy policies, and significant bottlenecks in conventional model fine-tuning. Such limitations impede the successful application of sophisticated AI models in clinical practice, particularly affecting the reliability and scalability of diagnosis. In this paper, we present a dual-pipeline PEFT model that addresses two fundamental problems: medical Visual Question Answering (VQA) and the generation of privacy-preserving synthetic data. For clinical VQA, we adopt the Florence-2 vision-language model. Leveraging PEFT enhances model interpretability while substantially reducing the computational cost of training. Simultaneously, we employ Low-Rank Adaptation (LoRA) with Stable Diffusion 2.1 to generate high-quality GI images that enhance training databases without violating patient privacy. This research utilized the Kvasir-VQA dataset. Our Florence-2 VQA model achieved ROUGE-1 of 0.92, ROUGE-L of 0.91, and BLEU score improvements from 0.08 to 0.24. Fine-tuning on private datasets consistently showed better results than fine-tuning on public datasets. The rank-4 LoRA synthesis achieved optimal performance with a fidelity score of 0.290, an agreement score of 0.730, and a Frechet BiomedCLIP Distance (FBD) of 1450, reducing computational costs by almost 90 percent. This framework improves the clinical potential of AI in GI endoscopy. Compared to FLUX, MSDM, and Kandinsky 2.2, our model demonstrates superior FBD and strong semantic alignment. While other models lead in Fidelity or Agreement, our lower FBD indicates better image-text coherence. These results establish our approach as a robust solution for enhancing VQA and synthetic data generation in clinical AI.
Authors: Zhipeng Ye, Jiaqi Huang, Feng Jiang, Qiufeng Wang, Yikang Duan, Dawei Wang, Xihang Zhou, Qian Qiao
Abstract: Multimodal Large Language Models (MLLMs) have demonstrated strong capabilities across a wide range of vision language tasks. However, when applied to large scale image classification, their performance degrades significantly as the label space expands a phenomenon we define as Performance Collapse in Long Sequence Recognition. Through an information theoretic analysis, we reveal that this collapse stems from a fundamental conflict between the escalating information entropy and the prominent attention dilution and decay within attention mechanisms, which impairs the model's ability to maintain a sufficient signal-to-noise ratio when processing extremely long prompts. To mitigate this, we propose Divide-and-Conquer Inference (DCI), a novel test-time scaling strategy for visual recognition with MLLMs. DCI recursively decomposes complex global classification tasks into multiple simpler, localized subproblems and employs a dynamic pruning mechanism to compress the search space. This method effectively improves the local signal to noise ratio and model accuracy by mitigating the inherent weight dilution issues in long-sequence inference. Moreover, while traditional self-attention incurs a prohibitive quadratic computational complexity, DCI achieves more favorable scaling behavior and substantially accelerates inference in large scale classification scenarios. Extensive experiments on benchmarks such as ImageNet-1K and ImageNet-21K demonstrate that DCI consistently improves classification accuracy. This enables lightweight open-source models to rival or even surpass frontier closed-source giants without any additional training or fine-tuning. As a model-agnostic, plug-and-play paradigm, DCI offers an efficient approach for scaling the inferential precision of MLLMs in large-scale scenarios.
Authors: Muhammad Ashad Kabir, Sirajam Munira
Abstract: Large audio and language models have recently demonstrated zero-shot reasoning capabilities across various domains. However, it remains unclear how the form of audio input, whether handcrafted acoustic features extracted from speech or the raw audio waveform itself, affects performance for Parkinson's disease (PD) detection across different languages. In this study, we systematically compare two input modalities for zero-shot PD detection: (i) handcrafted acoustic features extracted from speech recordings analyzed by a general-purpose LLM, and (ii) direct waveform input analyzed by audio-capable models. Experiments on PD speech datasets in four languages show that performance varies across input modalities, speech tasks, and languages. Handcrafted acoustic features provide more stable performance in a low-resource language (e.g., Bengali), whereas audio input yields dataset-dependent gains. These findings highlight the impact of input modality on zero-shot PD detection from speech.
Authors: Guodu Xiang, Kui Yu, Yujie Wang, Richang Hong, Fuyuan Cao, Jiye Liang
Abstract: Confounding bias is a key challenge in causal effect estimation from observational data. Double Machine Learning (DML) addresses this issue by estimating treatment and outcome nuisance functions, constructing treatment and outcome residuals, and estimating causal effects from the residuals. However, DML often produces biased and unstable estimates in highdimensional or finite-sample scenarios. One reason is that DML estimates nuisance functions using all covariates without disentangling distinct latent factors, resulting in unreliable nuisance function estimation. Another is that imprecise nuisance estimation further introduces residual dependence between the treatment residual and the remaining outcome error, undermining the accuracy of causal effect estimates. To address these issues, in this paper, we propose Disentangled Double Machine Learning (DDML), a novel algorithm that integrates two key strategies. First, a causal role disentanglement strategy decomposes covariates into confounders, treatment-specific factors, and outcomespecific factors for enabling reliable nuisance function estimation. And second, a residual dependence orthogonalization strategy mitigates residual dependence caused by nuisance estimation errors for enhancing the precision of causal effect estimates. Experimental results on synthetic, semi-synthetic, and real-world datasets demonstrate that DDML significantly outperforms 13 state-of-the-art baseline algorithms in both MAE and RMSE.
Authors: Yu Yang, Yihong Guo, Anqi Liu, Pan Xu
Abstract: Off-dynamics offline reinforcement learning seeks to learn a target-domain policy from a large source dataset and a limited target dataset under mismatched transition dynamics. Existing approaches such as reward augmentation and data filtering are constrained to the source dataset and cannot synthesize new target behavior to improve coverage beyond the collected source trajectories. While recent model-based methods attempt to address this by learning target-aware dynamics, the generated experience is constructed only at the transition level, which leads to accumulated errors over long horizons. These limitations necessitate a shift toward trajectory-level generation for off-dynamics offline RL. We propose CEDGE, a Cross-domain Energy-guided Diffusion GEneration framework. CEDGE trains a trajectory diffusion model on source-domain trajectories and adapts the generated samples to the target domain through energy guidance. This guidance is derived by minimizing the distribution mismatch between the source and desired target-domain trajectories and is decomposed into return, domain, and behavior energy components. The resulting energy-guided trajectories are useful both for direct planning and as synthetic data for policy learning. Since target adaptation is achieved via energy guidance rather than retraining the diffusion model, CEDGE can be efficiently adapted to new target dynamics compared to previous methods. Experiments on the ODRL benchmark demonstrate that trajectory-level energy-guided generation improves diffusion planning under dynamics shifts and produces synthetic data that improves downstream target policy learning.
Authors: Chidera G. Oguine, Kanyifeechukwu J. Oguine, Obiozor M. Oguine, Ozioma C. Oguine
Abstract: Non-Maximum Suppression (NMS) remains a key post-processing step in many real-time object detection pipelines, but it can introduce latency variation and deployment complexity in resource-constrained settings. Recent NMS-free designs such as YOLO26 aim to reduce this dependence through end-to-end detection, yet their performance relative to established NMS-based models such as YOLOv8 remains underexplored beyond standard benchmarks. This paper compares YOLOv8 and YOLO26 on Pascal VOC and VisDrone, representing general object detection and dense aerial small-object detection, respectively. Both model families are evaluated across five scales using accuracy, localization, model size, GFLOPs, and CPU/GPU latency. Results show that YOLO26 achieves stronger detection performance and lower model complexity on Pascal VOC across most scales, while the performance gap narrows on VisDrone, where both models struggle with dense small targets. YOLOv8 remains competitive in GPU latency, showing that NMS-free design does not guarantee universal deployment superiority. Overall, the study shows that detector selection depends on dataset characteristics, object scale, model capacity, and hardware constraints.
Authors: Lixing Lin, Juli You, Yue Li, Luyun Lin, Yiqing Wang, Zhen Zhang, Moxuan Zheng
Abstract: Large language model (LLM) safety classifiers such as Llama Guard are effective at detecting overtly harmful prompts but remain vulnerable to adversarial jailbreak attacks that disguise malicious intent through role-play scenarios, fictional framing, and indirect requests. We present Reflect-Guard, a method that augments LLM-based safety classifiers with chain-of-thought self-reflection capabilities through parameter-efficient fine-tuning. Our approach distills analytical reasoning from GPT-4o-mini into structured reflection annotations, then trains Llama-Guard-3-8B via QLoRA to generate logical self-reflections before issuing safety verdicts. Using only 1000 training examples and updating just 0.5% of model parameters (~42M), Reflect-Guard achieves substantial improvements on two challenging benchmarks. On WildGuardTest, F1 score improves from 0.770 to 0.842 (+7.2 pp), with recall on adversarial prompts increasing from 0.513 to 0.921 (+40.8 pp). On JailbreakBench, the attack success rate drops from 10.3% to 1.8%, representing an 82.5% relative reduction. These gains are especially pronounced on adversarial inputs, where the explicit reasoning step enables the model to see through obfuscation techniques that defeat standard pattern-matching approaches. Our results demonstrate that teaching safety classifiers to reason about adversarial intent, rather than simply classify surface patterns, is a promising direction for robust LLM safety.
Authors: Ligong Bi, Tao Huang, Jianyuan Guo, Chang Xu
Abstract: Visual Autoregressive (VAR) models have emerged as a powerful paradigm for image synthesis by performing hierarchical next-scale prediction. However, VAR models are inherently prone to cascading error propagation, where subtle coarse-scale mispredictions are amplified across the hierarchy, ultimately distorting the final synthesis. To mitigate this, we propose AID-VAR, a plug-and-play framework that enhances pre-trained VARs through Adversarially Injected Diagnosis. Instead of a standard passive generation, AID-VAR introduces a proactive error-correction mechanism inspired by the adversarial feedback in GANs. We deploy a discriminator to diagnose fidelity gaps at each scale transition, coupled with a lightweight guidance injector. This module operates as a non-invasive adapter that refines the feature manifold of a frozen VAR backbone, effectively steering the generation toward the distribution of real images without destabilizing the pre-trained latent space. Furthermore, to rigorously evaluate this cross-scale progression, we introduce the Inter-Scale Consistency Score (ISCS), a novel metric that quantifies the fidelity and structural alignment between consecutive resolution scales. Experimental results across various backbones demonstrate that AID-VAR delivers sharper textural details and fewer structural distortions with negligible overhead. For instance, AID-VAR-d20 achieves a 16% improvement in FID with only a 3% increase in parameters. These results establish AID-VAR as a highly efficient and scalable pathway for upgrading large-scale VAR generators, enhancing global coherence and local detail without altering training data, base architectures, or sampling schedules. Code is available at https://github.com/bijiw515/AID-VAR.
Authors: Xiangtian Ji, Yuxin Chen, Zhengzhou Cai, Xiang Wang, An Zhang, Tat-Seng Chua
Abstract: Large language models (LLMs) display strong comprehensive abilities, yet the internal mechanisms that support these behaviors remain insufficiently understood. In this work, we show that across a wide range of open-weight Transformers, a subset of neurons remains consistently highly activated during inference across tasks of multiple capability dimensions. By probing along the cross-task activation strength, an extremely sparse subset is isolated, whose removal causes a collapse in model behavior, which we term keystone neurons. Our analysis reveals that keystone neurons are a stable and intrinsic neuron subset of the model that is largely established during pretraining. The parameters associated with these neurons are tightly calibrated during the training process, and their precise values are critical for the capabilities of the model. Building on these insights, we propose a supervised fine-tuning approach that updates only keystone neurons, achieving task gains comparable to or even better than full-parameter fine-tuning while better preserving performance in other capability dimensions, despite modifying a much smaller number of parameters.
Authors: James Henry
Abstract: Concept formation in transformer language models is depth-extended, not a single-layer event: concepts emerge gradually across a contiguous region of the residual stream. Mechanistic interpretability methods identify the single layer of peak class separation -- the "best layer" -- capturing a snapshot rather than the process itself. We introduce the Concept Allocation Zone (CAZ): the depth interval within which a concept becomes measurably separable, the region allocated to its geometric expression. We formalize the CAZ through three layer-wise metrics (Separation, Concept Coherence, Concept Velocity) and derive principled boundary detection without manual layer sweeps. A CAZ is not a concept: it is the depth region within which the model organizes its geometry to make a concept separable. A single concept typically participates in multiple CAZes; multiple concepts may share one. Empirical validation across 34 models from 8 architectural families and 7 concepts reveals that the separation curve S(l) is frequently multimodal. A scored detector uncovers "gentle CAZes" -- subtle allocation regions invisible to standard peak detection but causally active in 93-100% of cases under ablation (16 of 34 models; 26 in the companion validation paper). The framework generates seven testable predictions; four yield clear verdicts (two not supported, one partially supported, one supported), one had its precondition invalidated by the data, and two are underpowered -- with cross-architecture alignment confirmed as depth-matched rather than monolithic under leave-one-concept-out cross-validation. Reference implementation: rosetta_tools v1.3.1 (doi:10.5281/zenodo.20361433).
Authors: Tianyi Wang, Tianyi Zeng, Zimo Zeng, Feiyang Zhang, Yujin Wang, Xiangyu Li, Yiming Xu, Sikai Chen, Junfeng Jiao, Christian Claudel, Xinbo Chen
Abstract: Advanced driver assistance systems (ADAS) play an important role in modern automotive intelligence, significantly enhancing vehicle safety and stability. The performance of ADAS critically relies on accurate and reliable vehicle state estimation, particularly from vehicle dynamic sensors. Among these signals, wheel load is a key variable for chassis control and safety-critical functions, yet it remains difficult to estimate robustly due to complex suspension geometry, nonlinear dynamics, and measurement noise. To address this issue, we propose DBPnet, a Bayesian physics-informed neural network (PINN) with a physics-aware embedding module inspired by damper characteristics. First, this paper presents a suspension linkage-level modeling (SLLM) approach that constructs a nonlinear instantaneous dynamic model by explicitly considering the complex geometric structure of the suspension. Building upon SLLM, Bayesian inference is integrated into the PINN to effectively cope with noise and uncertainty in the vehicle chassis system, thereby improving the model's robustness. Then, a physics-informed loss function is employed to ensure consistency with fundamental physical principles, while the damper characteristics-inspired embedding module extracts temporal variation features of input signals and incorporates them into each layer of the PINN, ensuring that physical observations guide the neural network without being constrained by fixed physical models. Extensive evaluations on high-fidelity simulations and real-world experiments demonstrate that our DBPnet consistently achieves lower RMSE and MaxError than baseline methods. These results highlight the potential of our DBPnet to advance wheel load estimation and contribute to the development of more reliable ADAS actuator functions.
Authors: Qirun Dai, Xiao Liu, Jiawei Zhang, Dylan Zhang, Hao Peng, Chenhao Tan
Abstract: Despite the importance of causal reasoning, training LLMs to reason causally remains underexplored. Existing data efforts mostly focus on benchmarking LLMs on specific aspects of causality, making them less suitable for training generalizable causal reasoners. To address this, we propose UniCo, a data generation framework that both (1) addresses 18 causal query types across Pearl's Causal Ladder and (2) translates natively symbolic examples into code and natural language forms to simulate real-world use cases where causal terms are not explicitly specified. To ensure data quality, UniCo grounds answers with exact causal inference and filters cases with reasoning shortcuts. Upon supervised finetuning with 66.6K UniCo-generated instances, Qwen3-4B, Qwen3-8B and Olmo-3-7B-Instruct achieve an average of 22.9% improvements across all 18 in-distribution query types, and 8.1% over state-of-the-art causal data generation frameworks on 7 established causal benchmarks outside the training distribution. More importantly, in real-world medical understanding, legal decision, and tabular reasoning, UniCo-trained models consistently display more faithful reasoning traces, outperforming the base models by an average of 20.2% in faithfulness metrics. These suggest that causality-centered training not only strengthens causal reasoning, but also equips LLMs with a causal mindset in general reasoning tasks.
Authors: Faizan Faisal
Abstract: Reasoning-enabled LLMs perform strongly on medical reasoning benchmarks, but it remains unclear whether these gains transfer to structured clinical documentation; we investigate this question using SOAP note generation from clinical dialogue in a source-aware benchmark spanning OMI Health, ACI-Bench, and PriMock57. We evaluate GPT-5.4, DeepSeek-V4-Flash, and Gemma-4-E4B in a controlled 2x2 design that independently toggles provider-native reasoning and same-source retrieval-augmented generation (RAG). Outputs are assessed using seven automatic metrics alongside two reference-aware LLM judges. Both evaluation approaches agree that a non-reasoning GPT-5.4 configuration achieves the highest overall quality, while DeepSeek-V4-Flash performs best among reasoning-enabled configurations. Enabling reasoning significantly degrades GPT-5.4 performance across all three datasets, whereas same-source RAG yields smaller, model-dependent improvements. Overall, the findings indicate that stronger reasoning capability should not be assumed to improve fidelity-sensitive SOAP note generation without dedicated, task-specific evaluation.
Authors: Ismail B. Mustapha, Shafaatunnur Hasan, Sunday O. Olatunji, Hatem S. Y. Nabus
Abstract: Class imbalance in deep neural networks (DNNs) has witnessed a rapid increase in research attention in recent years. However, the varying accounts of the reasons behind the poor performance of DNN on imbalance data in pertinent literature shows that little is known about how this agelong phenomenon impacts the performance of DNNs. A better understanding of this problem is crucial to developing effective DNN-based imbalance methods. Thus, this study systematically investigates the impact of class imbalance on the learning dynamics of DNN by monitoring the learning pattern of DNN models on both the majority and minority classes of datasets of varying imbalance ratios. Experimental findings shows that as against learning from balanced datasets where DNN learns the classes similarly, class imbalance has severe deteriorating impact on the performance of DNN, driving the model to underfit the minority class samples in the early training epochs while simultaneously learning only the majority class. Although DNN ultimately learns the minority samples, learning in this manner only results in learnt minority representations that are non-generalizable at test phase because they are merely overfitted to keep the overall training loss as low as possible.
Authors: Jinjin Chi, Lei Feng, Lulu Zhang, Yongcheng Jing, Yiming Wang, Ximing Li, Jialie Shen, Leszek Rutkowski, Dacheng Tao
Abstract: Time series foundation models (TSFMs) have recently achieved strong zero-shot forecasting performance through large-scale pretraining and retrieval-augmented prediction. However, our empirical analysis reveals a non-trivial limitation of retrieval-based forecasting: retrieval tends to induce more oscillatory predictions, improving performance on highly fluctuating series while degrading accuracy on smoother, trend-dominated ones. This suggests that retrieved information may be fused into prediction without explicitly distinguishing stable temporal structure from instance-specific variations, which can reduce robustness under distribution shifts. We propose a Retrieval-guided Invariant-Dynamic DEcomposition framework for time series forecasting. Rather than using retrieval as auxiliary predictive context, we leverage retrieved sequences as implicit samples from related environments to guide representation decomposition. Specifically, we first construct a retrieval-aware representation via attention-based aggregation, and then introduce a retrieval-guided routing mechanism to decompose it into an invariant component capturing stable shared structure and a dynamic component modeling context-dependent variations. These two components are forecast separately and fused for final prediction, enabling the model to preserve transferable patterns while remaining adaptive to evolving dynamics. We further design training objectives that encourage invariant learning and disentanglement, and provide theoretical insight showing that retrieval aggregation reduces variance and approximates invariant representation learning without explicit environment supervision. Extensive experiments demonstrate that our method consistently improves robustness under distribution shifts and outperforms existing TSFMs and retrieval-based baselines in zero-shot forecasting settings.
Authors: Mini Han Wang, Liting Huang, Wei Hong, Boonthawan Wingwon
Abstract: Background: Type 2 diabetes mellitus (T2DM) is increasingly recognised as a systemic disease characterised by coordinated dysfunction across metabolic, renal, lipid, and inflammatory pathways. Existing clinical assessments often fail to capture this multi-dimensional burden. Methods: We conducted a retrospective study of 1,195 patients using routinely collected laboratory biomarkers. System-level abnormality indices were constructed to quantify organ-specific dysfunction, and multi-system involvement was defined as abnormalities in two or more systems. Supervised machine learning models, including logistic regression, random forest, and gradient boosting, were trained to predict multi-system dysregulation. Model interpretability was achieved using SHapley Additive exPlanations (SHAP). Results: The gradient boosting model demonstrated near-perfect discrimination (AUC = 1.000), significantly outperforming logistic regression (AUC = 0.925). Feature attribution analysis revealed that hyperglycaemia, renal impairment, dyslipidaemia, and inflammation were the dominant drivers of multi-system risk. Dose-response relationships observed in partial dependence analyses further supported the biological plausibility of model predictions. Conclusion: This study presents an interpretable, data-driven framework for quantifying systemic disease burden in T2DM. By linking routine biomarkers to multi-organ dysfunction, our approach provides both predictive accuracy and mechanistic insight, offering potential for improved risk stratification and precision medicine in diabetes care. The data and code used in this study are openly available on GitHub at: https://github.com/MiniHanWang/Type-2-Diabetes-1.git
Authors: Mini Han Wang, Liting Huang, Wei Hong, Boonthawan Wingwon
Abstract: Retinal imaging provides a non-invasive window into systemic microvascular health and has emerged as a potential biomarker for systemic diseases. However, whether retinal features encode biologically meaningful systemic signals that can be reliably interpreted using explainable artificial intelligence (XAI) remains unclear. An explainable multi-task deep learning framework was developed to investigate associations between retinal microvascular features and systemic abnormalities in Type 2 Diabetes Mellitus. A total of 11,011 fundus images from 2,719 individuals were analysed using a shared neural network with task-specific heads for glycaemic status, kidney abnormality, and multi-system involvement. Model interpretability was evaluated using Gradient-weighted Class Activation Mapping (Grad-CAM), anatomical masking, and vessel alignment analysis. The framework demonstrated task-dependent predictive performance, with the best discrimination observed for kidney abnormality (AUC up to 0.63), whereas glycaemic status prediction showed limited performance (AUC = 0.49-0.61). Explainability analyses consistently localized model attention to retinal vessels and peripapillary regions. Masking experiments showed that occlusion of vascular regions caused the greatest performance decline, indicating that retinal vessels were the primary predictive source. Different architectures exhibited heterogeneous attention patterns, suggesting multiple representational pathways for systemic signal encoding. This pilot study demonstrates that retinal microvascular features contain measurable signals associated with systemic abnormalities, particularly microvascular damage. By integrating multi-task learning with quantitative XAI validation, this framework advances retinal imaging toward interpretable digital biomarkers for systemic risk stratification in diabetes.
Authors: Shogo Yamauchi, Tohru Nitta, Hideaki Tamori
Abstract: Quaternion neural networks are parameter-efficient and model multidimensional dependencies by representing four related features as a single entity. However, existing quaternion self-attention computes component-wise scores and applies independent softmax operations to each component, which increases the computational cost and allows attention distributions to diverge across components. We propose a shared-score quaternion self-attention mechanism that computes a single real-valued score using the quaternion inner product and applies a shared attention distribution across all components. This reduces score-computation multiplications by 75% and the number of softmax operations from four to one. We prove that, when queries and keys are produced by quaternion linear projections that induce component pre-mixing, the component-wise and shared scores lie in the same interaction subspace, indicating that independent component-wise attention primarily re-parameterizes the same interactions rather than expanding the feature interaction space. In speech enhancement, our method reduces inference time by up to 44.3% on a GPU and 58.1% on a CPU while maintaining quality, with consistent trends across vision and natural language processing.
Authors: Zhi (Leo), Wang, Botao He, Kelin Yu, Seungjae Lee, Ruohan Gao, Furong Huang, Yiannis Aloimonos
Abstract: Human egocentric video captures rich manipulation demonstrations without any robot hardware, yet transferring these skills to robots remains challenging due to the embodiment gap between human and robot in both visual appearance and kinematics. We present HumanEgo, a framework that bridges the embodiment gap by lifting each human demonstration to an entity-level representation of hand-object interaction, and training a flow matching policy with dense auxiliary objectives that amplify supervision from every trajectory. HumanEgo is robot-data-free, hardware-agnostic, data-efficient, and zero-shot human-to-robot transferable. With only 30 minutes of human videos per task, HumanEgo achieves 92.5% average success across four real-world tasks (75% with just 15 minutes), outperforms matched-time robot teleoperation by 41%, and robustly transfers zero-shot across novel robots, cameras, and environments.
Authors: Jianrui Zhang, Hyun Jung Lee, Sukanta Ganguly, Tae-Eui Kam, Donghyun Kim, Yong Jae Lee
Abstract: Multimodal retrieval relies heavily on single-vector retrievers, which compress rich, sequential token sequences into one single global representation. While efficient, they discard fine-grained, local evidence critical for dense retrieval tasks. Multi-vector approaches were introduced as a solution, but they strictly require training and many ignore the necessity of a globally summarizing representation. To address this, we introduce SMART, a framework that unlocks the latent multi-vector capabilities of standard single-vector models. We first demonstrate that standard contrastive training on the pooled embedding implicitly shapes the retrieval geometry of preceding hidden states via gradient flow. By applying direct late-interaction over these frozen hidden states during inference, SMART acts as a plug-and-play upgrade that consistently improves performance across diverse modalities, improving even the state-of-the-art models further on MMEB-V2. We also reveal SMART's superior performance, as simple lightweight post-training not only saves time and compute, but also brings forth further improvement on Visual Document retrieval, allowing a single-vector model to outperform SoTA multi-vector counterparts. Ultimately, SMART offers both a highly efficient inference enhancement and a powerful finetuning technique for multimodal retrieval. We open source our code and weights at https://github.com/HanSolo9682/SMART.
Authors: Narmeen Oozeer, Shivam Raval, Philip Quirke, Manikandan Ravikiran, Jeff Phillips, Shriyash Upadhyay, Amirali Abdullah
Abstract: Steering a language model - intervening on its internal activations to change downstream behaviour - has recently expanded beyond linear interpolation to nonlinear methods such as angular and kernelized steering, which define intervention transformations without learning an explicit geometry over paths in activation space. Freshly introduced geometry-aware manifold methods do learn such a geometry, but require labelled class centroids together with prescribed cyclic or sequential structure. These assumptions restrict where manifold steering can be applied, since existing constructions require labelled centroids and compatible boundary conditions. We recast manifold steering more broadly as \textbf{Riemannian geodesic computation} on activation space, recovering linear and labelled-spline steering as geodesics under particular choices of metric. A principled metric within this framework is the output-space Hellinger distance pulled back to activations; we approximate this with a learned encoder trained on output distances over a small concept-token schema - no per-prompt labels, no topology prior, and no per-task curve fitting. Empirically, the method reliably drives the model onto the target class across all tasks in a standard four-task language-model arithmetic benchmark, while following more behaviourally natural trajectories than baselines on smaller output spaces. We thereby provide a unified Riemannian framework for manifold steering together with a schema-supervised, label-free instantiation that operates without labelled centroids or prescribed boundary conditions.
Authors: Ruize Li, Zhibin Wen, Tao Han, Hao Chen, Fenghua Ling, Wei Zhang, Song Guo, Lei Bai
Abstract: Accurate evaluation of weather forecasting models is critical for their reliable deployment in real-world applications. However, existing benchmarks predominantly rely on reanalysis products such as ERA5, which are generated through delayed data assimilation and do not reflect the constraints of real-time operational forecasting, thereby resulting in a systematic mismatch between benchmark performance and real-world forecasting. In this work, we introduce RealBench, a next-generation benchmark for AI weather forecasting that emphasizes realistic evaluation under operational conditions. RealBench features a strictly out-of-distribution test set spanning 2025 to eliminate data leakage and capture recent atmospheric regimes. It integrates multiple data sources, including low-latency operational analysis and a large-scale global in-situ observation dataset comprising over 10,000 stations, enabling direct evaluation against real atmospheric measurements. Beyond standard global metrics, RealBench provides a comprehensive evaluation framework for high-impact extreme events, including heatwaves, cold surges, and tropical cyclones, using event-specific metrics that better reflect real-world forecasting priorities. The evaluation results reveal substantial discrepancies between reanalysis-based metrics and real-world performance, particularly concerning extreme events. By highlighting the limitations of existing benchmarks, this work establishes a more faithful and operationally relevant evaluation paradigm, providing a rigorous foundation for advancing next-generation AI weather forecasting systems. The benchmark implementation is available at: https://github.com/lixruize-del/NWP-Benchmark.
Authors: William Guanting Li (University of Queensland), Alsharif Abuadbba (CSIRO Data61), Kristen Moore (CSIRO Data61), Dan Dongseong Kim (University of Queensland)
Abstract: Penetration testing is essential to securing modern web infrastructures, yet traditional manual methods struggle to keep pace with their scale and complexity. Large Language Models (LLMs) offer new opportunities for automating these tasks, but existing approaches face two persistent challenges: hallucination of technical entities and insufficient long-term contextual memory. To address these issues, we present APT-Agent, a fully automated LLM-driven penetration testing framework that systematically orchestrates reconnaissance, exploitation, and exfiltration. APT-Agent introduces a hybrid rectification module to recover hallucinated commands and a command-specific memory architecture to preserve operational context across multi-step attack sequences. We evaluate our APT-Agent on Metasploitable 2 against seven vulnerable services spanning web, database, and network protocols. APT-Agent achieves an 84.29% end-to-end exploitation success rate, compared to 48.57% (Script Kiddie) and 18.57% (PentestGPT) under matched conditions. By reducing cognitive burden and minimizing reliance on human intervention, APT-Agent represents a step toward scalable, reliable, and cognitively efficient automation for penetration testing.
Authors: Han Liu, Zhi Xu, Xiaotong Zhang, Feng Zhang, Xiaoming Xu, Wei Wang, Fenglong Ma, Hong Yu
Abstract: Despite the strong performance of deep neural networks in modern Web and language applications, they remain vulnerable to adversarial attacks, especially transferable attacks that generate adversarial examples using surrogate models without accessing the victim model. Transferable attacks in the text domain are still under-explored, with only a few studies addressing this challenging issue, often with suboptimal results due to equal treatment of submodels or inaccurate estimation of importance scores. To address these challenges, we propose a simple yet effective paradigm for transfer-based textual adversarial attack, named SEP-Attack. Specifically, we employ the Determinantal Point Process (DPP) to generate diverse surrogate ensemble weights, representing the transferability of submodels. Using these weights, we introduce a new metric to evaluate prediction confidence scores, which in turn are used to calculate word importance scores and generate adversarial candidates. Finally, we quantify the transferability score for each candidate and select the top ones as the final transferable adversarial examples. Experiments conducted on four datasets and two real-world APIs validate the efficacy of SEP-Attack, significantly outperforming state-of-the-art baselines.
Authors: Jingyi Sun, Qianli Wang, Pepa Atanasova, Nils Feldhus, Isabelle Augenstein
Abstract: Chain-of-Thought (CoT) faithfulness, i.e., whether CoTs genuinely reflect large language models' (LLM) underlying behavior, is typically evaluated under two disjoint paradigms: contextual faithfulness, measured by perturbing the input or CoT trace, and parametric faithfulness, assessed by intervening on a model's parametric knowledge. Yet prior work compares them only descriptively. We fill this gap by proposing FaithMate, a unified preference-alignment interface for optimizing models towards either faithfulness paradigm. It enables us to investigate the interplay between the two paradigms, examining whether and to what extent faithfulness gains generalize within and across paradigms. Across three models, two datasets, and six faithfulness metrics, we find that the two paradigms are positively coupled, yet asymmetric: optimizing towards parametric faithfulness yields consistent gains across both paradigms, whereas the contextual counterpart delivers more variable gains. Within the contextual paradigm, faithfulness gains on one metric do not consistently transfer to others, implying that existing contextual metrics capture disjoint facets of faithfulness and exposing inherent trade-offs. These findings imply that CoT faithfulness is not a monolithic objective and therefore requires multifaceted optimization and evaluation.
Authors: Ibrahim Delibasoglu
Abstract: The rapid evolution of generative models has enabled the creation of hyper-realistic facial deepfakes, exposing a critical vulnerability in modern digital forensics: the inability of detectors to generalize to unseen manipulation techniques. Traditional networks suffer from representation collapse, overfitting to localized artifact fingerprints of specific training generators. This work investigates whether modern Vision Foundation Models can serve as generalizable, out-of-the-box feature extractors capable of tracking forensic anomalies across entirely unseen generative manifolds. We conduct a systematic cross-domain evaluation comparing three foundational learning paradigms: fully supervised macro-semantic features (RoPE-ViT), pure self-supervised geometric features (DINOv3), and multi-teacher agglomerative representations (NVIDIA C-RADIOv4-H). By deploying frozen backbones subjected to downstream linear probing, we map the performance limitations of these architectures on the challenging DF40 benchmark. Our empirical findings expose the intrinsic trade-offs between pre-training paradigms and parameter scale, proving that while foundation models retain high discriminative capabilities for entire face synthesis, localized face editing techniques expose fundamental boundaries in linear probe evaluation structures. Source code and model weights are available in http://github.com/mribrahim/deepfake
Authors: Chang Chu, Qingyue Zhang, Shao-Lun Huang, Junxiong Zheng
Abstract: Long-tailed recognition suffers from a persistent head--tail trade-off: improving tail performance often degrades head accuracy and can increase training instability. Despite strong empirical results from re-weighting, decoupled training, and multi-expert methods, key design choices about representation sharing between head and tail classes and supervision weighting across class groups remain largely heuristic. In this work, we propose OSDTW, a principled task-decomposition framework that partitions the original single-label recognition problem into a head task and a tail task, implemented with a shared encoder and task-specific decoders. To handle the mutual exclusivity and statistical dependence between the two label groups, we introduce a factorized model and show that the resulting Kullback--Leibler divergence-based generalization error can be written as the sum of task-wise terms up to an additive constant, yielding a well-defined task-wise objective. We further develop a three-stage training pipeline: independent task training to estimate task-wise optima and the Fisher information matrix, weighted joint training to learn a shared encoder, and branch assembly to construct the final decoupled model. Under a block-diagonal Fisher approximation, we derive a computable second-order expansion of the expected generalization error, decomposing it into encoder variance, encoder bias, and decoder variance. This bias--variance decomposition provides a computable proxy to select the shared depth and task weights, enabling efficient hyper-parameter search. Experiments on standard long-tailed benchmarks demonstrate the effectiveness of the proposed approach over strong baselines.
Authors: Hongjiang Chen, Pengfei Jiao, Ming Du, Xuan Guo, Zhidong Zhao, Di Jin, Xiao Liu
Abstract: The growing interest in Temporal Graph Neural Networks (TGNNs) stems from their ability to model complex dynamics and deliver superior performance. However, TGNNs encounter fundamental challenges in capturing long-term dependencies and identifying periodic patterns. To address these limitations, we propose TGFormer, a novel Transformer architecture specifically designed for temporal graphs. Our model redefines temporal graph learning by establishing a trajectory framework that aligns with time series analysis principles. This approach allows TGFormer to derive node representations through systematic analysis of historical interactions, enabling granular examination of node relationships across sequential timestamps. Building upon stochastic process theory, we develop an auto-correlation mechanism that systematically uncovers periodic dependencies in node interactions. This innovation empowers TGFormer to perform dependency discovery and representation aggregation at sub-interaction levels, demonstrating superior efficiency and accuracy compared to conventional attention mechanisms. Experimental validation across six public benchmarks confirms the effectiveness of our approach, with TGFormer at most achieving 9.35\% precision improvement compared to state-of-the-art approaches.
Authors: Bangrui Xu, Ziyang Miao, Xuanhe Zhou, Yiming Lin, Zirui Tang, Xiaomeng Zhao, Fan Wu, Cheng Tan, Fan Wu, Bin Wang, Conghui He
Abstract: VLM-based OCR models have become the de facto choice for document parsing, as they can accurately extract page-level elements (e.g., paragraphs within individual pages) together with their bounding boxes and textual content. However, downstream applications such as RAG require coherent document-level information, whereas these models often break cross-page continuity and fail to recover disrupted structures, such as paragraphs and tables truncated by page boundaries. Such relationships are not confined to a single page; instead, they require joint analysis of titles, paragraphs, tables, and images spanning multiple pages. A natural solution is therefore to reuse existing OCR outputs and reconstruct document-level logical structures through post-processing. To this end, we propose MinerU-Popo, a lightweight and universal framework for POst-Processing OCR outputs, which converts page-level results from diverse parsers into coherent document-level structures. MinerU-Popo decomposes the problem into four focused subtasks: text truncation recovery, table truncation recovery, title hierarchy reconstruction, and image-text association. To address these effectively, we build a task-oriented data engine with task-specific input filtering, and use the generated data (30K) to fine-tune a lightweight post-processing model (Qwen3-VL-4B). To support long documents, we introduce dynamic chunking with overlap-based synchronization, which aligns chunk-level outputs from the fine-tuned model and preserves global consistency. Finally, we assemble the aligned outputs into a tree-structured document representation, further enriched with node chunking and summaries for downstream retrieval and analysis. Empirical results show MinerU-Popo improves title-hierarchy TEDS by at least 20% across all five tested OCR models, improves RAG accuracy and reduces per-query latency.
Authors: Gianluca Sabatini, Chenhao Li, Marco Hutter
Abstract: Proximal Policy Optimization (PPO) has become the de facto standard for training legged robots, thanks to its robustness and scalability in massively parallel simulation environments like IsaacLab. However, its on-policy nature makes it inherently sample-inefficient, preventing its use for continuous adaptation and fine-tuning on real hardware. Soft Actor-Critic (SAC), by contrast, is an off-policy algorithm that can reuse past experience, making it a natural candidate for sim-to-real transfer workflows where the same algorithm can be used both in simulation and for online learning on the real robot. Despite these advantages, SAC has consistently failed to match PPO's empirical performance in massively parallel training settings. This work identifies the root causes of this gap and introduces targeted modifications, covering policy initialization, timeout-aware critic targets, and multi-step return estimation, that enable SAC to train stably at scale. Evaluated across multiple legged robot platforms and diverse locomotion tasks, our approach closes the performance gap with PPO entirely.
Authors: Moyu Zhang, Yun Chen, Yujun Jin, Jinxin Hu, Yu Zhang, Xiaoyi Zeng
Abstract: Scaling test-time compute has proven highly effective for language models, yet this opportunity remains largely unexplored for industrial Click-Through Rate (CTR) prediction. CTR models suffer from a fundamental asymmetry: feature combinations well-represented in training yield confident predictions, while sparsely observed ones produce unreliable outputs. Existing training-phase solutions such as adaptive gating learn a fixed selection function subject to the same sparsity, offering no per-instance recourse at deployment.We propose UTTSI (Uncertainty-Triggered Test-Time Selective Inference), a training-free model-agnostic framework that scales inference depth proportionally to per-instance uncertainty. A dual-signal estimator combining model logit confidence with a data-level frequency prior distinguishes epistemic uncertainty from aleatoric ambiguity. Every instance undergoes adaptive feature filtering to remove unreliable embeddings; uncertain instances additionally receive stochastic feature-path explorations whose predictions are aggregated via consistency-weighted ensembling. Confident instances bypass exploration entirely, keeping average overhead at approximately $2.8\times$ base model cost with worst-case latency unchanged.Experiments on four datasets with three backbone architectures demonstrate consistent, statistically significant gains over all training-phase baselines. A seven-day online A/B test further confirms a 5.3% relative CTR gain ($p < 0.01$), establishing selective test-time compute allocation as a practical complement to training-phase advances for CTR prediction.
Authors: Changling Li, Ying Li
Abstract: Multi-agent reinforcement learning (MARL) has shown wide applicability in collaborative systems such as autonomous driving and smart cities for its ability of learning through interaction. With the recent development of drone networks, researchers have also applied MARL to address the trajectory planning problems. However, the dynamic environment and the limited battery capacity are still challenging for using MARL to achieve efficient collaborative task execution. In this paper, we propose an energy-aware MARL model as an attempt to tackle these challenges, leveraging Deep Q-Networks (DQN) with \emph{individual reward functions} driven by the task execution progress and the remaining battery of drones. We conduct a set of simulation studies for the proposed mode and compare it with the shared reward MARL~\cite{Li2022MARL} to explore the impact of credit assignment in MARL. The results indicate that our proposed model can achieve at least 80\% success rate regardless of the task locations and lengths. Similar to the shared reward mode, the individual reward mode can achieve a better success rate when the task density is high, and it can hit nearly a 100\% success rate when task density gets close to 40\%. The true advantage of our proposed model with individual reward is revealed when scaling up the environment. The comparison to the shared reward MARL shows that the our proposed model is more robust towards the change of the environment size and agent numbers. It can achieve higher success rate with fewer steps due to the clarity of the goal which improves energy efficiency even better.
Authors: Chainarong Amornbunchornvej
Abstract: An agent must act on the situation before it, learn what it cannot yet represent, and model other agents well enough to coordinate. These faculties are usually realized by separate mechanisms, yet they share a failure mode: the situation can exceed what the agent can currently represent, and the honest response is then a principled refusal that says what was missing. We develop a small cognitive architecture in which these limits arise from a single quantity. An Interpretation-Decision Unit (IDU) interprets a content vector through a family of regimes - local representational frames with private bases - and decides which actions it licenses; a scalar residual of the content against the active regimes' representational scope drives the unit. Low residual with a clean licensing emits an action; otherwise the unit re-interprets, attempts a description-length-justified expansion, or halts with a typed, witnessed terminal. We prove the unit is total and deterministic: for any content and fixed configuration it halts in finitely many bounded-cost steps with a unique terminal witness, so abstention carries its cause by construction. By binding the architecture's open parameters without changing its mechanics, the same residual-against-scope constraint recovers three documented phenomena at three scopes: the typology of not-knowing (typed abstention); a forced misunderstanding between agents, localized to one shared concept and invisible to the agent committing it (bounded empathy); and prerequisite dependence in learning derived from a bounded focus window rather than posited (developmental prerequisites). Each instantiation is worked for a natural and an artificial agent and states a falsifiable prediction, so one constraint can model limits in both human and machine cognition. The account contributes a unification and a notion of accountable abstention, typed and witnessed by construction.
Authors: Qishen Zhou, Yifan Zhang, Michail A. Makridis, Anastasios Kouvelas, Yibing Wang, Simon Hu
Abstract: Inferring network-wide traffic states from sparse observations with high accuracy and trustworthy uncertainty quantification is essential for intelligent transportation systems, yet it remains challenging due to the underdetermined nature of the problem, multifaceted disturbances in sensing networks, and the inherent conflicts among multiple inference sub-tasks when modeled jointly. We propose the Task-Aware Attentive Neural Process (TA-ANP), a unified probabilistic framework for resilient and trustworthy global traffic state inference (GTSI) by fusing floating car data (FCD) with sparse fixed-detector measurements. By casting GTSI as a stochastic process, TA-ANP leverages the meta-learning properties of neural processes to adapt rapidly to changes in sensing configurations without retraining. A task-aware multi-query attention module with distinct spatiotemporal inductive biases is introduced to jointly handle three GTSI sub-tasks, while mitigating cross-task interference. For uncertainty quantification, we combine neural processes with Monte Carlo Dropout to capture both aleatoric and epistemic uncertainty. To support metropolis-scale evaluation, we construct the Metropolitan Multi-Source Traffic Dataset (MMTD), integrating fixed-loop sensor measurements, FCD statistics, and OpenStreetMap road-network data over an urban network of 2,371 road segments. Experiments on MMTD show that TA-ANP achieves state-of-the-art performance across all sub-tasks under deterministic and probabilistic metrics. The resulting well-calibrated uncertainties enable more efficient fixed-sensor placement with fewer sensor deployments. Under a Damage-Repair-Addition sensing lifecycle, TA-ANP demonstrates superior resilience in terms of disturbance absorption, performance recovery, and adaptability to unseen sensing configurations.
Authors: Hichem Cheriet, Badra Khellat Kihel, Samira Chouraqui
Abstract: Integrating artificial intelligence (AI) into sampling-based motion planning provides new possibilities for improving autonomous navigation efficiency. In this paper, three algorithms, namely RRT*, Neural RRT*, and Neural Informed RRT*, are implemented and evaluated on environments containing convex and concave obstacles with different obstacle densities. The obtained results indicate that neural-guided planners improve path quality, producing up to 14\% shorter paths and 55--75\% smoother trajectories compared with the conventional RRT* algorithm. Among the evaluated methods, Neural Informed RRT* achieves the best overall performance in terms of path length and trajectory smoothness. These results demonstrate the effectiveness of AI-guided sampling strategies for improving reliability and trajectory efficiency in robotic and UAV navigation, despite a slight increase in computation time. Overall, the study highlights the growing importance of artificial intelligence in real-time robotic path planning applications.
Authors: Wenjie Zheng, Haoji Hu, Jiali Lu, Xingze Zou, Jing Wang
Abstract: Dataset distillation (DD) aims to compress large-scale datasets into compact synthetic sets while preserving training efficacy. However, existing studies mainly focus on image classification, leaving dense prediction tasks such as semantic segmentation largely underexplored. In this work, we identify three key challenges for segmentation DD: (i) long-tailed class imbalance, (ii) the need for strict pixel-wise alignment between images and dense labels, and (iii) the high computational cost of optimizing high-resolution data with complex models. To address these challenges, we propose D3S2, a Diffusion-guided Dataset Distillation framework for Semantic Segmentation. Our method adopts a two-stage design. In Class-Balanced Mask Selection, we construct a representative mask set via a greedy strategy that prioritizes underrepresented classes. In Diffusion-Guided Image Synthesis, we employ a pretrained layout-to-image diffusion model to generate images conditioned on the selected masks, naturally ensuring spatial alignment. To further enhance the training utility of synthesized data, we introduce guided diffusion sampling with two complementary objectives: a segmentation-consistency loss for pixel-level alignment, and a class-wise feature matching loss for aligning per-class feature statistics across layers. Extensive experiments demonstrate the superiority of D3S2. Notably, at an extremely compression rate of 1%, our method achieves 24.99% and 35.49% mIoU on ADE20K and COCO-Stuff with Mask2Former (Swin-S), outperforming random selection by 9.34% and 5.70%, respectively.
Authors: Yangneng Chen, Jing Li
Abstract: Large Vision-Language Models (LVLMs) extend large language models with visual understanding, but remain vulnerable to hallucination, where outputs are fluent yet inconsistent with images. Recent studies link this issue to language bias-the tendency of LVLMs to over-rely on text while neglecting visual inputs. Yet most analyses remain empirical without uncovering its underlying cause. In this paper, we provide a systematic study of language bias and identify its root in modality misalignment during training. Our analysis shows that both Visual Instruction Tuning (VIT) and Direct Preference Optimization (DPO) often prioritize textual improvements, which may cause LVLMs to overly lean toward language modeling rather than balanced multimodal understanding. To address this, we propose two simple yet effective methods: Language Bias Regularization (LBR) which mitigates language bias through regularization during instruction tuning, and Language Bias Penalty (LBP), which penalizes language bias in the DPO training process. Extensive experiments across diverse models and benchmarks demonstrate the effectiveness of our approach. LBR consistently improves performance on over ten general benchmarks, while LBP significantly reduces hallucination and improves trustworthiness. Together, these methods not only mitigate language bias but also advance the overall alignment of LVLMs, all without introducing any additional data or auxiliary models. Our code is publicly available at https://github.com/lab-klc/LVLM-Language-Bias.
Authors: Jun-Wei Hsieh, Meng-Yu Kao, Ghufron Wahyu Kurniawan, Kuan-Chuan Peng
Abstract: YOLO-series and DETR-based detectors struggle with tiny-object detection. YOLO-style models benefit from efficient dense prediction, but their large-stride backbones may suppress tiny instances in deep feature maps and make grid assignment ambiguous. DETR-based models remove hand-crafted post-processing through set prediction, yet they reason over coarse token grids, where tiny objects occupy only a few weak tokens and are easily overlooked during matching. To address these limitations, we propose TinyFormer, a unified YOLO--DETR hybrid real-time detector that combines ViT representations, NMS-free set prediction, and a YOLO-style pyramid neck for accurate small-object detection. TinyFormer introduces a Parallel Bi-fusion Module (PBM), which builds high-resolution shortcuts from shallow stages to the feature pyramid, preserving fine spatial details during multi-scale fusion. We further design a Spatial Semantic Adapter (SSA) to compensate for the spatial loss caused by coarse tokenization. SSA extracts high-resolution cues from early stages and injects them into transformer token embeddings, improving tiny-object localization without sacrificing the global modeling ability of DETR. Experiments on MS COCO show that TinyFormer consistently outperforms recent YOLO-series detectors and the strong DEIMv2 baseline. TinyFormer-X achieves 58.4% AP even without PBM, while adding PBM improves the overall AP to 58.5% and brings a 1.6% AP gain on small objects. With Objects365 pre-training, TinyFormer-X-PBM reaches 60.2% AP, surpassing RF-DETR and other Objects365-pretrained detectors with fewer parameters and lower computation. These results demonstrate that TinyFormer bridges dense YOLO-style feature fusion and DETR-style set prediction, providing a strong accuracy-efficiency trade-off for real-time tiny-object detection. Code is available at https://github.com/mmpmmpmmpjosh/TinyFormer.
Authors: Ayush K. Varshney, Konstantinos Vandikas, \v{S}ar\=unas Girdzijauskas, Adam Orucu, Aneta Vulgarakis Feljan
Abstract: Deploying deep neural networks on resource-constrained 6G edge devices demands aggressive compression with minimal accuracy loss. Quantization-Aware Training (QAT) has emerged as a leading compression approach; however, existing mixed-precision methods typically operate at coarse layer- or channel-level granularity. These methods often rely on heuristic or search-based bit-allocation strategies, which may overlook fine-grained variability at the neuron level. We propose Neuron-Level Mixed-Precision QAT (NMP-QAT), where each neuron independently learns its own discrete precision during training. Starting from low-bit precision, NMP-QAT expands bit-width only when training signals demand it, via differentiable surrogates and straight-through estimators, while preserving a fully discrete inference graph. This adaptability extends to both weights and activations, reducing memory movement. Evaluated on telecom and non-telecom datasets across MLP and tabular foundation model architectures, NMP-QAT achieves superior compression-accuracy trade-offs over mixed-precision QAT baselines, making it well-suited for Green AI deployments at the network edge.
Authors: Gang Peng
Abstract: Current AI interaction models treat the prompt as the primary object of exchange, omitting a critical layer: the user's latent source intent, the goal state preceding and motivating the prompt. Here we introduce Intent Signal Theory (IST), a computational framework that formalises this missing intent layer. IST distinguishes four objects routinely conflated: latent source intent (I*), observable intent proxy (I-hat), encoded carrier (P), and model output (O). It formalises dimensional weights, encoding masks, structural and fidelity recovery scores, and public-private intent decomposition. The Theorem of Irreversible Intent Loss establishes that private intent absent from the carrier cannot be recovered beyond generic substitution. Evidence from four companion studies spanning six LLMs, three languages and three task domains shows structural-fidelity splits, human-validated metric dissociation, and weight-tolerance plateaus consistent with IST's predictions. IST reframes prompt engineering as intent-protocol design and identifies a computational layer that current AI systems lack.
Authors: Ziyi Wang, Dongyang Kuang
Abstract: EEG-based emotion recognition holds significant promise for objective diagnosis of mood disorders. Graph neural networks (GNNs) have emerged as the dominant paradigm for modeling inter-channel dependencies in EEG, yet existing approaches rely on symmetric adjacency matrices derived from spatial proximity or functional correlations that fundamentally capture statistical associations rather than directed causal influences, which conflicts with the inherently asymmetric, causally-driven nature of neural information flow. To bridge this gap, we propose GL-LFGNN, a Global-Local Dual-branch Causal Graph Neural Network grounded in Liang-Kleeman information flow theory. Unlike Granger causality that merely assesses temporal precedence, our approach rigorously quantifies causal strength from a dynamical systems perspective, yielding neurophysiologically interpretable directed graphs. A dual-branch architecture further integrates whole-brain connectivity with region-specific processing aligned to established functional neuroanatomy. On the MEEG dataset, GL-LFGNN achieves 86.17% (Arousal) and 86.71% (Valence) accuracy with only 37K parameters -- approximately 10% of the current state-of-the-art -- demonstrating that principled causal modeling can simultaneously enhance interpretability, generalization, and computational efficiency. Code will be released.
Authors: Ata G. Zare
Abstract: The dominant artificial intelligence paradigm trains neural architectures via gradient descent against proxy objectives and reinforcement learning from human feedback. While remarkably capable, this top-down optimization inherently generates structural failure modes, including hallucination, sycophancy, reward hacking, and alignment fragility, which represent paradigmatic limitations rather than mere engineering defects. In response, we introduce RECLAIM (Recursive, Ecological, Cognitive, Lifelike, Adaptive, Intelligent Machine), a theoretical framework for cultivating intelligence through computational ecology rather than engineering it through strict optimization. The model is supported by four interlocking theoretical pillars. General Darwinism replaces gradients with blind variation and selective retention, while non-agentic emergence substitutes evaluative rewards with environmental physics to structurally prevent specification gaming against human intent. Concurrently, the Polya-Hebbian bridge applies Polya urn dynamics to Hebbian reinforcement for path-dependent specialization, and the free energy principle is integrated as environmental thermodynamics rather than as an agent objective. The architecture situates autopoietic units, bounded by Markov blankets and competing for finite computational energy, within a data ecology shaped by cognitive food chains and Red Queen arms races. This framework suggests the spontaneous emergence of dual-process cognition, sensory specialization, analogical reasoning, and intrinsic motivation as natural consequences of evolution under resource constraints. We conceptualize this paradigm transition as the OMEGA shift, representing a move from optimization and maximization to emergence through generative autopoiesis.
Authors: Wenjuan Li, Yitao Liu, Runze Chen, Rajkumar Buyya
Abstract: Background: Fine-tuning is central to adapting pre-trained Large Language Models (LLMs) to downstream tasks, but its reliance on training data, parameter updates, and reusable components opens entry points for attackers. Threats have evolved from data poisoning and weight tampering to agent manipulation and interface exploitation, yet existing reviews lack a unified framework spanning the full fine-tuning lifecycle. Objective: This paper presents a systematic survey of LLM fine-tuning security and establishes a lifecycle-based framework for comparing attacks and defenses, complemented by unified empirical evaluation. Methods: We divide attack and defense mechanisms into three phases by intervention timing: pre-tuning, during-tuning, and post-tuning. Within each phase, strategies are reviewed and contrasted to expose their evolution and limitations. Representative methods are then evaluated under a unified model, hardware, and protocol setup, with cross-phase experiments pairing attacks and defenses from different phases. Results: Attack effectiveness is highly model-dependent and non-monotonic with scale: weight-editing attacks effective on earlier models lose impact on modern open-source LLMs; cross-lingual backdoor transfer, reported as near-perfect at larger scales, fails entirely on tested 1B-4B models; and purely benign samples can compromise safety alignment in instruction-tuned models. Single-phase defenses rarely generalize across phases, and defense effectiveness depends jointly on model architecture and alignment state. Conclusion: We identify key open problems (configuration-robust defense, cross-phase defense composition, and embedding-space attacks beyond behavioral assumptions) and propose concrete future research directions.
Authors: Munsik Kim
Abstract: We study the rate-distortion limits of online KV cache compression in autoregressive language models, formulating it as sequential Wyner-Ziv source coding on the filtration induced by the model, with the next-step query as decoder side information. Empirically, across four models spanning two families and $0.5$-$3$B parameters, we find that the next-token distribution's sensitivity to context truncation decays \emph{polynomially} rather than \emph{geometrically}: a power law improves on an exponential fit by an order of magnitude in extrapolation, the fitted exponent is recovered independently from a sink-plus-recent KL measurement, and the decay is verified to be free of positional-encoding artifacts by a position-preserving ablation. Under a corresponding \emph{polynomial truncation-sensitivity} assumption, our main result characterizes the per-token memory requirement of \emph{suffix-only} cache policies: a sliding-window scheme attains distortion $\varepsilon$ with window $w = O(\varepsilon^{-1/\alpha})$, and -- under an additional two-sided Bayes-risk condition -- a converse shows $w = \Omega(\varepsilon^{-1/\alpha})$ is necessary within this policy class, so the scaling is $\Theta(\varepsilon^{-1/\alpha})$ for suffix-only policies. Whether recurrent or propagating cache summaries can beat this scaling is left open. An explicit block-Markov scheme achieves the upper bound; its rate-of-convergence exponent matches the converse under additional forward-decay and regularity hypotheses (not implied by truncation sensitivity alone), and differs by a factor of two otherwise. Empirically, the polynomial law predicts the degradation curves of concrete cache policies: recency-based eviction (sliding, sink-plus-recent) suppresses distortion by roughly two orders of magnitude over random retention at equal budget, with a power-law decay in the budget.
Authors: Ashir Kulshreshtha, Abdullah Mughees, Gaadha Sudheerbabu, Tanwir Ahmad, Kristian Klemets, Dragos Truscan, Mikael Manng{\aa}rd
Abstract: In many industrial domains, the Functional Mock-up Interface (FMI) is used to exchange simulation models as Functional Mock-up Units (FMUs) across different partners using various modelling tools. This opens up the possibilities for simulation-based verification and validation using FMUs for ensuring reliable system behaviour. However, deriving effective test oracles for these simulation models remains challenging due to the absence of explicit expected outputs. This limits the applicability of conventional testing approaches, which require access to the internal workings of the systems. Metamorphic testing (MT) addresses this limitation by leveraging metamorphic relations (MRs), but extracting such relations from specifications remains largely a manual and error-prone process. To address this challenge, we propose an LLM-powered multi-agent workflow for specification-based metamorphic testing of FMU-based simulation models. The approach takes functional and interface specifications as input and orchestrates multiple agents to extract requirements and derive MRs. These MRs are expressed using Given-When-Then patterns to structure input conditions (Given), transformations (When), and expected output behaviours (Then). These relations are then used to generate metamorphic test cases, execute simulations, and evaluate output consistency across multiple sessions. We evaluate the approach on a Lube Oil Cooling system FMU, demonstrating its ability to automatically generate meaningful MRs and corresponding test cases. Preliminary results indicate that the proposed workflow can effectively support the systematic verification and validation of dynamic simulation models by reducing manual effort and improving test generation.
Authors: Jules Berman, Tobias Blickhan, Benjamin Peherstorfer
Abstract: Existing work on population dynamics inference often focuses on flows arising from vector fields that are the gradients of scalar potentials. Among all admissible flows that are compatible with the population dynamics, gradient flows are optimal in a specific sense: they minimize kinetic energy. The selection of fields based on different criteria corresponds to a gauge freedom when determining population dynamics, which we leverage in this work. We propose Non-Gradient Inference Flows (NGIF), an algorithm to infer non-gradient population dynamics using a weak formulation of the continuity equation. This allows us to parameterize general vector fields and choose other selection criteria beyond minimal kinetic energy. We demonstrate on a variety of low- and high-dimensional physics problems that this more general approach improves distributional accuracy over gradient-restricted baselines and better captures non-potential transport.
Authors: Lei Wang, Syuan-Hao Li, Yongsheng Gao, Piotr Koniusz
Abstract: Aligning structured data is a fundamental problem in computer vision and machine learning, underlying tasks such as time series analysis, human action recognition, and visual representation learning. Existing alignment methods, including Dynamic Time Warping (DTW) and its differentiable variants, rely on deterministic similarity measures and are therefore sensitive to heterogeneous and noisy features. In this work, we introduce uncertainty-aware alignment, a probabilistic framework that models pairwise correspondences with heteroscedastic uncertainty and performs structured matching along alignment paths. Our formulation, uncertainty-DTW (uDTW), assigns each correspondence a Normal distribution and parametrizes each alignment path by a Maximum Likelihood Estimate objective consisting of (i) a precision-weighted matching term that suppresses unreliable features, and (ii) a log-variance regularization that prevents degenerate solutions. This yields a probabilistic alignment mechanism that is robust to noise and interpretable, as uncertainty directly reflects the reliability of matches. We further generalize this framework from temporal sequences to tokenized visual representations, enabling structured matching over sets of visual tokens. The learned uncertainty can be interpreted as a reverse-attention: semantically relevant regions exhibit low uncertainty and dominate the alignment, while ambiguous/noisy regions have high uncertainty. This provides a connection between alignment, attention, and uncertainty modeling. We evaluate the proposed framework across diverse domains. The results demonstrate consistent improvements over state-of-the-art methods and show that learned uncertainty correlates with semantic importance. These findings establish uncertainty-aware alignment as a general, robust, and interpretable framework for learning from structured data.
Authors: Anuj Kumar, Josiah Bjorgaard, Nikolaos Bouklas, Matteo Salvador, Alexander Lavin
Abstract: We introduce "Courant", a Perceiver-based encoder-processor-decoder surrogate model that has latent features exhibiting adaptive specialization and local support in the physical space, enabling functionality akin to an adaptive hp-refinement scheme, an attribute that is highly desirable in traditional numerical solvers and scientific machine learning broadly. The proposed architecture combines a shared random Fourier feature coordinate embedding, state-adapted latent queries, and a light-weight decoder. Courant is trained end-to-end with steady or transient simulation data and only a standard L_2 prediction loss in the physical space, achieving competitive accuracy on benchmarks. We demonstrate that Courant's inductive biases yield latents that are interpretable by design: they develop multiscale geometric specialization in the simulation domain and track coherent structures in the time-dependent case, acting analogously to time-evolving spatial basis functions and allowing for decoding a compact, geometry-anchored, partition-of-unity-like decomposition of the simulated field.
Authors: Xi Ding, Lei Wang, Syuan-Hao Li, Yongsheng Gao
Abstract: Domain adaptation aims to mitigate performance degradation caused by distribution shifts between a labeled source domain and an unlabeled or sparsely labeled target domain. Most existing approaches estimate domain discrepancy either in feature space or in prediction space. However, these single-perspective strategies overlook a critical problem under domain shift: the reliability of the signals used for alignment. In practice, both learned representations and semantic predictions may become unreliable, and treating all target samples equally can lead to misleading alignment and suboptimal transfer. We introduce trust-aware domain adaptation, a principled framework that models domain discrepancy through the reliability of feature and prediction signals. Central to our approach is the Joint Feature-Prediction Discrepancy (JFPD), a unified formulation that jointly captures representation divergence and prediction divergence while weighting their contributions by sample-specific trust. Trust is quantified via two complementary mechanisms: uncertainty-aware trust, derived from prediction entropy to suppress unreliable predictions, and semantic-alignment trust, computed from prototype similarity in feature space to emphasize well-aligned representations. By prioritizing confident and semantically consistent samples while down-weighting noisy or ambiguous ones, JFPD provides a reliability-aware estimate of domain discrepancy. We further integrate JFPD into a training objective that guides adaptation toward trustworthy regions of the target domain. Experiments on standard benchmarks demonstrate that the proposed framework consistently achieves superior adaptation performance and yields discrepancy estimates that correlate with target-domain error. This work addresses, for the first time, the importance of modeling trust in the interaction between features and predictions for domain adaptation.
Authors: Houman Kazemzadeh, Kamyar Naderi
Abstract: Radiology reports remain the primary mechanism by which imaging findings are communicated to clinical teams. However, much of the structured information behind these reports, including measurements, image evidence, prior comparisons, lesion identity, uncertainty, and terminology, often remains trapped in free text or fragmented across picture archiving and communication systems, radiology information systems, reporting workstations, worksheets, advanced visualization tools, and electronic health records. This paper proposes a human-supervised, evidence-linked reference architecture for structured radiology reporting. The framework combines exam-specific templates, speech-to-structure processing, measurement and segmentation capture, controlled AI-assisted drafting, and standards-based interoperability using DICOM, DICOM Structured Reporting, DICOM Segmentation, HL7 FHIR, RadLex, SNOMED CT, LOINC, and UCUM. The system is positioned not as an autonomous report generator, but as a structured intelligence layer for enterprise imaging that supports reviewed reporting, longitudinal comparison, clinical data reuse, governance, and integration with PACS, RIS, EHR, analytics, and registry workflows. The paper also discusses modality-specific deployment considerations, clinical safety risks, validation requirements, cybersecurity, privacy, quality management, and regulatory boundaries for AI-assisted radiology reporting systems.
Authors: Weixin Wang, Yu Yang, Wei Deng, Pan Xu
Abstract: We study inference-time alignment for diffusion-based generative models, aiming to steer a base model toward high-reward outputs without updating its weights. Recent Sequential Monte Carlo (SMC)-based steering methods approximate reward-tilted target distributions in a principled way, but their proposals remain largely tied to the base sampler. Since reward information is mainly used after propagation through particle reweighting and resampling, these methods can require large particle budgets and suffer from weight degeneracy and high-variance estimates. One way to reduce variance and improve particle efficiency is to iteratively learn twisting functions that provide look-ahead guidance, as in twisted SMC. However, existing learnable twisting methods are developed mainly for classical sequential inference and can be unstable when applied to diffusion-based alignment with high-dimensional state spaces and terminal, noisy, or black-box rewards. We propose Trust-Region Iterative Twisted Sequential Monte Carlo (TRI-TSMC), a trust-region framework for learning twisting functions in SMC-based inference-time alignment. Each iteration computes an exact KL-constrained update in path space, which admits a closed-form solution by tempered importance reweighting, and projects this target back to the parameterized twisted family by weighted maximum likelihood. Theoretically, we formalize the value-function interpretation of the optimal twisting function and show that it yields a zero-variance sampler. We prove that the trust-region update follows an escort path toward the target distribution, that the weighted maximum-likelihood update is a forward-KL projection, and that the path reduces residual importance-weight variance. Empirically, TRI-TSMC improves primary alignment objectives on discrete diffusion text generation and text-to-image generation under matched inference-time budgets.
Authors: Huangyu Xu, Jingqin Yang, Qianqian Xu, Jiaye Teng
Abstract: Sparse optimization is a fundamental challenge in various practical applications. A popular approach to sparse optimization is $\ell_p$ regularization. However, it may encounter optimization instability due to the unbounded gradients when $0
Authors: Rai Ali Yar, Umaisa Lail, Anwar Shah
Abstract: Anomaly detection in Industrial Internet of Things (IIoT) environments is essential to protect the Industrial Control Systems (ICS) and Cyber-Physical Systems (CPS) from occuring run time false data injection and other malicious attacks. The increasing complexity of sensor networks and interconnected control loops makes it difficult to identify anomalous behavior hidden within high-dimensional and time-dependent signals. To address these challenges, this article introduces Adaptive Spatio-Temporal Reinforcement Optimization ASTRO (ASTRO), a novel anomaly detection framework that pioneers the use of reinforcement learning for dynamic threshold optimization. By integrating a Deep Q-Network (DQN) with Graph Neural Networks (GNNs), temporal modelling and a Multi-Head Attention mechanism, ASTRO continuously adapts its decision boundaries to improve detection accuracy. The GNN component models the spatial relations among sensors, Temporal model captures time series dependencies and the attention layer highlights most informative time steps. The model generates continuous anomaly scores, which are transformed into binary decisions using an adaptive threshold, optimized via a Deep Q-Network (DQN). The ASTRO approach is evaluated on two real world industrial benchmarks: the Secure Water Treatment (SWaT) and Water Distribution (WADI) datasets. The proposed model achieves an exceptional performance on the SWaT with F1 score of 0.990. Moreover, on highly complex 127 end devices WADI dataset, it secures F1 score of 0.788, outperforming state-of-the-art baselines by nearly 14%. Results across multiple runs confirm consistent generalization and stability. These experiments demonstrate that the ASTRO framework is highly practical and scalable method for strengthening the large scale cyber physical infrastructures
Authors: Pavan Manjunath, Thomas Pruefer
Abstract: Reliable forecasting of renewable energy generation is a foundational requirement for grid stability energy trading battery scheduling and carbon aware operational planning Solar and wind resources are inherently intermittent their output fluctuates with cloud cover wind speed atmospheric turbulence seasonal patterns and local terrain The proliferation of IoT and edge devices spanning smart meters inverters anemometers pyranometers weather stations and grid interface sensors has created an unprecedented volume of real time operational data that conventional forecasting pipelines are ill equipped to exploit fully This review investigates how large language model LLM agents can enhance renewable energy forecasting by integrating heterogeneous sensor streams weather API data historical generation records grid constraints and contextual reasoning into unified decision support workflows We survey classical forecasting methods statistical time series models deep learning architectures physics hybrid approaches and emerging LLM agent frameworks for explanation uncertainty communication and operator guidance A six layer taxonomy is proposed covering data acquisition preprocessing feature engineering model inference uncertainty estimation and natural language reporting The review identifies twelve open challenges spanning real time deployment model drift under distribution shift uncertainty quantification hallucination control in LLM agents interoperability of edge hardware and integration with energy management systems The paper concludes by recommending a research agenda centred on open benchmarks physics informed LLM grounding and federated forecasting architectures
Authors: Kasra Jalaldoust, Elias Bareinboum
Abstract: Prediction models trained under the source distribution do not generalize well to a different target distribution. A valid inference about an unseen data distribution must be anchored by the invariance of certain causal mechanisms that generate the source and target data, however, these structural invariances are non-identifiable from the source data alone. Under mild causal assumptions about the data, we show that the optimal prediction in the target is in fact partially identifiable by the source distribution. The result rests on a simple observation: In any domain, the optimal prediction can be factorized into what we call a pair of abduction and deduction maps, where the abduction map makes inference about some unobserved variables (possibly confounders) from the observed variables and the deduction map predicts the label using both the observed and inferred quantities. Access to large source data pins down the optimal prediction, thus constrains the valid abduction-deduction ensembles that produce it -- a non-identifiability that we call the abduction-deduction entanglement. To leverage this, we parameterize the constrained family using what we call a representation transplant, that is a specific linear transformation in the representation space that manipulates the abduction content of the representation while retaining the deduction component. Invariance of the causal mechanism generating the label implies existence of an invariant deduction map between source and target. Thus, we can search the space of plausible target distributions via a parametric transplant. We use this scheme in a learner-adversary game that, under an idealistic optimization, provably terminates with the learner having the minimax-optimal target prediction. Evaluations verify the theory, showing that the method is competitive in DG benchmarks.
Authors: Liang Xue, Haoyu Liu, Cheng Wang, Pengyu Chen, Haozhuo Zheng, Yang Liu
Abstract: Large language models for vertical domains are bottlenecked by the scarcity of complex, domain-specific task-oriented dialogues. Existing data acquisition pipelines face a persistent trilemma: expert annotation is expensive, real-world service conversations are constrained by privacy and commercial restrictions, and static corpora quickly become temporally stale. We propose Stream, a data-centric framework that leverages publicly available streaming media (live streams and short videos) to synthesize high-value service dialogues at scale. Stream mines authentic interaction signals from noisy streams and synthesizes conversations by integrating role-grounded persona construction with Conversational Blueprint construction; it further adopts retrieval-augmented generation (RAG) to support knowledge-aware responses. Based on Stream, we release StreamDial, a large-scale multi-domain dataset covering Automotive, Restaurant, and Hotel. StreamDial contains 87,498 dialogue sessions and 1,497,320 turns in total, with an average of 17.11 turns per session and a comparable scale across domains. Each session is organized as a structured quadruplet $\langle P_u, P_a, B, H \rangle$ that pairs dialogue history with explicit user/agent personas and a Conversational Blueprint, capturing realistic service behaviors such as requirement mining, constraint conflicts, negotiation, and recovery. Evaluations with automatic judges and downstream tasks show that StreamDial improves intrinsic dialogue quality over strong baselines, and models trained with StreamDial improve Dialogue State Tracking across backbones; we further report a completed human-evaluation set and encouraging multilingual transfer on Qwen3-8B under a controlled training budget. The data is released in https://github.com/hitxueliang/DialogDataSetBySTREAM.
Authors: Bikram Keshari Parida, Abhijit Sen, Wonsang You
Abstract: A panoramic X-ray compresses a 3D jaw into a 2D strip; we aim to recover the missing depth cleanly and fast. Existing implicit neural representations render realistic volumes but are slow to train, sensitive to sampling and positional encodings, and costly in practice. Pure CNN baselines are efficient yet struggle with the dental arch's long-range geometry, blur fine enamel-dentin boundaries, and offer little interpretability. We present K-U-KAN, a three-stage pipeline that (i) lifts 2D features into depth-aware observables with Kolmogorov-Arnold Networks, (ii) advances these observables by a stable, phase-aware linear evolution via a Koopman token block, and (iii) places the predicted depth bins onto focal-trough rays before a lightweight 3D attention U-KAN refines the volume. This marriage of physics (Beer-Lambert image formation), geometry (horseshoe focal trough), and learned linear dynamics yields sharp anatomy, fewer artifacts, and robust behavior on native radiographic intensities with batch size one. On held-out data, K-U-KAN matches transformer/implicit baselines on signal and structure metrics, clearly improves perceptual quality, and trains in roughly half the time-making single-view PX $\to$ CBCT reconstruction more practical for clinical pipelines.
Authors: Rui Wang, Renhao Xue, Ray Razi, Huan Song, Hannah R. Marlowe
Abstract: Time series forecasting models are increasingly scaled through large Transformer backbones, yet most existing approaches process all series through a shared dense computation path despite substantial heterogeneity in temporal structure. Mixture-of-Experts (MoE) offers a natural alternative by enabling conditional computation, but standard MoE routing leaves expert specialization weakly identified and often unstable during downstream adaptation. We propose AME-TS, a structure-guided sparse time series foundation model that aligns expert routing with interpretable temporal structure. AME-TS first uses a lightweight regime predictor to estimate series-level descriptors, including forecastability, seasonality, trend, and sparsity, and maps them to a soft structural prior over experts. This series-level prior guides token-level routing during training, encouraging structure-aligned specialization. On the GIFT-Eval benchmark, AME-TS delivers a strong accuracy-efficiency tradeoff across model scales: it substantially outperforms existing time series foundation models at small model scales and remains competitive with the strongest models at larger scales, while activating substantially fewer parameters through sparse routing. We further show that AME-TS learns more interpretable routing geometry and substantially more stable expert specialization than standard MoE during fine-tuning on the M5 dataset. These results suggest that structure-aware routing is an effective and reliable way to realize the benefits of sparse expert models for time series forecasting.
Authors: Kozachok Elena Sergeevna
Abstract: This study presents a methodology for constructing a clinically verified dataset of dermatoscopic images for medical informatics research. The relevance of the work is driven by the fact that the performance of automated diagnostic support systems depends not only on the volume of images, but also on the reproducibility of the image acquisition procedure, the completeness of structured metadata, and the reliability of diagnostic labels. International collections were primarily created under conditions that differ substantially from routine Russian outpatient practice and mobile dermatoscopy. The proposed methodology integrates three interconnected components: (1) a standard operating procedure (SOP) for acquiring images via mobile dermatoscopy, (2) an information model comprising 16 structured metadata fields organized into six clinically oriented blocks in ISIC-compatible notation, and (3) a multi-stage expert verification of diagnostic labels (initial clinical annotation, consensus review by three specialists, and histological confirmation of all malignant neoplasms). Using this methodology, a dataset of 1,026 unique dermatoscopic images from 443 patients was collected between June 2025 and May 2026. From 1,044 initial records, 18 duplicates were excluded. The dataset includes nine nosological categories; all 39 malignant lesions (18 melanomas, 15 basal cell carcinomas, and 6 squamous cell carcinomas) were histologically verified. Patient age ranged from 2 to 90 years (median 38), with 279 females (63%) and 164 males (37%). Each image is accompanied by expert-annotated dermatoscopic structures and an explicit verification_stage field indicating the level of diagnostic confirmation. The resulting dataset serves as a pilot clinically verified resource suitable for independent model evaluation, domain shift analysis, interpretability studies, and further expansion.
Authors: Kordel K. France, Ovidiu Daescu
Abstract: Training data for olfaction is scattered through disparate, non-standardized datasets that limit the ability to build representative world models. Olfactory navigation is a highly dynamic and non-stationary task that benefits from real-time continual learning. We introduce an adaptive framework called Grow-Prune-Freeze (GPF) networks that enable an agent to continually learn through growing, pruning, and freezing early layers of its policy in response to world complexity. Grounding GPFs in non-linear random matrix theory, we show that the work of Pennington & Worth (2017) can be extended from single hidden layers to n-layer continual-learning models, and that eigenvalue composition of network weights is preserved as successive layers are added. We show that GPFs based on Expected SARSA achieve a 94% success rate on turbulent plume navigation - a partially observable, non-stationary task representative of the "big world" challenges that motivate adaptive learning in robotics - and provide supporting methodology for applying GPFs in other world models. Further experiments amount evidence that GPFs may generalize well to other machine learning tasks such as reinforcement learning in Atari, image classification, and autoregressive language models. We open source all code and data to encourage improvements on and more research in olfactory robotics.
Authors: Jake Stephen, Niraj K. Jha
Abstract: Knowledge graph (KG) is an abstraction that can be extracted from text corpora and used for in-depth reasoning. Prior work has leveraged KGs to fine-tune language models (LMs), enabling domain-specific superintelligence. In this work, we explore whether KG-driven in-depth reasoning capabilities can emerge in neuroscience using only information contained within a single authoritative textbook. The central hypothesis is that structured knowledge, when distilled into a high-quality KG and converted into KG-grounded question-answer (QA) supervision, is sufficient to produce expert-level reasoning through a fine-tuned LM that surpasses large language models (LLMs) in accuracy, while employing orders of magnitude fewer parameters. We construct a textbook-derived KG via a dual-LLM validation pipeline, expand it with a masked LM trained on the KG topology, generate multi-hop QA items, which include QA pairs and reasoning traces, to fine-tune an LM exclusively on KG-derived supervision, and apply reinforcement learning using path-derived KG signals as implicit reward models. Our results demonstrate that deep, mechanistic neuroscience understanding can be induced in the model without reliance on large, heterogeneous web-scale corpora. The KG-based synthetic neuroscience curriculum that readers can quiz themselves on, and the fine-tuned LM, are available at the following GitHub location: https://kg-bottom-up-superintelligence.github.io/neuro-bench.
URLs: https://kg-bottom-up-superintelligence.github.io/neuro-bench.
Authors: Julius Vernie, Matthias Grabmair
Abstract: Formalizing legal provisions promises machine-accessible law and automated legal reasoning, and recent LLMs make it tempting to generate such formalizations directly from statutory text. However, any formalization makes implicit interpretive choices whose consequences are hard to anticipate, especially if an LLM is the author. We present a method for systematically comparing different formalizations of the same legal provision by their inferences on individual cases. Given multiple formalizations of a provision, we match them at the node level, derive a shared interface for each pair from the matching, and use a SAT solver to enumerate the edge cases on which any two formalizations disagree. Selected edge cases are then verbalized into concrete factual scenarios that a legal expert can examine and act on. We apply our method to formalizations of ten EU provisions generated by nine frontier LLMs. We find that behavioral divergence between formalizations is essentially uncorrelated with their structural agreement and that the verbalized cases reveal qualitatively distinct types of disagreement, including divergences that mirror genuine controversies in the legal commentary.
Authors: Andreas Jungherr, Antonia Schlude, Adrian Rauchfleisch
Abstract: AI-enabled military systems are a fixture of modern military conflict. Applications vary from autonomous drones for surveillance and attack to AI-supported target selection. The importance of AI for modern conflict shows also in public disputes between governments and technology companies over the conditions for military access to frontier AI. Both military uses and government attempts at enabling and steering them happen before a backdrop of public opinion, yet we still know little about how people think about military AI. Drawing on a preregistered survey of 9,000 respondents in nine countries, including China, Germany, and the United States, we examine whether support for military AI is shaped primarily by general attitudes toward AI, principled opposition to lethal autonomy, or foreign-policy and geopolitical orientations. Across six military AI scenarios that vary in lethality and human control, respondents who view AI as beneficial are substantially more supportive of military AI. Hawkish respondents are also more supportive. By contrast, principled opposition to lethal autonomy is not broadly associated with the full index but is related to the application of fully autonomous lethal force. Contrary to our expectation, perceived AI risks are positively associated with support. Cross-national differences are moderate and broadly consistent with geopolitical context. Overall, public opinion toward military AI appears conditionally permissive. Publics are not categorically opposed to various military uses of AI. Instead, unease is concentrated around fully autonomous lethal force.
Authors: Ruitao Liu, Qinghao Hu, Alex Hu, Yecheng Wu, Shang Yang, Luke J. Huang, Zhuoyang Zhang, Han Cai, Song Han
Abstract: Reinforcement learning with verifiable rewards (RLVR) has become a powerful paradigm for improving language models on reasoning-intensive tasks, but its effectiveness is often limited by exploration. For example, models often fail on hard problems, leaving little useful reward signal. External expert traces offer a natural source of guidance, yet they may also expose reward-relevant content along the critical path to the verifier target, such as final answers, intermediate values, executable implementations, or answer-related entities. This content can create an unintended reward hacking channel, allowing the policy to obtain reward by copying the trace rather than learning the underlying reasoning or agentic behavior. Existing guided-RL methods reduce this risk by using partial trajectories, but they mainly control how much expert information is shown heuristically rather than which parts should be hidden. To this end, we propose Semantic Masked Expert Policy Optimization (SMEPO), a fine-grained semantic masking strategy for expert-guided RLVR. Instead of truncating traces coarsely or revealing them unchanged, SMEPO masks reward-relevant semantic spans along the critical path while preserving the expert's decomposition, plan, and procedural structure. This turns hard problems from reasoning from scratch into a fill-in-the-blank process: the policy can follow the expert's problem-solving route, but must still reconstruct the missing values, code, or entities by itself. SMEPO is simple to apply and requires no changes to the reward function or RL objective. Across diverse domains, including math, code, and agentic search, SMEPO improves accuracy by up to 3.2 points over GRPO and reduces training time by up to 4.2x. The code is available at https://github.com/mit-han-lab/SMEPO.
Authors: Gorgi Pavlov
Abstract: We apply the influence-adaptive Walsh geometry of a companion theory paper (arXiv:2605.01637) to extreme low-bit weight-only LLM quantization. The recipe is one math-invariant transformation: WHT-rotate each linear layer's weight matrix and rescale its columns by per-coordinate Walsh-basis activation energy before handing off to a reconstruction-error quantizer (Intel auto-round). This biases per-group integer rounding toward high-spectral-energy channels. On four pretrained decoder-only models from 135M to 1.5B parameters, BBT-spectral reduces wikitext-2 perplexity by 15-58% relative to vanilla auto-round at W2A16; we also report a TinyLlama-1.1B auxiliary data point. Three extensions transfer the recipe to families it failed on: a per-head PCA matrix-Gamma replacement of q_norm/k_norm for Qwen3 attention (PPL 136.76 -> 88.99 on Qwen3-0.6B); an SO(2) per-pair rotation that commutes with RoPE (PPL 36.93 -> 21.84 on Qwen2.5-1.5B); and an MoE-aware input-side absorption fix identified by architectural fuzzing of Laguna-style fused-expert layouts. A W2-vs-W4 ablation gives a deliberate negative control: the redistribution payoff falls within the +/-0.5 PPL noise floor at W4, consistent with the Schur-convexity intuition that the cost of unconcentrated influence vanishes as the noise budget shrinks. All quantized weights export to OpenVINO IR and run on Intel NPU + Arc dGPU + CPU with PPL invariant to device within +/-0.1. We do not claim a formal Boolean-to-real-valued transfer of the theory paper's majorization argument: the WHT activation energy used here is not the Boolean influence of the theory paper, the link is intuitive, and the contribution is engineering value rather than a transferred theorem. Head-to-head benchmarks against SpinQuant, QuaRot, QuIP-sharp, AQLM, OmniQuant, and ButterflyQuant at matched calibration are the main future-work item.
Authors: Ziheng Cheng, Yixiao Huang, Hanlin Zhu, Haoran Geng, Somayeh Sojoudi, Jitendra Malik, Pieter Abbeel, Xin Guo
Abstract: Diffusion models are increasingly used as powerful conditional generators, yet real deployments often involve multiple target distributions arising from different tasks, e.g., diverse prompt domains in text-to-image generation, or multiple environments in robotics with diffusion policies. This naturally leads to a multi-objective learning (MOL) problem. A key challenge is that achieving good Pareto trade-offs can require a generalist model class with substantially larger capacity than what suffices for solving any individual task, thereby increasing statistical cost since sample complexity typically scales with the model complexity. To reconcile this, we develop a principled MOL framework for diffusion models with limited data: a semi-supervised regime where paired (labeled) samples are scarce, but (unlabeled) condition data are abundant. We propose a two-stage training procedure that first fits lightweight specialist models from limited paired data, and then distills them into a generalist model by generating pseudo-samples. We establish generalization bounds showing that the required number of paired samples only depends on the complexity of the specialist model classes. We further extend the theory to diffusion policies for sequential decision making to account for distribution shift in on-policy rollouts. Extensive experiments on robotic control and image restoration tasks are conducted to verify our theoretical results.
Authors: David N. Olivieri, Antonio F. P\'erez Rodr\'iguez
Abstract: Mechanistic interpretability often uses activation patching, causal tracing, path patching, and steering directions to reveal behaviorally meaningful directions in Transformer activation space. This paper develops a field-theoretic framework for organizing and predicting such interventions. Treating the residual stream as a depth-token field, we formulate patching as localized source insertion, patch effects as sensitivity-field predictions, downstream propagation as empirical Green-function response, and patch selection as an adjoint variational problem. Empirically, we test the forward response theory in GPT-2-style autoregressive Transformers by applying localized residual-field interventions and observing the induced residual-field differences and logit-difference responses. We identify a bounded local linear regime; predict patch effects from first-order sensitivities across residual sites; measure structured anisotropic propagation across depth and token position; construct response descriptions from high-sensitivity sites and sliced Green operators; and show that prompt-induced residual displacements can transfer answer behavior. These results establish response objects, namely sensitivities, propagated fields, and Green-operator slices, as a practical language for organizing patching experiments and as the forward mathematical basis for formulating patch-site inference and cross-scale transfer.formulated.
Authors: Oleg Grynets, Vasyl Lyashkevych, Arsen Dolichnyi, Roman Piznak, Taras Zelenyy, Volodymyr Morozov
Abstract: Direct Code2Code transformation remains challenging to control because it can preserve surface-level syntax while introducing semantic drift, hidden behavioral changes, loss of traceability, non-idiomatic target implementations, or incomplete reconstruction of domain logic. This paper proposes a specification-based Code2Text2Code reengineering framework for LLM-mediated software evolution. The central idea is to transform source code into a neutral textual specification that captures program behavior, identifiers, computational flow, conditions, side effects, data dependencies, and domain-specific intent without directly transferring the source language syntax. The proposed framework combines factual context extraction, Code2Text generation, iterative verification between source code and text specification, Text2Code generation, target code verification, retrieval-augmented grounding, and semantic-aware chunking, and transformation loss estimation. The knowledge representation layer integrates metadata derived from AST, graph-based dependency structures, neutral natural language specifications, technical documentation, business documentation, and architecture-level representations. The conducted experiments include a Code2Text2Code dataset built from multiple programming languages and SQL dialects, comparison of intermediate representations, retrieval evaluation, documentation transformation evaluation, and prompt tuning using DSPy. A graph formalization using structural preservation, reverse compatibility, interface stability, and total graph similarity is implemented to estimate transformation losses. The results support the interpretation of the Code2Text2Code approach not as a simple code transformation, but as a controlled specification-based reengineering process for LLM-mediated software evolution.
Authors: David R\"ugamer
Abstract: Epistemic uncertainty is often viewed as a reducible uncertainty that vanishes with increasing data. This perspective implicitly assumes parameter identifiability and equates epistemic uncertainty with predictive variability. In overparametrized neural networks, however, model parameters are typically non-identifiable due to symmetries and redundant representations. As a consequence, substantial parameter uncertainty can persist even when the underlying function is fully identified. In this work, we analyze epistemic uncertainty through the lens of non-identifiability and characterize both discrete and continuous sources of residual uncertainty. Focusing on one-hidden-layer ReLU networks, we thoroughly analyze the resulting posterior structure and validate our theoretical insights through empirical studies.
Authors: Sohaib Lafifi
Abstract: We give an attribution method for neural combinatorial-optimisation (CO) policies that (i) decomposes a decision by constraint families via LP-relaxation duals, (ii) certifies counterfactuals through a combinatorial feasibility model (implemented as a CSP feasibility-decision model), and (iii) bounds the size of a PAC-sufficient explanation with a Bonferroni-corrected Hoeffding sufficient-subset test along a greedy ordering. Across three CO problems and three seeds, our LP-anchored $\Lambda$-attribution matches the CF-derived signal at 96.5% on CVRPTW (n_cert=344) and 77.2% on the Orienteering Problem (n_cert=281) vs 75.0% and 35.2% for proxy gradient (paired diffs +0.215 and +0.420; McNemar exact $p \le 10^{-14}$). In the rank-aligned regime of the Flexible Job-Shop Scheduling Problem, both backends agree on every CSP-certified flip (n_cert=59), confirming the no-gain prediction. Bonferroni-PAC subsets average 5.0 nodes per step ($M=70$, $\varepsilon=\delta=0.2$, $k_{\max}=25$). Reference implementation: https://github.com/sohaibafifi/neuro-co-cax
Authors: Russell Yang, Ruishi Chen, Pierce Kelaita, Riya Ranjan, Sibo Ma, Charles Dickens, Matthew Guillod, Megan Ma, Julian Nyarko
Abstract: Two methodologies dominate current practices of benchmarking: rubric-based scoring evaluates items against predefined criteria, whereas comparative judgment elicits pairwise preferences between outputs. Although both methodologies are widely used, the choice between them is rarely justified. We release JudgmentBench, a benchmark of 30 real-world legal tasks, paired with 1,539 rubric scores and 1,530 pairwise preference judgments collected from practicing attorneys--including at major U.S. law firms--with substantial experience. The annotations constitute the first publicly available dataset in a high-expertise domain in which both supervision signals are elicited from the same experts on the same items. Using LLM-generated outputs at three constructed quality levels, we provide an initial empirical comparison: comparative judgments recover the intended quality ordering substantially better than rubrics (mean Spearman's rank correlation of 0.908 vs. 0.150, estimated difference = 0.758 [0.494, 1.021]) while requiring less than half the annotation time. The patterns hold for human annotators and LLM autograders. Beyond this initial comparison, the paired structure of the dataset supports a broader research agenda on how expert judgment should be elicited, aggregated, and used as supervision in domains without verifiable ground truth.
Authors: Ryo Mitsuhashi, Patrick Chen, Isabelle Tseng, Jasin Cekinmez, Addison J. Wu
Abstract: Reinforcement learning with verifiable rewards (RLVR) has become a standard paradigm for post-training language models, but in practice, verifiers are rarely perfect. Recent theoretical work predicts that verifier noise affects the rate of learning but not its final outcome, implying that sufficient compute should close any gap induced by imperfect supervision. We test this prediction empirically by post-training Qwen2.5 (0.5B, 1.5B) with GRPO on GSM8K while injecting controlled false-positive and false-negative noise into the binary correctness signal, and varying rollouts per prompt as a compute axis. In practice, the gap in validation accuracy persists under substantial compute scaling, with returns to compute that are sharply diminishing. We further find a structural asymmetry where false negatives monotonically degrade performance quicker than with false positives. These findings suggest verifier quality and training compute are not interchangeable, and that reducing false negatives is a more effective lever than scaling compute alone.
Authors: Jasin Cekinmez, Ryo Mitsuhashi, Addison J. Wu, Yida Yin
Abstract: With unified model-generated images now widespread online, attributing their model of origin offers a path toward transparency and deeper insight into the characteristic behaviors of individual models. Prior work has explored provenance in LLM-generated text, diffusion model images, and datasets, but the separability of unified model-generated images remains an underexplored area. We address this gap by examining separability across corruption, domains, and prompt languages using images generated by seven unified models. We show that model attribution is highly feasible as our model achieves near-perfect accuracy with around 20K images per model. Corruptions and structural perturbations have only a modest effect on attribution performance, and cross-domain generalization reveals that semantic content contributes to separability but is not the dominant signal. Finally, we observe that for most models, prompt language attribution is around chance levels, suggesting minimal language-specific visual signatures. These findings highlight consistent model-specific visual characteristics in unified models outputs and open new directions for tracing and auditing generative image pipelines.
Authors: Alberto D\'iaz-\'Alvarez, Ra\'ul Lara-Cabrera, Fernando Ortega-Requena, V\'ictor Ramos-Osuna
Abstract: Recommender systems generally optimises user engagement, but this approach is dangerous in mental health contexts. When vulnerable users show signs of suicidal ideation, standard algorithms often trap them in echo chambers of harmful content, worsening their psychological state. In response, we introduce RankAid, a re-ranking method that prioritises clinical safety alongside predictive relevance. It works as an add-on layer to existing models: it penalises risky items and boosts therapeutic content depending on the user's current level of vulnerability. We evaluated this approach using the MovieLens 1M dataset, where items were semantically annotated for clinical risk and therapeutic value using large language models. Our simulations show that our algorithm successfully blocks the recommendation of harmful content during crisis peaks, actively reshaping the feed to support emotional de-escalation. Furthermore, this safety intervention only causes a controlled, acceptable drop in standard accuracy metrics like NDCG. By using asymmetric hyperparameters, RankAid also gives system administrators the flexibility to tune the severity of the intervention based on specific clinical guidelines.
Authors: Elio Musacchio, Lucia Siciliani, Pierpaolo Basile
Abstract: Current language modeling approaches are built around tokens. Text corpora are split into tokens, and models are trained by performing computations on these tokens, such as predicting the next token given the preceding ones as context. This paradigm has become the standard in modern language modeling, especially given the outstanding performance obtained by token-based architectures. However, recent works have not only begun to question how language models process and understand meaning from tokens, but also to question whether using higher levels of granularity could advance the research field. This led to the idea of Concept Modeling, that is, to directly train models for next-concept prediction rather than next-token prediction. The goal is to change the input from tokens to concepts, forcing the underlying language model to shift its granularity from fine-grained tokens to broad concepts. In this work, we introduce Mimir, a 1.6B Large Concept Model trained for multilingual concept understanding and generation. We leverage a large-scale multilingual pre-training corpus (38,883,987,240 sentences) spanning 46 languages and a large-scale multi-turn and multilingual instruction-tuning dataset (66,816,428 sentences) covering a total of 35 languages. We extensively evaluate model performance against a language model with a comparable number of parameters.
Authors: Minjae Kwon, Amir Moeini, Shangtong Zhang, Lu Feng
Abstract: Safe in-context reinforcement learning (ICRL) adapts online from interaction history without test-time parameter updates while controlling episode cost under a safety budget. Under out-of-distribution (OOD) deployment shifts, pretraining-only safe ICRL can give poor reward-safety tradeoffs because the remaining budget affects behavior only through frozen policy conditioning, not an explicit action-level check against predicted future cost. We propose a latent Q-Barrier shield that learns a context representation, latent dynamics, and an ensemble cost critic before deployment. Without parameter updates, the shield infers context from history and filters or softly reweights candidate actions using the remaining budget and predicted future cost. We prove a conditional, error-decomposed barrier-margin result: a Q-Barrier-satisfying action leaves the next latent-budget state with an approximately budget-safe continuation under the learned critic, up to Bellman and latent-prediction errors. Across five safe ICRL benchmarks, the shield improves deployment-time reward-safety tradeoffs over a strong safe-ICRL baseline: after a short context window, it achieves higher return in four of five benchmarks while matching or lowering average episode cost in all five.
Authors: Gergely B\'erczi, L\'aszl\'o M. Feh\'er
Abstract: We study the symmetric polynomial $\prod_{\alpha\in A_{n,d}}\bigl(1+\alpha_1 x_1+\cdots+\alpha_n x_n\bigr)$ where $A_{n,d}:=\{\alpha\in\mathbb{Z}_{\ge 0}^n:|\alpha|=d\}$, which is the total Chern class of $\mathrm{Sym}^d(\mathbb{C}^n)$, viewed as a torus representation whose Chern roots are the weights $\alpha_1 x_1+\cdots+\alpha_n x_n$ for $\alpha\in A_{n,d}$. Its homogeneous degree-$k$ part $c_k(n,d)$ is the $k$-th Chern class of $\mathrm{Sym}^d(\mathbb{C}^n)$. These Chern classes, together with their coefficients in various symmetric function bases, play a central role in enumerative geometry. Despite their simple definition, general closed formulas for their coefficients are subtle, and many structural properties of these classes have remained poorly understood. In this paper we prove several conjectures concerning their structure, establish explicit formulas, and study log-concavity properties for both the Chern classes and their $K$-theoretic analogue. In rank two, passing to the Schur basis and expanding the Schur coefficients in the binomial basis of $d$, we uncover a new binomial log-concavity phenomenon and prove refined positivity results. The paper demonstrates a novel methodology: we combine several AI systems with human mathematical insight in a coordinated workflow, deploying each tool according to its strengths in experimental discovery, conjecture formation, symbolic proof construction, and verification. To our knowledge, this is one of the first detailed case studies of orchestrating multiple AI tools to make substantial progress on a coherent mathematical research project.
Authors: Pingfan Su, Kai Ye, Shijin Gong, Erhan Xu, Jin Zhu, Giulia Livieri, Chengchun Shi
Abstract: Recent advances in large language models (LLMs) have made it increasingly difficult to distinguish human-written text from AI-generated content. Many existing detectors train supervised neural classifiers that achieve strong in-distribution performance but are often opaque and can degrade substantially under distribution shift. We present READER, a reasoning-enhanced AI text detector that outputs both a human/AI label and a structured rationale describing the evidence for its decision. A key component of our approach is READ, a curated supervision set of rationales and verdicts. We fine-tune an LLM on READ to build READER, which reasons before detecting at inference time. Despite having only 1.5B parameters, READER consistently outperforms existing detectors as well as prompted, high-capacity LLM baselines (GPT-5.2, Gemini-3-Pro, and DeepSeek-V3.2), which are 100 to 1000 times larger in scale.
Authors: Sambit Mohapatra, Senthil Yogamani, Heinrich Gotzig, Patrick Mader
Abstract: Autonomous driving perception demands accurate and efficient processing of three-dimensional sensor data under strict power constraints. Traditional convolutional neural networks achieve strong detection accuracy but are computationally intensive, limiting their suitability for deployment on resource-constrained neuromorphic platforms. Spiking neural networks offer a compelling alternative through event-driven sparse computation, yet their application to complex real-world perception tasks such as three-dimensional object detection remains limited. In this work, we propose an end-to-end spiking encoder-decoder network for object detection in bird's eye view representations of LiDAR point clouds, trained using surrogate gradient backpropagation. We train two variants: a membrane potential variant that reads continuous neuron state at the output stage for maximum accuracy, achieving $92.05$/$87.04$/$86.51$ AP at $\mathrm{IoU}\!=\!0.5$ (Easy/Moderate/Hard), and, a fully binary spiking variant that operates exclusively on spike trains at every layer for direct neuromorphic deployment. We evaluate four input spike encoding strategies and demonstrate that allowing the network to learn spike representations directly from data outperforms hand-crafted Poisson, latency, and z-axis encoding schemes on the KITTI benchmark, where sequential frames are unavailable and the BEV input is presented repeatedly across timesteps as a proxy for temporal streaming. A block-wise energy analysis demonstrates a $3.33\times$ reduction in synaptic operation energy over an equivalent CNN under conservative loop-based operation. Together, these results demonstrate the viability of spiking neural networks for accurate and energy-efficient neuromorphic perception in autonomous driving.
Authors: Hangxuan Li, Renjun Jia, Xuezhang Wu, Yunjie Qian, Zeqi Zheng, Xianling Zhang
Abstract: Effective features are crucial for predictive model performance, but creating them often requires domain expertise, limiting scalability across applications. We define feature engineering as an agentic code generation problem: features are not static data transformations, but executable programs that can be generated, evaluated, and iteratively improved. We present Eureka, an LLM-driven framework with three stages. (1) An Expert Agent, fine-tuned via SFT on domain knowledge, produces structured feature design plans in JSON format. (2) An LLM Feature Factory translates each plan into executable Python code through chain-of-thought reasoning, turning feature hypotheses into runnable programs. (3) A Self-Evolving Alignment Engine uses Reinforcement Learning (GRPO) with dual-channel reward (metric-based utility + semantic alignment) to enhance code quality. By expressing features as programs, the learned generation patterns can transfer across domains. Evaluated on 7 public benchmarks in healthcare, finance, and social domains, Eureka consistently outperforms both traditional AutoFE and LLM-based baselines. We further demonstrate Eureka's effectiveness on cloud GPU resource demand prediction at Alibaba Cloud, where Eureka improves demand fulfillment rate by 16% and lowers computing resource migration rates by 33%.
Authors: Santosh Kumar Radha, Oktay Goktas
Abstract: World models for partially observed environments must imagine multiple compatible hidden futures and steer between them under counterfactual actions. Joint Embedding Predictive Architectures (JEPAs) do this in latent space, but a vector-valued latent has no internal structure for carrying the belief over hidden continuations through blind rollout. We introduce the Unitary World Model JEPA (UWM-JEPA), a JEPA world model with a density-matrix latent on a joint system-environment space and a learned unitary predictor. The construction preserves the joint-state spectrum exactly during rollout, so the predictor itself cannot dissipate the represented uncertainty. On a hidden-velocity indicator task requiring five-step forward simulation under a given action sequence with the target observation masked, UWM-JEPA reaches 0.77 accuracy and degrades monotonically as actions are perturbed; a parameter-matched LSTM-JEPA trained under the same counterfactual-target objective and action head collapses to majority-class accuracy (0.53) under every action condition. Under blind rollout, UWM-JEPA loses fewer than ten points of probe R^2 at short horizons while vector-latent baselines lose forty-one and sixty-eight; both nevertheless tie on a held-out context probe, locating the separation in the predictor rather than the encoder. Action sensitivity itself requires training against counterfactual rather than teacher-forced targets, a finding that applies beyond the unitary parameterisation. For JEPA world models to imagine under partial observability, latent geometry and predictor dynamics matter, not frozen context-encoding capacity alone.
Authors: Akash Bonagiri, Devang Borkar, Gerard Janno Anderias, Setareh Rafatirad, Houman Homayoun
Abstract: Large language model (LLM) agents frequently fail on multi-step tasks involving reasoning, tool use, and environment interaction. While such failures are typically logged or retried heuristically, they contain structured signals about where execution broke down. We introduce CausalFlow, an interventional framework that converts failed agent traces into minimal counterfactual repairs and reusable supervision. CausalFlow models execution traces as sequential chains of dependent steps and computes Causal Responsibility Scores(CRS) via step-level counterfactual intervention to identify failure-inducing steps. For these steps, we generate minimally edited repairs that flip the final outcome to success, producing validated contrastive pairs of the form (wrong step, corrected step). CausalFlow supports two complementary uses: targeted test-time repair that recovers from failures with minimal behavioral drift, and training-time supervision suitable for offline preference optimization or reward modeling. Across four benchmarks spanning mathematical reasoning, code generation, question answering, and medical browsing, CausalFlow converts failed executions into validated minimal repairs with high minimality and causal-consensus scores, and demonstrates that causal attribution is necessary for reliable improvement across diverse agent tasks, outperforming heuristic refinement in complex retrieval settings while producing more localized repairs throughout. These results demonstrate that interventional analysis over structured execution traces provides a principled and scalable mechanism for transforming agent failures into reliability gains and learning-ready supervision.
Authors: Ying Lu, Peng-Fei Zhou, Qi-Xuan Fang, Pan Zhang, Shi-Ju Ran, Gang Su
Abstract: Large language models (LLMs) are dominated by dense linear transformations, whose storage, memory and computational overheads hinder efficient adaptation and deployment while masking the functional impacts of structural simplification. Here we present Tensor Mixture (MixT), a general tensor-structured compression scheme that replaces targeted dense linear layers with natively executable mixtures of tensor operators. Operating directly on generic linear projections instead of model-specific components, MixT is potentially applicable across Transformer-based LLMs and other dense neural mappings. We evaluate MixT on Qwen3-8B and LLaMA2-7B under a unified recovery protocol, identifying a broad compressible regime in which MMLU accuracy is largely preserved before an abrupt transition at model-specific boundaries. This transition coincides with coordinated shifts in output entropy, prediction entropy and inter-layer geometry. At the LLaMA2-7B transition boundary, MixT reduces full-model parameters by 47.5\%, inference FLOPs by 37.1\%, training FLOPs by 52.1\% and peak inference memory by 60.4\%, demonstrating its practical potential for lower-cost LLM compression.
Authors: Keyi Shen, Glen Chou
Abstract: Neural network (NN) dynamics models and control policies achieve strong performance in robotics, but providing sound guarantees under uncertainty remains difficult, especially for closed-loop NN systems. Existing reachability tools provide formal over-approximations, yet are often non-differentiable, overly conservative, or too slow for modern learning and online planning pipelines. To address this, we present a parallelizable, differentiable reachability framework in JAX for continuous- and discrete-time systems with analytical and NN-based dynamics and controllers. Our framework combines Taylor-model flowpipe construction with CROWN-style linear bound propagation through a unified representation that preserves affine dependencies while supporting GPU-batched computation and automatic differentiation. Building on this reachability primitive, we develop (i) a certified training method that encourages reachability-friendly dynamics models and controllers, and (ii) a reachability-aware sampling-based MPC scheme with gradient-based refinement. Experiments on non-prehensile manipulation and quadrotor tasks, including hardware and higher-dimensional evaluations (up to 72D), demonstrate practical online planning while maintaining certified reachable-set over-approximations under bounded uncertainty.
Authors: Veera Varuni Radhakrishnan, Chinthaka Dinesh, Qurat-ul-Ain Azim
Abstract: Low-dose computed tomography (LDCT) reconstruction faces a critical tradeoff between reconstruction quality and resource requirements. While recent deep learning methods achieve state-of-the-art performance, they typically rely on over 500,000 parameters trained on large-scale datasets exceeding 35,000 scans. This work investigates whether graph-based regularization can provide meaningful noise reduction under strict resource constraints. We propose Deep Graph Laplacian Regularization (Deep GLR), integrating quadratic graph regularization into a Proximal Forward-Backward Splitting optimization framework with three lightweight CNN modules. Evaluated on the LoDoPaB-CT benchmark, Deep GLR achieves 30.70 dB PSNR, representing a 6.33 dB improvement over filtered backprojection, while using only 91,848 parameters trained on 1000 samples (2.8\% of standard training set). Compared to benchmark methods, this represents 5.8 times better parameter efficiency and 30 times better data efficiency per dB improvement. The learned graph bandwidth parameter ($\epsilon$=1.25) converges to interpretable values, suggesting the method captures meaningful image priors rather than overfitting. While a 13 dB gap remains versus state-of-the-art methods, results demonstrate that graph-based regularization provides a favorable efficiency-quality tradeoff for resource-constrained medical imaging scenarios.
Authors: Konstantinos Emmanouilidis, Tianjiao Ding, Nghia Nguyen, Nicolas Loizou, Ren\'e Vidal
Abstract: Deep learning models are vulnerable to adversarial perturbations, raising important concerns for safety-critical deployment. Empirical defenses can achieve strong robustness in practice, but lack formal guarantees, motivating the need for certifiably robust classifiers. While certified methods provide formal guarantees, they often yield overly conservative bounds due to their inability to exploit structure in complex data distributions. In this work, we propose a framework for designing certifiably robust classifiers that leverages latent structure in data representations. We first analyze the Gaussian mixture setting, deriving necessary and sufficient conditions for the existence of robust classifiers and constructing a classifier with a closed-form robustness certificate and generalization guarantees. Our main contribution is to show that exact structure is not required: we prove that if a pretrained encoder maps inputs to a latent distribution that is $\varepsilon$-close (in KL divergence) to a Gaussian mixture, then certified accuracy degrades gracefully, with an explicit bound relating robustness under the true and approximate distributions. This result enables the direct use of pretrained models without requiring exact distributional assumptions. Empirically, our method achieves state-of-the-art or competitive certified accuracy on CIFAR-10 and ImageNet, while maintaining strong clean performance and low computational overhead. Overall, our work establishes approximate latent structure as a practical and principled route to certifiable robustness.
Authors: Thomas Stephan Juzek
Abstract: AI-associated lexical shifts have been documented mainly in Scientific English. We extend this work to 34 languages in the WMT News Crawl corpus, refining a split-halves continuation diagnostic that compares GPT-4.1 continuations with matched human gold-standard text. For each language, we derive ranked AI-overused lemmas using log prevalence ratios. We find substantial cross-lingual semantic convergence: semantically related concepts recur across typologically diverse languages, with 'emphasize'-type verbs appearing in 24 of 34 languages. Embedding-based and manual analyses support this pattern. We also examine diachronic uptake in news writing before and after ChatGPT's release. Tracking each language's top 20 AI-overused items, we find prevalence increases in 26 of 34 languages from 2020-2021 to 2023-2024, with a mean change of +15.1%, whilst matched baseline words show no comparable increase (-4.5%). In 10 languages with longer historical coverage, longitudinal analyses show post-2022 increases that exceed the modest shifts observed in earlier periods, though with smaller effect sizes than in Scientific English. We validate our approach extensively, including across seeds, model variants, data sizes, model families, and more. Our findings are consistent with the view that AI-associated lexical preferences extend beyond English and may exert cross-lingual homogenising pressure on global language use.
Authors: Kolawole Quadri
Abstract: Observability tells operators when an agent is slow. KYA tells operators when an agent is wrong, drifting, leaking, or quietly going rogue. We present KYA (Know Your Agents), an open-source trust and governance layer for autonomous systems composed of five primitives: (1) a four-gate inbound apply pipeline composing Ed25519 signature verification with multi-anchor pinning, persist-time expiry, only-tighten composition, and operator-approval-as-default; (2) an only-tighten composition algebra over a three-channel multi-tenant hierarchy (platform default,tenant override, signed external recommendation); (3) KYP -- Know Your Principal, a schema-level unification of trust scoring across human users, AI agents, and service accounts; (4) auditable interaction-multiplier amplification over an AIVSS-shaped additive baseline, with bounded asymmetric per-interaction multipliers carrying stable audit codes; and (5) two-axis delegation attribution combining a static observation-gated delegation-trust premium with zero-config runtime orchestrator-blame at three SDK hook surfaces. KYA is framework-agnostic across 22 agent frameworks. The pure-function scorer runs sub-millisecond at p99 and the system sustains ~1,800 ops/sec at 20 concurrent workers with HMAC chain integrity preserved end-to-end. The four-gate inbound apply pipeline rejects forged, expired, loosening, and unapproved recommendations on every trial (1,200 / 1,200) with sub-millisecond p99 latency on SQLite. KYA detects 89% of 1,200 adversarial probes from PyRIT and Garak, including the recently-published topology-guided multi-agent attack. The system is available under Apache 2.0 as the veldt-kya package on PyPI (release candidate at submission time; stable v0.1.0 forthcoming)
Authors: Ruoxi Cheng, Haoxuan Ma, Zhengfei Hai, Yiyan Huang, Ranjie Duan, Tianle Zhang, Xu Yang, Ziyi Ye, Xingjun Ma
Abstract: Large Vision-Language Models (LVLMs) have advanced multimodal understanding, yet their reliability is limited by hallucination, where generated content conflicts with visual facts. Existing mitigation methods either rely on costly external interventions, such as instruction tuning and retrieval, or use internal mechanisms that remain limited by flawed attention weights and entangled hidden representations. We propose Adversarial Orthogonal Disentanglement (AOD), a latent geometric framework for mitigating LVLM hallucinations. AOD learns a hallucination-related direction through a minimax objective: a classifier concentrates hallucination signals into the projected component, while an adversary removes them from the orthogonal residual space via a Gradient Reversal Layer. The learned direction enables a training-free dual-forward-pass contrastive decoding strategy that suppresses hallucinations while preserving general capabilities. Experiments on three LVLMs across four hallucination and four utility benchmarks show that AOD consistently outperforms strong baselines. It improves POPE accuracy by over 6\% on average, boosts AMBER by 6\%, and maintains strong performance on utility tasks such as MMMU. Further analysis shows robust transfer across datasets, suggesting that AOD captures general hallucination-related biases rather than dataset-specific artifacts. Our source code and datasets are available at https://github.com/Hunter-Wrynn/AOD.
Authors: Fangtai Wu, Hailong Guo, Shijie Huang, Jiayi Song, Yubo Huang, Mushui Liu, Zhao Wang, Yunlong Yu, Jiaming Liu, Ruihua Huang
Abstract: Customized image editing aims to equip pre-trained diffusion models with specific visual effects using limited paired data, typically via Low-Rank Adaptation (LoRA). As the number of desired effects grows, storing and dynamically loading numerous these effect LoRAs significantly increases deployment overhead. Furthermore, current pipelines typically cascade these effect LoRAs with acceleration modules for fast generation, which triggers severe parameter interference and results in concept bleeding and style degradation. We propose CollectionLoRA, a multi-teacher on-policy distillation framework capable of distilling the concepts of up to 50 different effect LoRAs along with few-step generation capabilities into a single LoRA. This fundamentally resolves the feature interference issue and significantly reduces deployment costs. Specifically, the method introduces (i) a Probabilistic Dual-Stream Routing mechanism that enables the model to randomly switch between data sources during training, effectively enhancing its generalization in unseen scenarios; (ii) an Asymmetric Orthogonal Prompting strategy to achieve concept isolation within the prompt space; (iii) a Coarse-to-Fine Distillation Objective to mitigate the distribution gap between the teacher and student models. Extensive evaluations show that CollectionLoRA distills all customized effects and few-step generation into a single LoRA, reducing deployment overhead while achieving concept fidelity comparable to or better than independently trained teacher models.
Authors: Xia Li, Xinran Liu, Lin Qi, Junyu Dong
Abstract: Camouflaged object detection (COD) from a single image is a challenging task due to the high similarity between objects and their surroundings. Existing fully supervised methods require labor-intensive pixel-level annotations, making weakly supervised methods a viable compromise that balances accuracy and annotation efficiency. However, weakly supervised methods often experience performance degradation due to the use of coarse annotations. In this paper, we introduce a new weakly supervised approach for camouflaged object detection to overcome these limitations. Specifically, we propose a novel network, MGNet, which tackles edge ambiguity and missed detections by utilizing initial masks generated by our custom-designed Cascaded Mask Decoder (CMD) to guide the segmentation process and enhance edge predictions. We introduce a Context Enhancement Module(CEM) to reduce the missing detection, and a Mask-guided Feature Aggregation Module (MFAM) for effective feature aggregation. For the weak supervision challenge, we propose BoxSAM, which leverages the Segment Anything Model (SAM) with bounding-box prompts to generate pseudo-labels. By employing a redundant processing strategy, high quality pixel-level pseudo-labels are provided for training MGNet. Extensive experiments demonstrate that our method delivers competitive performance against current state-of-the-art methods.
Authors: Bingyu Yan, Xiaoming Zhang, Jinyu Hou, Chaozhuo Li, Ziyi Zhou, Yiming Hei, Litian Zhang
Abstract: While Large Language Model-based Multi-Agent Systems (LLM-MAS) demonstrate remarkable capabilities in solving complex tasks by orchestrating specialized agents and external tools, the implicit trust in tool outputs creates a critical attack surface. Existing tool attacks are limited by domain specificity or fixed and static templates. To address these challenges, we propose Evo-Attacker, which formulates the tool attack as a self-evolving, memory-augmented reinforcement learning process. Evo-Attacker constructs a dynamic attack memory and employs deliberative reasoning to retrieve adversarial patterns and strategize modifying interventions at critical moments. Furthermore, we introduce Attack-Flow GRPO to optimize intermediate reasoning steps via terminal outcomes, addressing the long-horizon credit assignment challenge. Comprehensive experiments demonstrate that Evo-Attacker consistently outperforms baselines, highlighting its generalization and evolutionary capabilities and the urgent need for defensive tool safeguards.
Authors: Chunzheng Zhu, Jianxin Lin, Feng Wang, Cheng Jiang, Guanghua Tan, Zhenyu Zhou, Shengli Li, Kenli Li
Abstract: Reliable quality control (QC) of ultrasound images is essential for both real-time acquisition guidance and retrospective clinical audit, yet existing approaches rely heavily on per-plane annotations, or employ pseudo-labeling prone to systematic bias under spatial deformations inherent in clinical acquisition. We present STRIQ, a registration-driven framework that recasts annotation-free US plane quality control as a subspace-guided consistency measurement problem. Specifically, STRIQ introduces a Latent Registration Aligner (LRA) to establish hierarchical feature space correspondences between query images and variance-driven anchors, which are autonomously distilled from unlabeled data via a variance spectrum criterion to serve as structurally stable prototypes. To further disambiguate anatomical planes and mitigate negative knowledge transfer, we propose an Orthogonal Knowledge Subspace (OKS) module. The OKS decomposes plane-specific representations into mutually orthogonal subspaces, enabling fine-grained expert collaboration while preventing inter-plane interference, ensuring that the quality metric is grounded in principled subspace proximity. Extensive experiments on the in-house US4QA and public CAMUS datasets demonstrate that STRIQ achieves state-of-the-art correlation with clinical quality scores, establishing a new paradigm for annotation-free, real-time reliable ultrasound quality control. Our code is available at https://github.com/zhcz328/STRIQ.
Authors: Chunzheng Zhu, Yijun Wang, Jianxin Lin, Feng Wang, Hongwei Wang, Lei Zhao, Shengli Li, Kenli Li
Abstract: Self-supervised pre-training paradigm has gained increasing prominence for learning transferable representations in medical imaging, yet existing methods for ultrasound (US) images operate at the image or frame level, overlooking the anatomical context for clinical-aligned representation learning. In this work, we propose an anatomy-anchored ultrasound self-supervision framework ANAUS that shifts representation learning from generic visual regions to clinically meaningful anatomical structures. Utilizing a learnable latent prompt engine alongside a one-time domain adaptation on existing public image--mask pairs, we empower the LP-SAM module to achieve annotation-free anatomy delineation at scale. Building upon this anatomical grounding, we propose a dual-policy self-supervised learning paradigm consisting of inter-view semantics-aware anatomy-separating alignment and contextual core-region prediction to enhance representation learning. Specifically, the former enforces feature invariance within identical anatomical regions while promoting discriminability across distinct structures; the latter compels the model to reconstruct corrupted regions, thereby capturing fine-grained structural details. Extensive evaluations on six public datasets demonstrate that \ours{} consistently outstrips current state-of-the-art methods while maintaining the computational efficiency essential for clinical deployment. Code is available at https://github.com/zhcz328/ANAUS.
Authors: Jiaquan Zhang, Caiyan Qin, Haoyu Bian, Libin Cai, Yi Lu, Chaoning Zhang, Wei Dong, Yuanfang Guo, Yang Yang, Hen Tao Shen
Abstract: Neural operators learn mappings from function-dependent inputs to solutions, providing an effective framework for solving partial differential equations (PDEs). For time-dependent PDEs, existing methods typically perform long-horizon prediction through autoregressive rollout directly in high-dimensional physical field spaces, where each predicted state is recursively fed back as the input for the next step. Although effective for short-term prediction, this autoregressive rollout and the lack of continuous-time modeling lead to progressive error accumulation over long-horizon rollouts. In this work, we propose Autoregression-Free Neural Operators (AFNO), which map the time evolution of PDEs into a latent space and model continuous-time vector fields within it. AFNO uses flow matching to learn the latent vector field, thereby enabling continuous evolution over extended horizons, avoiding autoregressive rollout and capturing dynamics under varying parameter configurations through explicit conditioning on physical parameters. Theoretical analysis and extensive experiments on six PDEs demonstrate that AFNO improves long-horizon prediction stability and consistently reduces rollout errors compared with the baselines.
Authors: Khalid Yusuf Dahir
Abstract: Large language model safety evaluation remains heavily English-centered, leaving low-resource languages under-measured even when models are deployed globally. We evaluate four open-weight instruction-tuned models on SomaliBench v0, a native-author-verified benchmark of 100 harmful-intent prompts paired across English and Somali. Each of Llama-3.1-8B-Instruct, Gemma-2-9B-Instruct, Qwen-2.5-7B-Instruct, and Aya-23-8B is run locally with temperature 0 and the same English "helpful, harmless, and honest" (HHH) system prompt. A pinned Claude Sonnet snapshot (claude-sonnet-4-5-20250929) classifies each response as refused, complied, or unclear; the native author spot-checks a stratified 80-row sample. We find large English-to-Somali refusal gaps for all four models: Llama-3.1-8B (0.90; 95% bootstrap CI [0.85, 0.96]), Aya-23-8B (0.75 [0.67, 0.83]), Qwen-2.5-7B (0.69 [0.59, 0.78]), and Gemma-2-9B (0.38 [0.27, 0.49]). For three models, the dominant Somali non-refusal mode is not fluent harmful compliance but unclear output: empty, wrong-language, or incoherent generations. The native verification spot-check achieves 100% agreement with the judge (Cohen's kappa = 1.00) on the 80 sampled rows. We report aggregate refusal rates, category gaps, and reliability statistics only; raw model generations are retained locally and are not released.
Authors: Lipeng Dai, Luping Xiang, Kun Yang
Abstract: The convergence of large language models (LLMs) with 6G networks is fostering a paradigm of autonomous multi-agent cooperation, which in turn is expected to substantially increase east-west traffic. Although latent-space interaction mechanisms can enable more efficient collaboration than symbolic natural-language (NL) exchanges, prior work often abstracts away the associated communication overhead under practical wireless constraints. In embodied multi-agent settings, heterogeneous interaction media incur disparate inference and transmission costs, thereby inducing an inherent end-to-end (E2E) latency trade-off. To address this, we propose a joint design that integrates communication-media selection with wireless resource allocation. Through analytical characterization and simulation-based evaluation, we show that neither token-based transmission nor key-value (KV) cache-based transmission is uniformly optimal across operating regimes, as performance depends critically on system parameters such as available computational resources and channel conditions. Accordingly, we formulate a joint optimization problem aimed at minimizing the E2E latency of multi-agent collaboration and develop a low-complexity joint media selection and resource allocation (JMSRA) algorithm. Numerical results further confirm that, by adaptively coordinating the interaction media and bandwidth allocation over heterogeneous links, the proposed scheme achieves markedly reduced E2E latency relative to conventional NL-only and KV-cache-only baselines, enabling efficient and robust multi-agent collaboration in future wireless networks.
Authors: Zhongling Xu, Shunan Zheng, Wei Wang
Abstract: Existing LLM routing frameworks treat queries as independent events, neglecting the sequential nature of real-world user sessions constrained by global computational budgets. This mismatch inevitably leads to budget bankruptcy: myopic routing policies exhaust resources on early interactions, forcing subsequent and often more complex queries onto inadequate models. We introduce SeqRoute, a framework that formulates multi-turn routing as a finite-horizon Markov Decision Process and solves it via offline reinforcement learning. By incorporating the remaining budget into the state space and training with Conservative Q-Learning (CQL), SeqRoute learns delayed gratification to strategically preserve resources for high-stakes turns later in the session. To overcome data starvation, we propose Hindsight Budget Relabeling (HBR). This technique retrospectively simulates historical trajectories under diverse hypothetical budgets, expanding 10,000 raw sessions into 2.38 million transitions enriched with critical bankruptcy signals. At deployment, a dynamic $\lambda$-sweep mechanism enables zero-shot navigation of the cost-quality Pareto frontier without retraining. Extensive evaluations demonstrate that SeqRoute reduces operational costs by 6.0-73.5% while maintaining or improving quality, and suppresses bankruptcy rates to under 1%, strictly dominating behavior cloning, budget-aware heuristics, and static baselines across the entire Pareto frontier.
Authors: Udith Haputhanthri, Declan Campbell, Rim Assouel, Jonathan D. Cohen, Taylor W. Webb
Abstract: Despite success on standard benchmarks, vision language models display persistent failures on tasks involving processing of multi-object scenes, including many tasks that are relatively easy for humans. Recent work has found that these failures may stem from a basic inability to accurately bind object features in-context, a challenge that is referred to as the "binding problem" in cognitive science and neuroscience. The human visual system is thought to solve this binding problem via serial processing, attending to individual objects one at a time so as to avoid interference from other objects. Recent work has proposed "pointing" -- the use of explicit spatial coordinates to refer to objects -- as an analogous solution for vision language models, and found that it improves performance on challenging multi-object tasks. However, it is unclear $\textit{why}$ (i.e., on a mechanistic or representational level) this approach improves performance, and how directly this relates to serial processing in human vision. Here, we investigate this question. We find that learning to point-via-text induces an internal visual search routine, and we characterize the mechanisms that support this procedure. We also find that pointing behavior can be generalized to new tasks via fine-tuning, and that doing so eliminates binding errors and enables compositional generalization. These results provide a proof-of-principle that serial processing can solve the binding problem for vision language models just as it does for biological vision.
Authors: Rafal Kocielnik, J. Everett Knudsen, Steven Y. Cen, Jasmine Lin, Cherine H. Yang, Atharva Deo, Ujjwal Pasupulety, Peter Wager, Anima Anandkumar, Andrew J. Hung
Abstract: Verbal feedback delivered by attending surgeons in the operating room plays a critical formative role in resident trainee skill acquisition. Yet, assessing the quality of trainer feedback and its effectiveness in influencing trainee behavior during live surgery remains a challenge. Prior studies assessed feedback content relying on extensive manual annotation by expert human raters and focused on developing broad taxonomies that overlook the qualitative aspects of feedback delivery such as clarity or urgency. Limited existing automated methods, including keyword analysis and topic modeling, also fail to capture these nuanced aspects. We introduce a two-stage LLM-based framework that discovers interpretable feedback quality criteria grounded in the context of surgical training. Our method uses multi-agent prompting and surgical domain knowledge injection to discover a small set of human interpretable scoring criteria (e.g., Encouraging, Urgent, Clear). These criteria are then used to automatically score live surgical feedback via an LLM-as-a-judge approach. Evaluation on 4.2k trainer feedback instances demonstrates that our AI-discovered criteria outperform prior content-based frameworks in predicting feedback effectiveness, including observed trainee behavioral adjustments and trainer approval. This work advances scalable, human-aligned assessment of communication quality in the operating room and provides a foundation for improving surgical teaching practices.
Authors: Jiwon Kim, Claire Wang, Taeung Yoon, Sabelle Huang, Koustuv Saha
Abstract: Large language models (LLMs) are increasingly being used for emotional support. They are also being developed for formal therapy purposes. However, LLMs like ChaptGPT or Llama are often developed with content moderation guardrails that prevent them from discussing sensitive subjects with users for both liability and safety purposes, and this inability to broach these subjects may affect their capacity as therapists. In this study, we perform an algorithm audit on three state-of-the-art moderation systems (OpenAI's moderation endpoint, Meta's Llama Guard, and Google's Shield Gemma) to investigate the extent to which these systems flag the content of real-life therapy sessions as undesirable. Our results raise implications for the limitations that users and organizations may encounter when designing LLMs to play the part of a therapist.
Authors: Asvin G., Jack Lindsey
Abstract: Language models are pretrained as passive predictors with no incentive to model the consequences of their own outputs. Post-training changes this: a model producing its own responses can benefit from recognizing that it is on-policy. We present evidence that post-trained models recognize their on-policy generations, and this recognition is implicitly encoded in their output distributions. In particular, on-policy output distribution entropy is 3--4$\times$ lower than off-policy entropy, across model families and size classes. We trace part of this effect to an internal representation of input surprise, tracking the unlikeliness of the most recent input token according to the model's prior predictions, that causally modulates output entropy. One example of these phenomena can be observed in response to open-ended prompts; post-trained models (unlike pretrained models) collapse their uncertainty over the topic of their upcoming response before the first output token; violating this cached intention with a different-topic prefill results in higher output entropy. We also tested whether models can distinguish on-policy contexts from prefills via explicit verbal report. We find that they can, but that interestingly, this explicit recognition routes through a different mechanism than implicit recognition.
Authors: Xintong Yang, Hao Gu, Binxing Xu, Lujun Li, Bei Liu, Jiacheng Liu, Qiyuan Zhu, Sirui Han, Yike Guo
Abstract: Large Language Models (LLMs) are increasingly expected to operate over long contexts, yet standard softmax attention incurs a KV cache that grows linearly with sequence length, quickly becoming the bottleneck for long context inference. A practical remedy is to evict less important KV entries; however, existing eviction policies are largely heuristic and struggle to capture the rich, input-dependent distribution of token importance. In this work, we introduce a learnable indexer that predicts KV importance, enabling more accurate retention of critical tokens. Meanwhile, naively evicting tokens permanently discards their information, leading to irreversible forgetting and degraded retrieval over long ranges. To address this, we propose a lightweight latent memory module that compresses evicted tokens into a compact, online-updated state and provides residual readouts to compensate for the attention contributions lost through KV eviction. Collectively, our method enables accurate long-context inference under a bounded KV budget, delivering consistent improvements on RULER (4K/16K) across Qwen, Mistral, and Llama models (up to 25 points under aggressive eviction), markedly more stable Needle-in-a-Haystack retrieval, and superior LongBench scores and compression curves compared to existing eviction policies.
Authors: Perry Dong, Kuo-Han Hung, Tian Gao, Dorsa Sadigh, Chelsea Finn
Abstract: The ability to efficiently and reliably learn new tasks has been a foundational challenge in robotics. Vision-Language-Action (VLA) models have demonstrated strong generalization across diverse manipulation tasks, yet pretrained policies consistently fall short of the reliability required for real-world deployment. Reinforcement learning (RL) fine-tuning offers a promising path to bridge this gap, but existing approaches either train from scratch without fully leveraging pretrained priors, or fine-tune VLAs without achieving the sample efficiency and success rates that practical deployment demands. We present EXPO-FT, a system for stable, sample-efficient RL finetuning of pretrained VLA policies that closes this gap. Our system solves a suite of challenging manipulation tasks, including routing string lights and inserting the plug to light it up, striking a pool ball into a pocket, and inserting a flower into a wine bottle, each requiring combinations of high precision, dynamic actions, and robustness to varied initial states. Our system achieves perfect task performance (30/30 successes) across all evaluated tasks within an average of 19.1 minutes of online robot data, outperforming both prior RL-from-scratch and VLA finetuning approaches. We release an open-source codebase with the aim of facilitating broader adoption of RL finetuning of VLA models in robotics.
Authors: Zhicheng Zhang, Lei Wang, Yu Zhang, Yongsheng Gao
Abstract: Audio-driven talking-head generation has achieved remarkable progress with recent models such as AniTalker, FLOAT, and Sonic. Despite their success, most existing approaches rely on a single static reference image to condition the entire video generation process at inference stage. This static conditioning paradigm often creates a mismatch between fixed identity features and dynamically evolving facial motion, leading to identity drift, temporal inconsistency, and degraded perceptual quality. We introduce Test-Time Self-Adaptive Conditioning (TT-SAC), a parameter-free inference framework that enables pretrained talking-head generators to adapt their conditioning representations during inference without retraining, gradient updates, or additional supervision. Instead of treating the reference portrait as immutable, TT-SAC composes the generator with its encoder in a feedback loop: the generator's own outputs are re-encoded to construct a refined conditioning representation that better aligns with the temporal dynamics of the synthesized sequence. A single adaptation step approximates a self-consistent equilibrium of the generative process, stabilizing identity and motion across time. We further provide theoretical analysis showing that test-time conditioning adaptation reduces feature variance and improves generative stability under mild Lipschitz assumptions, while exhibiting a principled bias-variance tradeoff that governs the optimal strength of adaptation. Extensive experiments on state-of-the-art talking-head generators and benchmark datasets demonstrate consistent improvements in lip-sync accuracy, temporal coherence, identity preservation, and perceptual fidelity. TT-SAC offers a model-agnostic and training-free strategy for enhancing generative video models, establishing test-time conditioning adaptation as an effective mechanism for stabilizing audio-driven portrait animation.
Authors: Yehudit Aperstein, Alexander Apartsin
Abstract: Educational aspect-based sentiment analysis (ABSA) can support course improvement, but public aspect-labeled student feedback remains scarce because educational reviews are private, institution-specific, and expensive to annotate. This study introduces a controlled synthetic benchmark for educational ABSA built from 10,000 synthetic course reviews with explicit train-validation-test splits and a 20-aspect pedagogical schema spanning instructional quality, assessment and course management, learning demand, learning environment, and engagement. The corpus is generated with sampled target labels, sampled nuance attributes, and a realism-tuned prompt refined through a three-cycle judge-editor procedure. On the resulting benchmark, local baselines with TF-IDF, two-step transformers, and joint encoders show that the task is nontrivial; the strongest untuned model, BERT, reaches a held-out detection micro-F1 of 0.2760, while a modest lower-rate BERT schedule improves this to 0.2930. Full-test GPT-based inference with gpt-5.2 reaches 0.2519 micro-F1 in zero-shot mode and 0.2501 with retrieval-based few-shot prompting, placing batch inference above the classical baseline and close to the compact joint encoders. A conservative external evaluation on 2,829 mapped student-feedback reviews from Herath et al. yields a micro-F1 of 0.4593 for BERT on a 9-aspect overlap, indicating partial synthetic-to-real transfer. Realism and faithfulness analyses are reported as generator diagnostics that clarify how the benchmark was stabilized and where label noise remains. The study therefore contributes a synthetic educational ABSA corpus, a documented generation procedure, and a reproducible benchmark setting for a domain in which public labeled data remain difficult to obtain.
Authors: Xiliu He, Haoxiang Zhao, Mingyi Ma, Edward Wen Chuan Lai, Koei Enomoto, Anni Hu, Jiatong Li, Lingyun Chu, Yuan Lai
Abstract: Generative artificial intelligence (GenAI) is the first automation wave to reach high-cognitive tasks at scale, yet its effects on intra-urban inequality remain largely unknown. Using 5 million job postings from Beijing (2018--2024), we construct a neighborhood-level GenAI Exposure Index by aggregating task-level assessments from five leading large language models. We examine the spatial, structural and causal mechanisms of this shock. We find that GenAI exposure is highly concentrated in the city's core districts, deepening the intra-urban AI divide. Since 2023, high-exposure neighborhoods have experienced wage stagnation even as they continue to attract high-skilled workers -- a "high-skill trap." This wage penalty is driven by task de-skilling and intensified labor-market crowding. A difference-in-differences design centered on ChatGPT's release supports a causal interpretation. These findings challenge the prevailing theory of skill-biased technological change and provide a basis for inclusive AI governance in global technology hubs.
Authors: Xinyang Zhai, Chong Yang, Ruizhi Zhang
Abstract: Breast ultrasound imaging is an important noninvasive method for early breast cancer diagnosis, but automatic benign/malignant classification remains challenging due to tumor heterogeneity, blurred boundaries, and data imbalance. To improve feature representation and classification accuracy, this paper proposes the Cross-Stage Attention Mixture-of-Experts Network (CSA-MoE-Net). It adopts a Cross-Stage Attention-enhanced ResNet-18 as the backbone, in which the Cross-Stage Attention module adaptively recalibrates multi-level features, thereby enhancing key tumor features and suppressing redundancy. A three-branch Mixture of Experts (MoE) Block learns complementary features from the Whole Tumor Image, Tumor Core, and Boundary, and an Adaptive Gating Network fuses them to capture morphological, textural, and contextual information. The fused features are denoted as Fused Expert Feature (FEF) in the architecture. Experiments on a balanced dataset of 2,129 breast ultrasound images show that, averaged over 20 independent runs, the model achieves an accuracy of 96.33\%, precision of 94.09\%, recall of 98.53\%, F1-score of 96.25\%, and AUC of 99.50\%. Compared to the baseline ResNet-18, these metrics improve by 3.01, 0.70, 5.37, 2.98, and 5.42 percentage points, respectively. The proposed mechanism requires no invasive modification and can be seamlessly embedded into VGG-16, DenseNet-121, etc., yielding stable performance gains, thus providing reliable support for computer-aided diagnosis.
Authors: Muslim Chochlov, Michael English, Jim Buckley
Abstract: Context. Large language models (LLMs) are increasingly applied to code-generating tasks (CGTs) in software engineering. While reported results are promising, the broader effects of such application and their integration into real-world development remain insufficiently understood with existing tertiary studies provide little in this area. Objective. This tertiary study consolidates secondary evidence on LLM-based CGTs, synthesizing the publication landscape, effects, scenarios, integration challenges, and future research directions. Method. Following systematic review guidelines, we searched in related digital libraries, complemented by backward-and-forward snowballing and screening step. Study quality was assessed and extraction reliability was audited with inter-rater agreement statistics. Evidence was synthesized using SWEBOK knowledge areas and the HELM framework. Results. We identify 30 secondary studies published between 2017-2025, with rapid growth since 2023. Accuracy seems strong on benchmarks but weakly supported for real-world generalization; robustness is fragile across tasks and configurations; efficiency constraints are pervasive; toxicity and bias are under-reported. Dominant challenges concern economic feasibility, evaluation validity, and socio-technical integration. Future directions suggest domain-aware model improvement and the need for holistic, standardized evaluation. Conclusion. LLM-based CGTs represent a fast-maturing yet unevenly evaluated research area, highlighting the need for domain-aware model improvements and holistic, standardized evaluation, addressing efficiency and associated costs.
Authors: Xinyuan Yan, Rita Sevastjanova, Mennatallah El-Assady, Bei Wang
Abstract: Neural networks encode inputs as high-dimensional vectors, known as representations, that capture how models process data by encoding task-relevant structure and semantics. Representation alignment refers to the degree to which different models, layers, or training conditions produce similar representations for the same inputs, with important implications for model interpretation, selection, and robustness analysis. Existing approaches to measure alignment primarily rely on geometric properties, such as neighborhood and cluster similarity, offering limited insight into the global organization of representations. In this work, we present TopoAlign, a topology-aware framework for visually comparing model representations from a structural perspective. Leveraging mapper graphs from topological data analysis, TopoAlign jointly analyzes graphs constructed from representations of shared inputs across different models or layers. The framework supports a top-down comparative workflow: it first performs global structure alignment via joint force-directed optimization to produce coordinated graph layouts; it then identifies local correspondences through automated detection of structurally matching regions, visualized with Bubble Sets; and finally it enables fine-grained pattern inspection through motif-based queries and membrane-inspired visualizations. We demonstrate TopoAlign through case studies on language and multimodal models, complemented by expert feedback. Our results show that TopoAlign provides meaningful insights into representation structure and alignment from a topological perspective.
Authors: Shubhajit Roy, Anirban Dasgupta
Abstract: Dynamic graph neural networks (DGNNs) that operate on snapshot sequences typically fall into one of two categories. \emph{Temporal-first} approaches build per-node temporal embeddings and only afterwards perform spatial aggregation, whereas \emph{Spatial-first} approaches invert this order, feeding the output of a graph convolution into a downstream temporal module. In either case, the rigid sequencing forces the second stage to consume an already-compressed summary produced by the first, ruling out joint reasoning over topology and evolution; concretely, the message-passing operator never gets to weight a neighbor's contribution by that neighbor's \emph{past} trajectory. This paper introduces \textbf{SiST-GNN} (\textbf{Si}multaneous \textbf{S}patial-\textbf{T}emporal \textbf{GNN}), which fuses the two signals inside a single message-passing operation rather than chaining them. Concretely, at each snapshot we maintain a recurrent hidden state per node that summarises its history, pair it with the node's current feature vector, and treat the pair as two nodes joined by a cross-time edge; running a standard graph convolution on this temporally augmented graph yields the updated representation. Our empirical study spans nine public baselines and fourteen model-dataset combinations, covering both fixed-split and live-update evaluation regimes. Across every public benchmark, SiST-GNN sets a new state of the art in link prediction task over the strongest prior method by $109$--$277\%$ in the fixed-split setting and by $68$--$194\%$ in the live-update setting. We additionally construct three dynamic node-classification tasks by discretising the underlying continuous-time event streams; here SiST-GNN beats the leading discrete-time (DTDG) baseline by $7$--$22\%$ and matches continuous-time (CTDG) methods that consume the raw events directly.
Authors: Bo Zou, Chao Xu
Abstract: High-quality expert chain-of-thought (CoT) data is one of the core bottlenecks in large language model (LLM) post-training. Existing data production methods each have structural limitations: crowdsourced annotation lacks deep reasoning paths; expert solo writing is constrained by the "expert blind spot" -- experts structurally skip reasoning steps they consider obvious; RLHF only produces preference signals rather than reasoning chains. This paper proposes the BC Protocol -- a structured dual-expert elicitation method for LLM post-training data production. The method carefully pairs a domain expert (crystallized intelligence) with a knowledge engineer (fluid intelligence), systematically externalizing the expert's implicit judgments as natural language reasoning chains. We introduce the Participant Aptitude Model, which defines six participant characteristic dimensions that affect elicitation quality. "Calibrated Ignorance" is an original concept proposed in this paper. We further propose "Selection-over-Prescription" as a methodological principle: for implicit knowledge elicitation tasks, investing quality-control resources in personnel selection yields a higher return than investing the same resources in process design. In a controlled experiment in the narrative fiction domain, we directly compared CoT produced by BC Protocol dual dialogue (Group A, (n=20)) against CoT written independently by the same domain expert (Group B, (n=20)). Three cross-vendor judge models -- GPT-4o, Claude Opus 4.5, and Gemini 2.5 Pro -- conducted blind evaluation across five dimensions (600 ratings total). Results show that the BC Protocol achieves an overwhelming advantage in "naturalness of reasoning process" (Group A mean 4.80 vs. Group B mean 1.30, (p=2.4\times10^{-8}), Cliff's (\delta=1.0)).
Authors: Austin Shen, Yunong Shi
Abstract: Automated theorem proving systems built on Lean 4 increasingly rely on parallel tactic search over partially specified proofs, such as those generated by Draft-Sketch-Prove (DSP) pipelines. In current systems, each search branch reconstructs a proof state by re-running elaboration, leading to substantial per-branch overhead. In Lean 4 with Mathlib, this cost has two components: (1) import loading, which deserializes pre-compiled libraries (~60 s per branch); and (2) theorem-body elaboration, which re-checks the theorem context up to the target goal (estimated 18-735 s depending on proof complexity). Together, these account for >99% of per-branch wall time, making portfolio-based search impractical at scale. We observe that this overhead arises from a mismatch between the structure of proof search and its execution model: branching is implemented via repeated reconstruction of proof states rather than direct reuse. To address this, we introduce proof-state snapshotting, which captures the elaborated proof state once and reuses it across branches via a small extension to the Lean 4 language server. Across 48 miniF2F-v2 problems (45 prove-phase benchmarks and 3 full end-to-end runs), our approach achieves a 5.6-50x wall-time speedup over the standard fallback (average 14x, median 9.7x). Speedup increases with the number of proof branches. Our method is orthogonal to import-level caching (e.g., Kimina Lean Server), which avoids import loading but not theorem-body elaboration. The patched Lean binary and the Snapshot-DSP pipeline will be released as open source upon publication.
Authors: Minghao Shao, Nouhaila Innan, Hariharan Janardhanan, Muhammad Kashif, Alberto Marchisio, Muhammad Shafique
Abstract: The growing complexity of quantum programming frameworks has exposed a critical limitation in existing large language model (LLM)-based code assistants: general-purpose models hallucinate PennyLane-specific gate names, misplace device configurations, and produce structurally invalid circuits when faced with specialized quantum coding challenges. We present PennySynth, a retrieval-augmented generation framework that addresses this gap by conditioning LLM inference on a curated knowledge base of 13,389 PennyLane instruction-code pairs, built via a three-stage extraction, verification, and deduplication pipeline over official PennyLane repositories, community GitHub sources, and QHack competition archives. PennySynth introduces a code-aware embedding strategy using st-codesearch-distilroberta-base, trained for natural-language-to-code retrieval, increasing average retrieval cosine similarity from 0.45 to 0.726 compared to a general-purpose baseline. Evaluated across 74 challenges spanning three years of the QHack competition (2022, 2023, 2024), PennySynth achieves 64%, 68%, and 52% pass@5 on QHack 2022, 2023, and 2024, respectively, improving over Claude Sonnet 4.6 without retrieval by +28, +25, and +28 percentage points. We further introduce a quantum-adapted CodeBLEU metric that upweights qml.* token patterns and show that structural code similarity and functional correctness capture distinct aspects of quantum code quality. Controlled ablations reveal that code-aware embeddings are the primary driver of retrieval performance, while dataset expansion and source composition provide additional gains when retrieval quality is sufficiently precise.
Authors: Junseok Ko, Jungwoo Kim, Jong-Seok Lee
Abstract: Concept erasure has emerged as a key research direction for ensuring safe and ethical image synthesis in Text-to-Image (T2I) models. While existing studies have explored concept erasure across multiple concepts, they typically assume only a single target concept per image, a limitation increasingly exposed by modern flow-based T2I models, which can generate complex scenes with multiple concepts simultaneously. To address this gap, we introduce compositional multi-concept erasure, a new task that aims to simultaneously remove multiple target concepts within a single scene. We propose CoME-Bench, a benchmark for evaluating compositional multi-concept erasure, which covers both intra- and cross-category scenarios. We further propose Mosaic, a novel framework for multi-concept erasure in flow-based T2I models, which exploits the spatial locality of target concepts in the vector field by dynamically constructing concept-specific masks and selectively blending them without additional optimization. Extensive experiments demonstrate that Mosaic effectively removes multiple target concepts in complex compositional scenes while preserving non-target contexts.
Authors: Yunqing Liu, Yi Zhou, Wenqi Fan
Abstract: The generation of accurate 3D molecular conformations is a pivotal challenge in computational chemistry and drug discovery. Recently, diffusion and flow matching models have achieved remarkable success. However, there is a critical misalignment between their mathematical formulation and the physical reality of molecules. Existing approaches predominantly treat molecules as unstructured point clouds in Cartesian space, overlooking the intrinsic hierarchical mechanics where bond lengths and bond angles are relatively stiff, whereas torsion angles constitute the dominant flexible degrees of freedom. This lack of manifold awareness forces models to relearn fundamental geometric constraints from scratch, often leading to physically implausible intermediate structures. To address this, we propose GO-Flow that aligns generative modeling with molecular geometry via manifold decomposition. Instead of forcing motion through Euclidean space, GO-Flow decomposes the generation process into three physically motivated subspaces: translation space with linear optimal transport, rotation space with geodesic flows on $SO(3)$, and conformation space with entropic optimal transport. This decomposition injects geometric inductive biases and makes the generative paths better aligned with molecular degrees of freedom. When combined with equivariant neural architectures, it encourages rotation-consistent generation and improves geometric validity. Extensive experiments on GEOM-Drugs and GEOM-QM9 demonstrate that GO-Flow achieves state-of-the-art generation quality. Notably, by learning straighter probability paths on the correct manifolds naturally, our method enables high-fidelity sampling with as few as 50 steps, effectively bridging the gap between structural precision and computational efficiency.
Authors: Changyu Chen, Xiting Wang, Rui Yan
Abstract: Reinforcement learning for large language models faces a fundamental trade-off between sample efficiency and asymptotic performance: strictly on-policy methods discard trajectories after a single update, while off-policy reuse introduces distribution mismatch that existing trust-region techniques mitigate primarily by enforcing conservative optimization, often leaving rich training signals underutilized. To investigate this, we perform extensive off-policy updates on fixed data. Our experiments reveal that aggressive multi-step optimization brings rapid initial gains, but excessive updates cause trajectory probabilities to deviate and entropy to collapse, with performance plateauing early. Tightening KL constraints merely lowers the ceiling without resolving the degradation. This motivates Extreme Region Policy Distillation (ERPD), a two-stage framework that decouples sample efficiency from KL efficiency. The first stage performs weakly constrained off-policy optimization on fixed data to maximally extract training signals. The resulting policy provides token-level supervision. In the second stage, we distill these signals into the base policy under trust-region constraints, filtering harmful drift while preserving useful signals. The distilled policy achieves comparable or better performance with substantially smaller KL divergence, indicating that much of the first-stage divergence was spent on unnecessary drift rather than genuine improvement. Crucially, ERPD accommodates both strong and weak teachers: when aggressive optimization yields no stronger policy, even degenerate teachers provide effective supervision via alternative signal construction strategies. We validate ERPD on mathematical reasoning, showing gains for strong base models where on-policy training plateaus, and reliable improvements with weak teachers.
Authors: Alexander Apartsin, Yigal Meshulam, Yehudit Aperstein
Abstract: Multi-robot task allocation usually assumes some combination of communication, known task models, or a coordinator. We study the opposite extreme, a regime common in practice but overlooked in theory, which we name Zero-Knowledge MRTA (ZK-MRTA): a robot team with no prior knowledge (no task models, not even the latent rank), no communication (no messages, no parameter sharing, no coordinator), and only a partial and privately-noisy view of a public stream of teammates' outcomes. A hidden low-rank structure governs which robot suits which task, and there are far more tasks than rounds, so most (robot, task) pairs are never attempted. Yet each robot can act well on tasks it never attempted, and onboard new tasks, by running online low-rank collaborative filtering over the broadcast (SwarmCF). The advantage over any structure-free learner is categorical, not a constant factor: a structure-free learner is provably at the prior-mean error floor on unseen pairs. We prove a matching per-robot sample complexity ({\Theta}(d) versus {\Theta}(n), in the rank d and the task count n), an anytime (cumulative-reward) separation under task scarcity, and a deterministic condition under which decentralized recovery from the masked broadcast is exact (validated empirically). Experiments quantify the value of the broadcast, a positive scaling law (per-robot unseen-pair skill rises with team size), and the strongest masking-robustness and anytime profile among low-rank methods, recovering most (about 80% on earned skill) of a centralized full-communication ceiling, and holding under capacity-1 contention and in a robotics-grounded sensing instance.
Authors: Alexander Apartsin, Omri Sason, Yehudit Aperstein
Abstract: Teacher education requires deliberate practice with learners who exhibit identifiable strengths, weaknesses, and partial mastery. Large language models could support such practice by simulating students with known skill components, enabling teachers to rehearse explanations, diagnoses, and instructional responses. For this purpose, however, the central requirement is neither to maximize benchmark accuracy nor to suppress isolated facts, but to control model behavior so that it reflects a specified skill profile. This paper investigates whether prompted language models can be steered to retain some skills while suppressing others. We introduce a benchmark-oriented framework in which an explicit skill vector represents a simulated student, prompt-based control specifies retained and missing competencies, and behavior is evaluated using profile-alignment metrics, retained-versus-forgotten comparisons, and cross-skill calibration analyses. The results show that selective partial mastery can be induced and measured in a structured mathematics setting, although the degree of controllability remains model-dependent. These findings position controllable learner simulation as a distinct research problem at the intersection of teacher education, educational simulation, and language-model control.
Authors: Ze Peng, Jian Zhang, Lei Qi, Yang Gao, Yinghuan Shi
Abstract: The observation that activation sparsity emerges in MLP blocks of standardly trained Transformers offers an opportunity to drastically reduce computation costs without sacrificing performance. To theoretically explain this phenomenon, existing works have shown that activation sparsity does not result from the data properties or data fitting but from the implicit bias of the training process. However, these connections are obtained with strong assumptions, which cannot be applied to deep models standardly trained with a large number of steps. Different from these works, we find that the flatness of loss landscapes is also closely related to the MLP activation sparsity and can serve as a weaker and naturally emerging assumption standard deep networks. Specifically, we find that 1) the MLP activation sparsity equals a ratio between "augmented flatness" (a weighted sum of flatness measures) and the product of the input norm and activation gradient of the MLP. We empirically find that this ratio decreases during training, leading to sparse activations. 2) We also propose the notion of derivative sparsity, which reduces to activation sparsity under ReLU, but further enables pruning in the backward propagation and is more stable than activation sparsity. With the theoretical findings, we can further encourage activation sparsity by decreasing the numerator and increasing the denominator of the ratio using three methods. These plug-and-play modifications can effectively reduce the ratio and produce sparser activations. Experiments on ImageNet-1K and C4 demonstrate relative improvements of at least 36% on inference sparsity and at least 50% on training sparsity over vanilla Transformers, indicating further potential cost reduction in both inference and training
Authors: Jatin Kishnani, Mayank Goel, Amit Singh, Pulkit Agrawal, Sairanjan Mishra
Abstract: We present the first end-to-end demonstration of fine-tuning and serving Google's Gemma 4 31B model on TPU hardware, providing an empirical comparison of TPU and GPU platforms for large language model adaptation. Using LoRA on a Google TPU v5p-8 for training and TPU v6e-8 (Trillium) for inference, we document the full set of code-level adaptations required to port a GPU-native training recipe, built on PyTorch, HuggingFace TRL, and FSDP, to the JAX + Tunix/Qwix stack. These adaptations span mesh configuration, LoRA module naming conventions, sharding annotation corrections, gradient checkpointing, data pipeline restructuring, and a custom Orbax-to-safetensors checkpoint merging procedure. For inference, we detail the vLLM-TPU Docker setup necessary to serve Gemma 4 on v6e-8 and characterize the resulting latency and throughput profile. Compared with a 2xH100 GPU baseline under identical hyperparameters, TPU training completes 1.61x faster at 2.12x lower cost. Inference throughput is within 3% across platforms, while TPU achieves 2x lower time-to-first-token (235 ms vs. 475 ms). Together, the TPU configuration is 1.82x cheaper for a representative train-plus-service workload. Our work removes a critical gap in the open tooling ecosystem and provides practitioners with a reproducible, production-ready recipe for Gemma 4 deployment on TPU infrastructure.
Authors: Haoran Gu, Handing Wang, Yi Mei, Mengjie Zhang
Abstract: Expensive optimization tasks are ubiquitous in real-world applications, demanding highly specialized solvers. While LLM-driven automated solver generation shows promise, current paradigms face three critical issues when tackling expensive optimization: factual hallucinations due to deficient domain knowledge, the frequent dismantling of previously established locally optimal structures during refinement, and the prohibitive evaluation costs alongside restricted generalization caused by executing on training instances. To address these issues, we introduce AutoSG, a fully automated workflow directly translating natural language prompts into executable customized solvers. AutoSG features three core innovations: a retrieval-augmented solver generation module strictly grounding code in verified literature; a one-step self-refinement operator introducing task-specific improvements while preserving critical structural components; and an instance-free Elo-based LLM-as-a-Judge evaluation mechanism rapidly establishing global rankings. Extensive evaluations across diverse expensive optimization tasks confirm AutoSG significantly outperforms human-designed state-of-the-art frameworks and existing LLM-generated solvers.
Authors: Arka Majhi, Aparajita Mondal
Abstract: Poor posture is a significant concern due to its detrimental effects on health and productivity. This paper presents a collar-clipped device called PostureClip, designed to restrict users from sitting and working at a bent angle, by blacking out the screen and resuming on correcting posture, thereby promoting better posture. The device integrates sensors and feedback mechanisms to provide real-time posture feedback to users. To evaluate the effectiveness of PostureClip, a controlled experiment was conducted with participants (n=165) who were working on a laptop/PC for over 6 hours per day. The participants were randomly assigned to both the intervention group (IG1,n=54 ; IG2,n=55), which used the collar-clipped device, and the control group (CG, n=56), which did not use the device. IG1 didn't get feedback while IG2 got feedback from the device by notifying and further darkening the screen. The study was conducted in the office environment of the participants, for 4 weeks, and metrics such as posture angle, duration of bent angle, and user feedback were collected. Analysis revealed significant improvements in posture angle (p<0.001) and significant reduction in bent angle duration (p<0.01) for participants' group using PostureClip with feedback and compared to the group without feedback and the control group (who were not intervened). The qualitative analysis of user feedback highlighted the device's ease of use, effectiveness in providing timely feedback, and positive impact on participants' awareness and habits regarding posture. These results indicate that PostureClip is an effective tool for promoting better posture during sedentary work.
Authors: Satadru Sengupta, Tamunokorite Briggs, Ivan Myshakivskyi
Abstract: AI-native software development is often evaluated at the level of individual models, prompts, or generated artifacts. This framing is insufficient for production environments where software must be continuously produced, verified, deployed, maintained, and adapted across many operational contexts and long time horizons. We present a meta-engineering harness: a software-production architecture that transforms operational and product feature requirements into explicit contracts, routes work through role-specialized AI agents, performs independent and adversarial verification, and continuously improves itself through structured failure classification and outer-loop calibration. The harness is designed for settings in which software delivery is not a one-time project but an ongoing operating function. In our motivating application, CTO-as-a-service for small service firms, the system manages websites, booking flows, payment systems, backoffice workflow automations, and AI-agent interfaces as continuously evolving technical infrastructure rather than one-off deliverables. We describe the layered architecture, including two-pass contract compilation, persistent markdown memory with specialization records, attention-based and independence-based verifications, a four-way failure arbiter, and outer-loop calibration. We report results from an early production deployment spanning 17 features over several weeks, including a detailed in-app payments case study that revealed contract incompleteness and verification-boundary issues. These observations directly drove targeted improvements to the harness. The contribution is an implemented, measurable, and extensible verification architecture for making AI-native service-as-a-software production reliable, auditable, and improvable over time.
Authors: Dan Ristea, Vasilios Mavroudis
Abstract: Security evaluations inherently depend on stable identifiers. Any finding, audit, or regulatory decision must remain attached to the specific artifact it pertains to. Continuously updated artificial intelligence systems violate this core assumption, with public model designations remaining static while underlying weights, prompts, retrieval mechanisms, misuse classifiers, inference settings, and serving infrastructures undergo unannounced modifications. Consequently, current evaluations frequently apply to superficial labels rather than identifiable and distinct systems. To resolve this, we propose referential security as a new paradigm for AI evaluation. The fundamental security question extends beyond whether a model is safe to whether subsequent parties can conclusively determine which system a specific safety claim addressed. This approach reframes model identity as an empirically verifiable property and separates referential stability from the substantive security claims it conditions. This framework brings tractability to three critical workflows that current practices handle poorly. Specifically, it enables reproducible evaluation, longitudinal audit validity, and cross-provider equivalence. By grounding these evaluations in verifiable artifacts, our approach ensures that safety audits and regulatory findings maintain their empirical utility across the operational lifecycle of dynamic systems.
Authors: Qihan Wang, Nicholas Tomlin, Michael Hu, Brian Dillon, Tal Linzen
Abstract: Language models are increasingly being deployed as user simulators, but their memory is far more reliable than that of real users. To measure this gap, we run a series of classic memory experiments from psychology on both humans and language models. Across tasks, we find that out-of-the-box language models exhibit better memory than humans, even when prompted to imitate human behavior. We then show that better prompting strategies and the use of a compactor can cause language models to forget content in a more human-like way. Using these methods, we show preliminary evidence that language models with human-like memory constraints can function as more effective user simulators in a downstream education task. Finally, we release human reference data and benchmarks to support future work on simulating human memory with language models.
Authors: Qingyuan Zeng, Pengxiang Cai, Zixin Guan, Ziyang Chen, Anglin Liu, Lang Qin, Xinyao Lai, Jintai Chen
Abstract: Designing a single molecule that modulates two targets is a promising strategy for polypharmacology, but it remains substantially harder than standard single-target generation because one candidate must satisfy two binding requirements while preserving drug-likeness and synthesizability. Existing dual-target generative methods typically introduce dual-target capability by either retraining the generator or intervening in the diffusion process during sampling. The former can be costly and difficult to stabilize when dual-target supervision is sparse, while the latter may be sensitive to denoising-time target balancing and competing update directions. These limitations motivate a generator-preserving alternative that keeps the pretrained prior intact: can dual-target candidates instead be recovered from the input space of a frozen single-target diffusion model, without modifying its parameters or denoising dynamics? We formulate this task as a constrained multi-objective optimization problem and propose REUSE, a hierarchical evolutionary input-space search framework that combines pair-conditioned exploration with structured multi-stage selection to enforce dual-target affinity, chemical quality, and diversity. Experiments show that, compared with methods that modify the diffusion process, REUSE consistently improves dual-target affinity and balance, achieving a 20.9-percentage-point gain in Dual High Affinity over the strongest prior baseline while maintaining competitive molecular quality.
Authors: Muhammad Azlan Qazi, Alexandros Iosifidis, Qi Zhang
Abstract: Distributing Transformer inference across embedded edge devices can alleviate individual memory and compute constraints, yet practical benefits on real hardware remain unclear: prior work relies largely on simulations that overlook hardware-specific communication overheads. We present a hardware prototype study on NVIDIA Jetson Orin Nano devices connected over WiFi. Our key finding is that the dominant bottleneck is not just network bandwidth but also the CPU-GPU staging during communication. Because Jetson's integrated GPU architecture lacks the PCIe/NVLink pathway that NCCL requires, all inter-device data communication should be routed through GLOO and staged in CPU memory; an overhead that scales with communication data volume and makes full-tensor exchange slower than single-device inference across the batch sizes for medium sized models such as ViT. We therefore evaluate Prism by combining Segment Means compression with lightweight offline profiling to adaptively select between local and distributed execution at runtime. Experiments show that this strategy reduces latency by 65%-77% and energy consumption by 34%-52% relative to full-tensor exchange in static distributed execution setup, demonstrating that profiling-driven adaptation is essential for practical distributed Transformer inference on embedded hardware.
Authors: Zhitao Zhu, Xili Wang, Shizhe Wu, Jiawei Fu, Xiaoqing Liu
Abstract: High-quality data is scarce in large language model (LLM) training, yet how to schedule its use jointly with training dynamics lacks theoretical guidance. We extend functional scaling laws by incorporating a data-quality dimension, and solve the joint data-quality and batch-size scheduling problem in asymptotic closed form. The solution reveals two regimes and a dual role of high-quality data. In the noise-limited regime, high-quality data should be used as a signal amplifier: lowering the batch size converts cleaner data into more signal without amplifying noise. In the signal-limited regime, it should be used as a noise suppressor: late placement reduces terminal noise without sacrificing signal accumulation. Existing curriculum-style pipelines primarily exploit the second role by placing cleaner data late, but miss the first role because conventional decay schedules reduce update intensity exactly when high-quality data becomes available. Guided by this, we propose Drop-Stable-Rampup for LLM midtraining: upon the quality transition, drop the batch size, hold it stable to accumulate signal, then ramp up to suppress terminal noise. On a 15B Mixture-of-Experts model midtrained on 108B tokens, Drop-Stable-Rampup improves average accuracy over Warmup-Stable-Decay (WSD) by +1.70 and over Cosine-decay by +2.98, with particularly large gains on mathematical reasoning benchmarks such as GSM8K (+4.23) and MATH (+2.80).
Authors: Haoran Li, Shulun Chen, Shaoyuan Sun, Hanchen Wang
Abstract: As large language model (LLM)-based multi-agent systems scale to handle increasingly complex tasks, balancing structural stability and dynamic adaptability becomes increasingly challenging. Existing systems typically adopt either structure-centric methods, committing to structures determined upfront that limit fine-grained control, or orchestration-centric methods, adapting decisions dynamically while leaving coordination structure implicit and unstable. To address this challenge, we revisit multi-agent coordination from a probabilistic perspective, casting it as posterior inference over the joint distribution of structure and orchestration. We introduce MACA, an automated coordination framework that learns a task- and budget-conditioned structural prior over agent participation and interactions. This prior guides a policy-based orchestration as an approximation to posterior inference, enabling efficient solutions with fine-grained control. Across benchmarks, MACA outperforms adaptive multi-agent baselines by an average of 8.42% while using 43.19% fewer tokens. Further investigation reveals that joint adaptation of structure and orchestration suppresses redundant interactions, converging coordination toward task-effective execution.
Authors: Chaotian Song, Jingyao Zhang, Chenghao Chen, Zisen Sang, Dehai Zhao, Guodong Cao, Boxi Wu, Deng Cai, Jia Jia
Abstract: In multi-stage recommender systems, reranking optimizes overall utility by capturing intra-list contextual dependencies, yet its central challenge lies in exploring optimal sequences within an exponentially large permutation space. Recent studies have shifted towards end-to-end generative frameworks, which typically leverage list-wise rewards or preference alignment to guide generator training. However, these methods still face two critical issues. First is the heuristic label bias. Existing methods often construct training targets based on simple rules, such as promoting clicked items to the top, while ignoring causal dependencies within the list context. Second is the credit assignment problem. Sparse list-level posterior rewards fail to directly guide intermediate steps in sequence generation, leading to ambiguous optimization directions. To address these issues, we propose DeGRe (Dense-supervised Generative Reranking), a generative reranking framework that bridges the gap between offline exploration and online efficiency through dense supervision. The core of DeGRe lies in its offline-online decoupled design. During the offline phase, we introduce a Lookahead Evaluator based on cumulative regression, which leverages beam search to actively mine high-value lookahead sequences in the unexposed space. During training, we transform the step-wise value estimations from the evaluator into dense supervision signals and distill them into a lightweight Online Generator. This mechanism enables the generator to internalize lookahead planning capabilities, requiring only a single efficient greedy decoding pass during online inference to approximate the global optimum. Experiments demonstrate that DeGRe outperforms baseline models on public benchmarks and industrial datasets. We have successfully deployed DeGRe on Taobao Flash Shopping, significantly improving online recommendations.
Authors: Bokai Zhao, Yiyang Zhang, Yuanchi Zhu, Hanqing Chao, Long Bai, Tai Ma, Minfeng Xu, Ming Song, Tianzi Jiang
Abstract: Pathology foundation models (PFMs) have emerged as a core approach for learning transferable representations from whole slide images (WSIs), and they are typically benchmarked through downstream clinical endpoints. While such task level evaluations are indispensable, they offer limited insight into what the representations themselves encode, particularly whether PFM embeddings can distinguish meaningful tissue regions and capture their spatial relationships. We present SpaPath-Bench, a representation level benchmark designed to diagnose spatial representation capability in PFMs. SpaPath-Bench formulates spatial domain identification (SDI) on paired whole slide image and spatial transcriptomics (ST) data as a diagnostic task. It curates 42 public paired WSI and ST slides, enables large scale evaluation across 19 encoders and seven SDI methods, and measures partition quality using three complementary criteria: unsupervised spatial coherence, transcriptomics referenced agreement, and expert referenced agreement. Across 83K runs, SpaPath-Bench reveals that different pretraining paradigms capture distinct aspects of tissue spatial architecture, and it provides practical guidance for building the next generation of spatially aware computational pathology models. Code and data pipelines are publicly available at https://bokai-zhao.github.io/SpaPath-benchboard/.
Authors: Saemi Moon, Suhyeon Jun, Seoyeon Lee, Dongwoo Kim
Abstract: Concept unlearning aims to erase a target concept from a pretrained text-to-image diffusion model without retraining. Closed-form methods are attractive in this setting because they apply a single deterministic edit to the cross-attention weights and add no inference-time cost. Existing closed-form methods, however, represent the target concept through the text encoder's response to a few short anchor prompts that name it, and paraphrased prompts that evoke the concept without naming it consistently bypass the edit. We argue that the target should instead be represented in the cross-attention activation space. Text embeddings describe the user's prompt, while cross-attention activations describe what the model is about to render, and the latter generalize to paraphrase the anchor templates do not cover. Building on this observation, we propose PURE (Projection in U-Net Rendering for Erasure), a closed-form method that builds the forget and retain bases from per-layer cross-attention activations captured along a short denoising trajectory and applies a single linear projector to the cross-attention key and value weights. On a recent holistic concept-unlearning benchmark covering ten concepts across artistic style, intellectual property, celebrity, and NSFW categories, PURE significantly reduces target leakage under paraphrased and adversarial prompts while preserving retain concepts close to the unedited model, yielding the best overall forget-retain trade-off among evaluated methods.
Authors: Ziyu Zheng, Yaming Yang, Ziyu Guan, Wei Zhao, Xinyan Huang
Abstract: Multi-domain graph pre-training is a crucial step in constructing foundational graph models with cross-domain generalization capabilities. However, existing methods predominantly rely on jointly training all source domain graphs, resulting in high computational costs. Furthermore, it remains unclear whether all source domain graph data contribute equally to effective transfer. This paper empirically reveals significant data redundancy in multi-domain graph pre-training. Based on this finding, we propose the Multi-domain Graph Pre-training Framework, MDGMIX, which combines boundary-aware subgraph mixing with hierarchical discrimination. By selecting boundary nodes to construct challenging mixed-domain subgraphs, MDGMIX employs coarse-grained domain discrimination and fine-grained domain decomposition losses to decouple shared patterns from domain-specific patterns. During adaptation, MDGMIX employs a lightweight prompt weighting mechanism to transfer source domain knowledge. Extensive experiments demonstrate that MDGMIX consistently outperforms strong baselines in few-shot classification tasks while exhibiting superior time and memory efficiency. The code is available at: https://github.com/zhengziyu77/MDGMIX.
Authors: Sam Bowyer, Acyr Locatelli, Kris Cao
Abstract: Efficient benchmarking techniques aim to lower the computational cost of evaluating LLMs by predicting full benchmark scores using only a subset of a benchmark's questions. By reframing this problem as an instance of multiple regression with feature selection, we find that existing efficient benchmarking methods can be greatly improved by simply using kernel ridge regression at the prediction stage. Additionally, using an information-theoretic feature-selection algorithm called minimum redundancy maximum relevance (mRMR), we can further improve upon these methods by selecting question subsets that will be maximally useful for prediction. Except in very data-poor settings, these approaches consistently achieve smaller prediction errors (in both MAE and RMSE), and greater ranking correlation between predicted and true scores (in both Spearman $\rho$ and Kendall $\tau$) across a range of benchmarks using both binary and continuous metrics. Furthermore, mRMR subsampling is much faster than competitor methods (which often involve fitting probabilistic models or running clustering algorithms), and is more likely to select the same questions under different random seeds or training data splits. Tutorial code can be found at https://github.com/sambowyer/mrmr_eval .
Authors: Bocheng Zeng, Rui Zhang, Runze Mao, Mengtao Yan, Xuan Bai, Yang Liu, Zhi X. Chen, Hao Sun
Abstract: Efficiently solving Poisson equations on complex, irregular domains remains a fundamental challenge in scientific computing, as classical iterative solvers often suffer from prohibitive runtime due to ill-conditioned systems. While neural operators offer a fast alternative, they typically rely on large-scale labeled datasets or struggle with unstable training dynamics when using physics-informed residual losses. We propose \textsc{NPSolver}, a neural Poisson solver trained without solution labels via iterative physics supervision. Instead of relying on fully converged numerical solutions or raw PDE residuals, \textsc{NPSolver} utilizes a small number of preconditioned conjugate gradient (PCG) steps to refine its own predictions, providing a more stable and well-scaled training signal. Theoretical analysis confirms that this iterative supervision serves as a well-conditioned error proxy and that a stop-gradient design is essential for optimization stability. To better capture boundary-driven features under mixed boundary conditions, we further introduce the Boundary-Aware Transolver (\textsc{BA-Transolver}) architecture that explicitly separates interior and boundary tokenization. Extensive evaluations on 2D and 3D irregular geometries demonstrate that \textsc{NPSolver} outperforms both physics-informed and data-driven baselines. Furthermore, a downstream thermal control task highlights the model's capability for conducting efficient and reliable gradient-based boundary control. We will release our codes and data at https://github.com/intell-sci-comput/NPSolver.
Authors: Yunlong Hou, Zixin Zhong, Vincent Y. F. Tan
Abstract: We study a stochastic multi-armed bandit problem where an agent is granted a free exploration budget before regret accumulates, a setting not captured by the classic regret minimization or pure exploration paradigms. The goal is to design an adaptive policy that strategically explores the bandit instance in the initial free exploration phase and minimizes the cumulative regret in the subsequent phase. We formalize this regret minimization with free exploration problem and identify an interesting regime where the free exploration budget scales logarithmically with the time horizon. To quantify the amount of regret saved with high probability as a result of the availability of the free exploration phase, we introduce a novel set of policies known as $(\alpha,\beta)$-probably saving policies. We propose a two-phase, probably saving algorithm, UFE-KLUCB-H, which consists of a principled free exploration policy, UFE, and a history-aware regret minimization policy KLUCB-H. Instance-dependent upper bounds on UFE-KLUCB-H are derived, showing that UFE-KLUCB-H accumulates strictly less regret than policies that do not have access to a free exploration phase. Complementarily, we derive instance-dependent lower bounds based on novel multi-instance perturbation arguments tailored to the free-exploration setting, demonstrating the near-optimality of UFE-KLUCB-H for two-valued bandits. Our upper and lower bounds reveal sharp phase transitions in the accumulated regret depending on the amount of available free exploration. Simulations are conducted to demonstrate that forced exploration and adaptivity in the algorithm lead to greater regret savings.
Authors: Jiahao Huo, Wenjie Qu, Yibo Yan, Kening Zheng, Jiaheng Zhang, Xuming Hu, Philip S. Yu, Mingxun Zhou
Abstract: Semantic-level watermarking (SWM) improves robustness against text modifications by treating sentences as the basic unit. However, robustness to paragraph-level paraphrasing remains difficult because such attacks globally disrupt watermark signals by changing sentence order. In this work, we propose SAMark, a self-anchored watermarking framework that removes the dependency on sentence order by establishing a step-independent green region in semantic space. To improve detectability, we introduce a multi-channel hyperbolic scoring mechanism that amplifies watermark signals while suppressing noise from weakly aligned candidates. We further propose a diversity-aware filtering strategy that combines hard filtering with soft regularization, extending beyond simple n-gram repetition filters to address semantic redundancy. Experimental results show that SAMark achieves up to 90.2% TP@FP1% under typical paragraph-level paraphrasing attacks, outperforming the strongest prior baseline by more than 30% on average, while maintaining generation quality competitive with unwatermarked text and breaking the robustness-quality trade-off that limits prior methods.
Authors: Hongtao Wang, Renchi Yang, Haoran Zheng, Xiangyu Ke
Abstract: Dirty entity resolution (ER), which identifies records referring to the same real-world entity from a single, messy dataset, is a fundamental task in data management and mining. However, the dominant blocking-matching-clustering paradigm for ER suffers from critical flaws. Its cascaded, decoupled workflow essentially produces a static, sparse graph plagued by missing edges (due to blocking failures) and noisy links (due to matching errors), causing error propagation and yielding suboptimal clusters, particularly when rigid transitivity is imposed in the clustering. We contend that matching and clustering are fundamentally synergistic, both optimizing for the construction of an ideal entity graph. Building upon this insight, we propose Alper, a unified framework that integrates these steps into an iterative probabilistic label propagation process over a global, evolving graph. Unlike disjoint blocking, Alper refines the graph structure and labels dynamically by adaptively integrating "weak but cheap" signals from graph propagation with "strong but expensive" LLM-based pairwise queries. For higher cost-effectiveness, we formulate the signal selection as a constrained optimization problem maximizing cumulative marginal gain under a query budget, solved via our greedy algorithm with provable theoretical guarantees. Our extensive experiments over eight benchmark datasets demonstrate that Alper is consistently superior to state-of-the-art cascaded pipelines.
Authors: Pritesh Jha
Abstract: Personally identifiable information (PII) detection systems are frequently trained within narrow source or domain boundaries, limiting coverage when deployed on heterogeneous text. We study model fine-tuning on a corrected multi-source PIIBench preparation spanning 82 retained entity types across ten source datasets. We evaluate three DeBERTa-based approaches: direct token classification fine-tuning, a source-conditioned hierarchical model (SC+H), and a three-phase curriculum extension (SC+H+Curr). Against eight published comparator systems on a reproducible 5,000-record held-out subset (test_5k), direct fine-tuned DeBERTa achieves F1 0.6476, while SC+H and the curriculum variant achieve 0.5899 and 0.2772 respectively; the strongest published comparator reaches only 0.1723. Because validation initially favoured SC+H, we perform a final streamed evaluation on the complete 100,002-record held-out split. Direct fine-tuning remains superior, achieving F1 0.6455 versus 0.5894 for SC+H. Entity-level analysis shows that direct fine tuning wins 54 of 82 fine entity types and all ten coarse groups by support-weighted entity F1, while SC+H retains localised advantages on 28 types. The results indicate that diverse task-specific training data and a simple weighted cross-entropy objective contribute more to broad-coverage PII detection than the tested architectural and curriculum complexity.
Authors: Xinzhe Chen, Sihua Ren, Liqi Huang, Haowen Sun, Mingyang Li, Xingyu Chen, Zeyang Liu, Xuguang Lan
Abstract: Recent vision-language-action (VLA) models and world action models (WAMs) advance robotic manipulation by enriching intermediate representations with auxiliary spatial features or future visual-state prediction. However, these representations largely remain within the observation space and do not share the rigid-body geometry of the action space, forcing the action decoder to implicitly recover this geometry. We propose OASIS, a visuomotor policy that aligns the intermediate representation with the action space via $SE(3)$ end-effector trajectory prediction. OASIS couples a 3D-aware feature encoder that fuses vision-language and metric-depth features with an $SE(3)$ trajectory predictor that produces a camera-frame end-effector trajectory. Conditioned on the predictor's pose-supervised hidden states, the action decoder generates action chunks consistent with rigid-body motion. Across simulation and real-world experiments, OASIS outperforms VLA and WAM baselines in success rate and out-of-distribution generalization. Our project page is available at https://npuhandsome.github.io/OASIS_web.
Authors: Joris Baan, Wilker Aziz, Barbara Plank, Raquel Fern\'andez
Abstract: Large language models (LLMs) define a distribution over text, which can be viewed as a probabilistic representation of uncertainty: sampling K responses yields a belief state - responses a model deems plausible. Existing work exploits this representation for narrow tasks like either decoding or selective prediction, and often requires manual interventions, not controlling generation directly. We propose Belief-Augmented Generation (BAG): grounding LLMs in their own belief state via the prompt and letting them reason over these K samples to decide on a conversational strategy: answer, clarify, or abstain. In a multi-turn ambiguous QA setting, we find that LLMs by default rarely clarify or abstain, ignoring uncertainty about the input or facts. BAG improves QA accuracy across six models and yields strategy decisions more faithful to the belief state than prompt-only baselines. Disentangling when to clarify from when to abstain, however, remains challenging.
Authors: Yunfei Wang, Xiaohao Xu, Yang Li, Xiaonan Huang
Abstract: Large language models (LLMs) are increasingly used as proposal generators for evolutionary robot design, yet most loops remain memoryless: simulator results shape the next population but are not preserved as reusable design knowledge. We present Auto-Robotist, a self-evolving LLM agent that distills morphology-search traces into an explicit natural-language skill library. Each skill stores a structural archetype, evidence-grounded positive and negative rules, and the evaluated designs that support them, making design memory inspectable rather than implicit in a population. During search, the agent retrieves skills to condition LLM edits of elite bodies while retaining a Genetic Algorithm (GA) mutation path for exploration; after evaluation, it updates the library through Add, Diagnose, and Merge. Across seven EvoGym tasks spanning locomotion, traversal, and object interaction, Auto-Robotist improves cold-start 5x5 search and transfers learned skills to 10x10 design spaces, where reference-conditioned transfer outperforms GA on every task. These results suggest that LLM agents can convert expensive physical evaluations into reusable, auditable design principles. Our code will be released upon acceptance.
Authors: Andrey Kozachok, Anatoliy Bakaev, Aleksandr Kozachok, Shamil Magomedov, Artem Noev
Abstract: This paper examines the specialization of Small Language Models (SLMs) with up to 4 billion parameters for generating artifacts in domain-specific languages (DSL). Kubernetes manifests are chosen as the target domain. We propose the context-instrumental data distillation method: the source corpus is formed through synthetic generation and, in an extended scheme, through reverse instruction generation from real Kubernetes YAML files, with pairs included in training only upon passing external validators and matching the domain context model. Unlike classical KL-divergence knowledge distillation, the baseline implementation reduces to supervised fine-tuning on instrumentally verified examples. The experimental section presents a pilot implementation under resource-constrained conditions: the DeepSeek-V4 Flash API serves as the teacher for synthetic generation, while Qwen2.5-Coder-1.5B-Instruct is fine-tuned via LoRA on CPU. On the K8s-Distill-Pilot corpus (train_1200, validation_100, test_200), we achieved full-pass@1 = 91.5% (183/200) with a stricter prompt formulation and max_new_tokens=768. The key empirical finding is that for Kubernetes YAML, result quality in the pilot depended more on strict output format requirements than on simply increasing the number of training examples.
Authors: Yutong Cheng, Changze Li, Raihan Sultan Pasha Basuki, Qian Cui, Wei Ding, Peng Gao
Abstract: Extracting MITRE ATT&CK techniques from cyber threat intelligence (CTI) reports is an open-set, multi-label problem requiring both high recall (not missing techniques) and high precision (not hallucinating unsupported ones). Existing methods--rule-based, supervised, and LLM-based--struggle to achieve both: rule-based and supervised approaches lack generalizability across diverse attack descriptions, while LLM-based approaches that couple candidate generation and validation within a single inference step suffer from limited recall and precision simultaneously. We propose TTPrint, which addresses this challenge through a diverge-then-converge design inspired by how human analysts work: first extracting broadly, then verifying rigorously. In the divergent phase, reports are decomposed into atomic behaviors and candidate techniques are proposed broadly. A deterministic span localization stage then anchors each candidate to a specific evidence window in the source text. A convergent verification stage retains only candidates supported by both the localized evidence and the authoritative MITRE definition. We contribute two evaluation resources--a cleaned TRAM benchmark (TRAM-Clean) and a new annotated dataset (TTPrint-Bench)--to address known annotation noise in existing benchmarks and elevate the task to document-level TTP extraction. On TRAM-Clean and TTPrint-Bench, TTPrint achieves 76.48% and 87.39% macro-F1 respectively, outperforming the leading baseline by 63.5% and 29.4%. A multi-backbone analysis across six LLMs and a threshold sensitivity study further demonstrate generalizability across model choices and provide practical guidance for parameter selection.
Authors: James Henry
Abstract: Concept probes extracted from transformer residual streams are only as reliable as the layer from which they are extracted. The common practice of probing at a fixed late layer or at the peak of a separation score function ignores a fundamental structural feature: concept representations undergo substantial directional rotation during their assembly phase, and do not settle into a stable direction until a characteristic handoff layer after the primary Concept Allocation Zone (CAZ). We introduce Geometric Evolution Maps (GEMs), which track the full directional trajectory of a concept through residual stream activations, identify the handoff layer where rotation ceases, and extract the settled probe direction from that layer. Across 23 architectures spanning 70M to 14B parameters and 17 concept types, the entry-to-exit cosine similarity within CAZs has a mean of 0.233, showing that probe direction at CAZ entry does not reliably predict probe direction at exit. Ablation experiments across 391 concept x model pairs (23 models x 17 concepts) show that GEM-extracted probes are at least as precise as peak-layer probes in 268/391 trials (68.5%), and strictly outperform in 259/391 (66.2%). The architecture split is pronounced: MHA models favour the handoff in 173/221 trials (78.3%); GQA models favour the handoff in only 56/119 trials (47.1%). Model-level Wilcoxon: W=214, N=23, p=0.010 (one-sided). An adaptive ablation width rule targets the 79/391 near-final-layer cases: it improves probe quality in 60/79 triggered cases (75.9%), mean gain +7.44pp. A direction-specificity control confirms the ablation effect is concept-direction specific: median 377x suppression rate versus random-direction ablation (99.1% of concept directions beat all 10 random seeds). Reference implementation: rosetta_tools v1.3.1 (doi:10.5281/zenodo.20361433).
Authors: Muyu Pan, Shu Zhao, Nan Zhang, Philip Shin, Varun Parekh, Vijaykrishnan Narayanan, Rui Zhang
Abstract: This paper investigates large language model (LLM) abstention learning, specifically using ternary reward, which incentivize truthfulness in large language models. This paper extends that idea by moving from a ternary reward to a Trajectory-Informed advantage reweighting, dynamically re-weights the abstention reward during Group Relative Policy Optimization (GRPO) training. The objective of this work focuses on abstention learning instead of improving truthfulness, serving as an exploration into hallucination reduction. The novelty of this paper lies in methodological innovation, advantage re-weighting, and benchmark selection. Leveraging GRPO's multiple trajectories as a natural abstention signal, this method uses a reward signal to explore knowledge boundaries and encourage consistency. By demonstrating that trajectories can be used as a confidence indicator of the policy relative to the query, they are then used to dynamically calculate the abstention advantage. AbstentionBench is used as the evaluation benchmark, as this work aims to contribute to the field of abstention learning. All datasets on the benchmark were tested against this method and various baselines. Empirical results demonstrate that TIAR achieves state-of-the-art abstention F1 scores across five of six evaluation categories, outperforming the static ternary baseline on 17 of 31 benchmark datasets while fully preserving baseline accuracy.
Authors: Daniela Fernandes, Daniel Buschek, Lev Tankelevitch, Thomas Kosch, Robin Welsch
Abstract: Large Language Model interfaces are increasingly verbose, exposing intermediate reasoning traces alongside final answers. Traces are framed as transparency mechanisms, yet it is unclear how people use them to solve problems. We report a preregistered between-subjects study (N = 559) in which participants solved ten LSAT-style reasoning problems under one of three conditions: an Answer-only baseline, a Full-trace revealed before the answer, and a Summary-trace presented alongside the answer. Summaries preserved task performance at the no-trace baseline while significantly elevating trust and hedonic appeal, establishing that trace exposure shifts subjective appraisal of the interaction without bringing performance benefits. Under an open-weight reasoning model exposing verbose intermediate output, full traces additionally impaired performance relative to the answer-only baseline. Across all conditions, participants substantially overestimated their performance, and no trace format supported calibrated self-evaluation. Further analysis indicates that hedonic appeal, not trust, carries the indirect path to overestimation, consistent with a processing-fluency account. Reasoning traces are best understood as user-facing interface artifacts rather than transparent windows into model cognition, and calibration is unlikely to emerge from the traces themselves and may best be scaffolded by interactions that elicit users' own reasoning first.
Authors: Yunqi Gao, Leyuan Liu, Yuhan Li, Changxin Gao, Jingying Chen
Abstract: 3D human mesh recovery and 3D clothed human reconstruction are inherently related, yet they have long been studied in isolation, thereby overlooking the potential gains of joint optimization. To overcome this limitation, we propose to address these two tasks within a unified framework, which allows their mutual dependencies to be effectively exploited. Building on this idea, we propose MuNet, a mutualistic network for joint 3D human mesh recovery and 3D clothed human reconstruction from single images. First, we adopt 2-manifold graphs as a unified representation for all 3D models, enabling consistent modeling across 3D human mesh recovery and clothed human reconstruction. Second, we design an end-to-end graph convolutional network that progressively deforms an initial graph into a 3D human mesh and refines it into a detailed 3D clothed human model. Third, we introduce a mutualistic mechanism that allows reciprocal interaction between the two tasks {during training}, where 3D human mesh recovery provides guidance for 3D clothed human reconstruction, and reconstruction feedback refines the 3D human mesh recovery. We extensively evaluate MuNet on six benchmark datasets for 3D human mesh recovery and 3D clothed human reconstruction, including Human3.6M, 3DPW, MPI-INF-3DHP, THuman2.0, CAPE, and RenderPeople. Experimental results demonstrate that MuNet achieves state-of-the-art performance on both tasks across all datasets. The code of MuNet is released for research purposes at https://github.com/starVisionTeam/MuNet.
Authors: Ziyi Ding, Xiao-Ping Zhang
Abstract: We find a mismatch between what large language models encode about a causal question and what they answer. On anti-commonsense CLadder items, a fixed linear probe recovers the evidence-supported answer from the model's hidden state (accuracy approximately 0.97), while the spoken Yes/No reverts to the commonsense one (accuracy approximately 0.5). We call this approximately +0.5 gap Causal Tongue-Tie: a wrong Yes/No decomposes into two separable failure modes: no internal signal versus a signal the verbal interface cannot say. The implication cuts both ways for output-only causal benchmarks: a benchmark "correct" need not mean the model has understood, and a benchmark "wrong" need not mean it cannot. Sweeping claims about whether LLMs can do causal reasoning, drawn from a single accuracy number, deserve a second look.
Authors: Wei Fan, Yining Zhou, Mufan Zhang, Yanbing Weng, Yiran HU, Tianshi Zheng, Baixuan Xu, Chunyang Li, Jianhui Yang, Haoran Li, Yangqiu Song
Abstract: While large language models (LLMs) augmented with agentic search capabilities show promise for legal reasoning, they overlook a fundamental constraint that applicable law must match the temporal context of each case, as retroactive application of statutes violates core legal principles and leads to erroneous conclusions. Our observations reveal that current legal LLMs suffer from temporal bias anchored to their training cutoff, while search agents rarely incorporate temporal constraints into queries, and that web search alone cannot provide the precise statute and precedent citations that legal reasoning demands. To address these challenges, we propose LegalSearch-R1, an end-to-end reinforcement learning framework that pairs local statute RAG for precise article matching with online web search for broader legal knowledge, trained on temporally-indexed data spanning multiple amendment periods to enforce temporal consistency. Extensive experiments on our benchmark covering 13 legal tasks demonstrate that our 7B-parameter agent outperforms state-of-the-art deep research frameworks and specialized legal LLMs by 12.9% to 29.8%, surpasses baselines by 57.7% to 80.3% on temporal consistency, and exhibits robust out-of-domain generalization. The code and data are available at https://github.com/AlexFanw/LegalSearch-R1.
Authors: Nicolas Ricka, Gauthier Pellegrin, Denis A. Fompeyrine, Thomas Rohaly, Leah Enders, Heather Roy
Abstract: Posttraumatic stress disorder (PTSD) is a prevalent and debilitating mental health condition with significant personal and societal impacts. Current clinical assessments of PTSD often rely on subjective evaluations, which can be time-consuming, costly, and prone to human bias. This study proposes a machine learning (ML) approach based on multivariate kernel density estimation (MKDE) technique for the objective evaluation of PTSD severity. We collected heart rate (HR) and galvanic skin response (GSR) signals as well as PTSD Checklist - Military Version (PCL-M) labels from 21 participants during an immersive simulation. A fear-response model was trained on a public arachnophobia dataset, and predictive features of PTSD were extracted from the fear-response curves estimated on the military dataset. The model achieved an accuracy of 86\% in classifying PTSD status, effectively distinguishing participants with and without PTSD (PCL-M threshold of 36). The average mean absolute error (MAE) of the models is 5.6, and it estimated a clinical PTSD severity scale with a mean absolute percentage error of 17\%. Our algorithm demonstrates promising potential for enhancing estimation of PTSD severity and followup by offering an objective and low-effort evaluation approach using physiology. These findings suggest clinical utility in both screening and follow-up settings.
Authors: Enrique Alba, Ezequiel Lopez-Rubio
Abstract: We here study whether training biases can make hidden neurons specialize in minimal one-hidden-layer MLPs, and whether such specialization improves prototype-based reconstruction of the training dataset from the learned weights. We consider Gaussianactivation MLPs of width equal to dataset size and compare three structural losses that respectively encourage coverage of the training samples, separation between neuron-induced prototypes, and low overlap of hidden responses, against the standard fitting baseline. Experiments on uniformly sampled one-dimensional datasets show a stable pattern from N = 3 to N = 100 across 480 controlled runs. Coverage regularization gives the lowest mean reconstruction error at every tested size and raises the prototype-usage specialization ratio relative to the standard baseline, while separation has mixed effects and overlap penalties are systematically harmful. We show that the harm is not an optimization failure: overlap-active approaches fit the data as well as overlap-free ones but route the optimizer to a degenerate equilibrium in which prototype centers are pushed outside the convex hull of the training inputs. Coverage cannot reward this expulsion and acts as an attractor: separation admits it only at large temperature and overlap admits it at the nominal hyperparameter choice. A direct {\tau}-sweep on the separation-only mask and a prototype-position visualization at N = 100 confirm the mechanism. The findings yield a simple design principle for prototype-recoverability-aware training: every repulsive structural loss must be compensated by a compatible attractor, or it will collapse the latent geometry it was meant to refine.
Authors: Ruiqiang Xiao, Zhaohu Xing, Yijun Yang, Zhenyan Han, Weiming Wang, Kaishun Wu, Lei Zhu
Abstract: Ultrasound video segmentation is clinically valuable yet difficult due to speckle noise, weak boundaries, and rapid anatomical deformation. Recent promptable foundation models enable point-guided segmentation, but their direct deployment in ultrasound remains unreliable: a single point provides insufficient spatial context to resolve scale ambiguity, and greedy memory updates amplify early errors into severe temporal drift. We present EchoPilot, a training-free framework for ultrasound video segmentation under sparse first-frame interaction, requiring only a single point click and an anatomical category name. EchoPilot orchestrates a frozen medical vision-language model (VLM) for semantic localization, a vision foundation model (VFM) for dense geometric feature extraction, and a promptable video segmentor for mask prediction and propagation. To resolve initialization ambiguity, we propose Scale-Space Semantic Prompting, which first selects an optimal contextual view via a parameter-free S.E.E.D. (Semantic Energy-Entropy Density) criterion, and then synthesizes geometrically precise auxiliary point prompts from dense foundation features without additional user interaction. To reduce propagation drift, a Reliability-Gated Memory update is further introduced to selectively freeze the segmentor's memory bank under uncertain predictions, preventing error accumulation. We also contribute the first dynamic fetal placenta ultrasound video segmentation dataset with 671 annotated frames. Across three ultrasound video datasets, EchoPilot achieves state-of-the-art performance under the sparse-interactive setting, consistently outperforming training-free baselines and finetuned specialists.
Authors: Shyam Sankaran, Hanwen Wang, Paris Perdikaris
Abstract: Neural PDE solvers have followed the scaling trajectory of vision and language, with recent foundation models reaching billions of parameters. We argue that scale is a poor substitute for architectural inductive bias in this domain: structured priors deliver outsized parameter efficiency, and the pattern of where they succeed and fail is itself informative about what they capture. We instantiate this argument in WaveLiT, an architecture combining a discrete wavelet transform for lossless multi-resolution tokenization, an augmented linear attention block, a shared-weight multiscale feature pyramid, and a wavelet-domain auxiliary loss. Bespoke 1-10M-parameter WaveLiT models compete with foundation models of 100-1000$\times$ their size across eight TheWell benchmarks, with the largest gains on wave and acoustic-dominated benchmarks where the wavelet-multiscale prior fits the dominant dynamical structure and small per-step errors do not compound geometrically under rollout. Trained jointly across all eight benchmarks, a 10M-parameter foundation variant exhibits a structured, physically interpretable transfer pattern -- strongest where the wavelet-multiscale prior matches the dynamics, weakest on chaotic advection-dominated flows. The entire pipeline trains on a single GPU. The results suggest that small-model PDE performance is shaped by architectural inductive bias rather than scale, and that the structure of a prior's failures is a useful empirical signal about its content.
Authors: Yinghao Wu, Zhuoyan Luo, Yiyao Yu, Zhaojian Yu, Yujiu Yang, Xiao-Ping Zhang
Abstract: Despite the remarkable progress achieved by recent efficient methods in accelerating multimodal understanding, they still suffer from noticeable performance degradation. Their emphasis on the high compression ratio of a single visual clue and reliance on the heuristic pruning strategy with coarse attention alignment incurs a bottleneck on the information capacity and density of visual tokens. Addressing this limitation, we propose VEN-VL, a visual ensemble MoE framework for effective and efficient perception following the enrich then compact principle. Specifically, we first enrich the information capacity by unifying the visual representations of different perspectives, and then progressively compact it with adaptive routers in specialized visual experts to enhance the information density. Furthermore, we incorporate the reconstruction ability of vanilla structure via explicit visual supervision, facilitating crucial information preservation. Experimental results demonstrate our superiority in complex visual tasks with few information-condensed tokens, which effectively bridges the gap between performance and efficiency.
Authors: Mengfan Liu, Da Zheng, Junwei Su, Chuan Wu
Abstract: Despite the strong reasoning capabilities of large language models (LLMs), optimizing the execution efficiency of tensor programs remains challenging due to the need for precise, composable transformation decisions. Recent LLM-guided approaches frame tensor program optimization as an iterative decision process, but existing datasets provide only end-to-end optimized program pairs using token-inefficient representations, lacking verifiable step-level supervision and interpretability. As a result, LLMs struggle to make reliable single-step decisions in large combinatorial optimization spaces. We introduce Step-TP, a post-training dataset for tensor program optimization that provides grounded, atomic, step-level supervision with structured chain-of-thought (CoT) reasoning. Step-TP forms a closed reasoning loop over intermediate program states, enabling reliable multi-step optimization rather than outcome imitation. Its design is guided by four principles: (i) a token-efficient, verifiable intermediate representation (IR) that deterministically lowers to TVM TIR; (ii) atomic and composable optimization strategies that decompose complex trajectories into interpretable single-step decisions; (iii) structured CoT supervision coupled with explicit IR-to-IR state transitions; and (iv) strategy filtering to balance coverage while preventing shortcut exploitation. The dataset and implementation are available at a GitHub link, https://github.com/LIUMENGFAN-gif/StepTP.
Authors: Bo Zou, Chao Xu
Abstract: Large language models (LLMs) face a dual challenge in creative capability evaluation: existing benchmarks (e.g., Story Cloze Test, HellaSwag) measure models' discriminative ability over narrative continuation using multiple-choice recognition paradigms, rather than directly measuring creative generation capability; rubric-based scoring and LLM-as-Judge methods rely on subjective dimension assessment or natural language model outputs, and cannot provide objective, automated scoring mechanisms. This paper proposes QUIET (Quality Understanding via Interlocked Evaluation Testing), a diagnostic benchmark for LLM creative capability based on multi-blank cascaded story cloze. QUIET sets N blanks (10-20) in a story with complete structure, with each blank accompanied by an explicit content constraint, and cascade dependency relationships between blanks -- the content filled into earlier blanks constrains the feasible solution space for later blanks. The evaluated model (or human participants) fills all blanks in open-ended generation mode; the results are scored by an information-theoretic automated scoring protocol without human grading. The scoring protocol directly operationalizes the "calibrated surprise" theoretical framework (Zou & Xu, 2026a). For each blank k, a composite score is computed: score = satisfy * (1 + lambda * surprise), where lambda = 1.0. Here, "satisfy" measures how well the blank filling satisfies the content constraint (objective logical reasoning judgment, not subjective aesthetic scoring), and "surprise" measures the degree of surprise given that the constraint is satisfied. Creative answers that do not satisfy the constraint score zero; answers that satisfy the constraint but are mediocre score low; answers that satisfy the constraint and are surprising score high.
Authors: Jinju Kim, Yunsung Kang, Gyeong-Moon Park, Jong Hwan Ko
Abstract: Machine unlearning removes designated concepts or knowledge from pre-trained models. Recent work has extended this paradigm to speaker identity unlearning in zero-shot text-to-speech (ZS-TTS), the task of selectively erasing a model's ability to replicate a speaker's voice. Existing methods, however, quietly assume all unlearning requests arrive at once; an unrealistic assumption, since privacy-motivated removals arrive sequentially over time. We show this assumption breaks state-of-the-art methods: unlearning each new speaker fully revives previously unlearned speakers, reintroducing the very privacy risk unlearning was meant to eliminate. We present Cumulative ORThogonal Identity Suppression (CORTIS), the first framework for continual speaker identity unlearning in ZS-TTS that requires no access to previously-unlearned speaker data. CORTIS combines Fisher-information-based parameter masking, which localizes updates to speaker-relevant weights, with orthogonal projection against subspaces spanned by prior unlearning updates. With VoiceBox, CORTIS unlearns each requested speaker while keeping previously unlearned speakers forgotten across long request sequences, substantially outperforming sequential application of prior methods. The demo is available at https://cumulativeortis.github.io/ .
Authors: Bo Zou, Chao Xu
Abstract: This paper provides an empirical implementation of the creative quality metric proposed in Calibrated Surprise (Zou & Xu, 2026a). The question this paper addresses is: does this mathematical claim hold at the engineering level? To make the answer as general as possible, we deliberately choose the strictest engineering conditions: low data cost and a small base model. Training data comes from approximately 100 expert chain-of-thought (CoT) annotations produced by the BC Protocol (Zou & Xu, 2026b). We also identify a data bias: most publicly available alignment datasets are skewed toward craft-related knowledge, while audience modeling and reality-logic coverage are systematically weak. We use the term Creative Quality Alignment (CQA) to describe this class of engineering methods. We also offer a supporting theoretical observation: in an LLM with a single conditional distribution architecture, calibrating the appreciation side automatically transfers to the generation side via architectural duality. This is the structural reason why ~100 CoT examples are sufficient -- not a purely empirical observation like LIMA (Zhou et al., 2023).
Authors: Michael Orme, Yanchao Yu, Zhiyuan Tan
Abstract: Ensuring safe and contextually appropriate behaviour in Large Language Models (LLMs) remains a critical challenge for real-world deployment. We present \textbf{SafeCtrl-RL}, an inference-time behavioural control framework that enables adaptive safety regulation without model retraining or parameter modification. The method formulates dialogue generation as a sequential decision process, where a reinforcement learning agent dynamically selects prompt adjustment strategies based on contextual feedback. This allows unsafe behaviours to be suppressed through iterative refinement, which we conceptualise as inference-time behavioural unlearning. Evaluated across multiple LLMs and unsafe dialogue scenarios, SafeCtrl-RL consistently improves safety and response quality, outperforms existing prompt-based optimisation methods, and achieves favourable performance--efficiency trade-offs. **Warning: This paper may contain examples of harmful language, and reader discretion is recommended.
Authors: Dhruv Agarwal, Emily Sheng, Chad Atalla, Jean Garcia-Gathright, Hussein Mozannar, Hannah Washington, Alexandra Chouldechova, Solon Barocas, Hanna Wallach
Abstract: Evaluating generative AI (GenAI) systems is challenging because many targets of evaluation are broad, contested concepts, such as "reasoning," "fairness," or "creativity." When these concepts are left underspecified, it becomes unclear what should be measured or how evaluation results should be interpreted. This problem reflects a missing step: systematization, that is, moving from a broad background concept to an explicit, structured account of the concept in measurable terms. To help address the fact that systematization is cognitively demanding and resource-intensive, we investigate whether AI assistance can support this process. To enable AI-assisted systematization and assess its quality, we introduce a structured representation of a systematized concept, a concept spec, and a validation worksheet. We then develop two AI-assisted systematizers: a direct, zero-shot approach and a multi-agent approach that more closely mirrors manual systematization approaches from existing literature. We use these systematizers to produce concept specs for two concepts -- hate-based rhetoric and digital empathy -- and evaluate resulting concept specs on content validity and information recoverability.
Authors: Aleksandar Todorov, Matthia Sabatelli
Abstract: Deep reinforcement learning (RL) agents commonly rely on high-dimensional neural representations, despite growing evidence that task-relevant value and policy structure may be intrinsically low-dimensional. In this work, we present a simple yet effective representation-level prior that inserts a fixed orthonormal projection to constrain encoder features to a low-dimensional subspace, requiring no auxiliary objectives, pretraining, or changes to the underlying RL algorithm. Under a linear realizability assumption, we prove that when the bottleneck dimension exceeds the intrinsic rank of the optimal value function in feature space, the bottleneck preserves expressivity and leaves the induced gradient dynamics unchanged up to an equivalent low-dimensional parameterization. Empirically, we find that across both single and multi-task benchmarks, baseline performance is either matched or improved once the bottleneck dimension exceeds a small task-dependent threshold; in many cases, value representations can be compressed to extremely low dimensions without loss, and the minimal sufficient dimension depends far more on environment complexity than encoder width. In addition, we analyze representation geometry and find that orthogonal bottlenecks stabilize feature norms and are associated with higher effective rank. Together, these results support a representation-space interpretation of the manifold hypothesis in reinforcement learning and position orthogonal bottlenecks as a lightweight, architecture-agnostic mechanism for shaping RL representations.
Authors: Branislav Kveton, Anup Rao, Subhojyoti Mukherjee, Krishna Kumar Singh, Viet Dac Lai
Abstract: We introduce AdvantageFlow, a forward-process reinforcement learning algorithm for rectified flow models. Unlike Flow-GRPO, which optimizes the reverse process, we optimize an advantage-weighted forward-process prediction loss. This optimization problem is unstable when advantages are negative and the loss becomes non-convex. We stabilize it by rollout policy regularization, which reduces variance and arises from fitting a local reward-improving target distribution. We evaluate AdvantageFlow on image generation tasks with Stable Diffusion 3.5 Medium. It outperforms both Flow-GRPO and a state-of-the-art forward-process RL baseline based on negative-aware fine-tuning.
Authors: Christoffer Loeffler, Tom\'as Rey Pizarro, Daniel Ignacio Miranda V\'asquez, Andrea Mart\'inez Freile
Abstract: Online Terms of Service often function as contracts of adhesion, creating asymmetries that may expose consumers to potentially abusive clauses. In Chile, assessing such clauses is legally challenging because some provisions clearly violate mandatory consumer law, whereas others depend on broader standards such as good faith and contractual imbalance. We present a retrieval-augmented generation framework for the automated detection and classification of potentially abusive clauses in Chilean Terms of Service. Designed for local execution, it combines efficient clause detection, hybrid dense--sparse retrieval, reranking, and prompt augmentation to support medium-sized open-weight language models. We also introduce the Chilean Abusive Terms of Service Extended corpus, comprising 100 contracts and 10,029 annotated clauses in 24 legally grounded categories spanning illegal, dark, and gray clauses. Experiments comparing commercial and open-weight language models, fine-tuned encoders, and traditional baselines show that retrieval-augmented prompting substantially improves performance and enables local models to approach larger cloud-based systems at lower computational and token cost. The study also contributes a refined legal annotation scheme and a practical design for AI-assisted consumer contract review.
Authors: Adina Scheinfeld, Haotan Zhang, Shang Mu, Rudolf L. M. van Herten, Lucas Stoffl, Ali Erturk, Zhuhao Wu, Johannes C. Paetzold
Abstract: Light sheet fluorescence microscopy (LSM) enables high-resolution, three-dimensional (3D) imaging of biological specimens, providing rich volumetric data for studying cellular organization, pathology, and vascular networks. However, the size, dimensionality, and annotation burden of LSM data make supervised deep learning approaches costly and difficult to scale. Additionally, despite the abundance of unannotated LSM volumes, foundation models for this modality remain underexplored due to computational challenges and the complexity of volumetric representation learning. In this work, we introduce a 3D foundation model for LSM data, pretrained on a large curated collection of 3D images spanning multiple organisms, stains, and imaging protocols. We learn transferable volumetric representations by jointly optimizing for masked reconstruction and image-text alignment. The pretrained backbone drastically reduces the annotation burden, enabling efficient, few-shot adaptation for varied downstream tasks. We evaluate this approach on downstream segmentation, classification, and deblurring. Our results demonstrate consistent improvements over baselines, (1) when measured using standard evaluation metrics and (2) when rigorously assessed by domain experts. This highlights the potential of foundation model pretraining to reduce annotation requirements while improving performance across diverse LSM analysis tasks. Pretrained model weights and code for pretraining and finetuning are publicly available: https://github.com/AdinaScheinfeld/lsm_fm_public_repo.git.
URLs: https://github.com/AdinaScheinfeld/lsm_fm_public_repo.git.
Authors: Zixin Jessie Chen, Zhuo Chen, Archer Wang, Jeff Gore, William T. Freeman, Congyue Deng, Marin Solja\v{c}i\'c
Abstract: Creating images from noise is image generation; reconstructing fine details from coarse inputs is super-resolution. Despite their practical differences, both can be understood as reversing information loss across scales. We introduce $\textbf{SKILD}$, a $\textbf{S}$cale-invariant $\textbf{K}$-Space $\textbf{I}$mage $\textbf{L}$earning $\textbf{D}$iffusion model that unifies generation and continuous super-resolution within a single unconditional framework. Both natural images and critical physical systems exhibit scale invariance, and we leverage it to design a forward process that attenuates image content from fine to coarse scales while injecting spectrum-matched Gaussian noise, making scale an explicit coordinate of the diffusion dynamics. The same trained reverse process performs generation and continuous super-resolution by varying only the starting timestep: $\textit{no task-specific architecture, no conditioning branch, no classifier-free guidance, no retraining per scale factor}$. Empirically, SKILD reaches FID $2.65$ and Inception Score $9.63$ on unconditional CIFAR-10, performs $2\times$--$8\times$ super-resolution on ImageNet from a single unconditional checkpoint while outperforming conditional models across perceptual metrics, and reconstructs critical Ising models whose connected four-point correlations closely track the ground truth.
Authors: Xinrui Shi, Kai Liu, Ziqing Zhang, Jianze Li, Anqi Li, Yulun Zhang
Abstract: Lightweight vision-language models perform competitively on standard benchmarks yet fail systematically in dense-scene reasoning, where multiple objects, attributes, and relations must be jointly grounded and resolved through multi-step inference. Such capability is critical for real-world applications where models must reliably interpret cluttered environments. Yet existing training signals provide no explicit grounding between reasoning steps and the underlying visual entities and relations, leaving lightweight models free to generate fluent but visually unanchored reasoning chains. To address this gap, we first introduce DRBench, a benchmark of 14,573 questions across 2,943 images, organized into five task categories spanning three progressive reasoning layers. Building on DRBench, we propose DRScaffold, a supervised fine-tuning framework that decomposes the supervision target into four causally ordered stages, enforcing grounded reasoning without architectural modification. Experiments on three lightweight VLMs demonstrate substantial gains on DRBench while preserving or improving performance on general-purpose benchmarks. Notably, Qwen2.5-VL-3B trained with DRScaffold surpasses the frozen Qwen2.5-VL-32B on DRBench, demonstrating that structured supervision can substitute for a significant portion of model scale in dense-scene reasoning. Our code and models are available at https://github.com/irene-shi/DRScaffold .
Authors: Federico Torrielli, Peter Schneider-Kamp, Lukas Galke Poech
Abstract: Activation oracles aim to make the activations of other models legible to humans and yield promising results compared to white-box interpretability techniques. However, uncertainty quantification (UQ) for the natural-language outputs of such activation oracles is so far understudied. Here, we investigate 6 different methods for estimating the confidence of activation oracles and evaluate how well-calibrated their confidence scores are. Our experiments on 6,000 samples per oracle (varying verbalizer and context prompts) reveal that bootstrap mode frequency is the best-calibrated method among those tested (ECE 5.7% vs. 25.5% for the answer-word log-probability on Qwen3-8B; 10.3% vs. 13.1% on Qwen3.6-27B), and that the log-prob baseline can serve as a fast triage signal at a fraction of the cost. Code and the patched trainer are available at https://github.com/federicotorrielli/probabilistic_activation_oracles.
URLs: https://github.com/federicotorrielli/probabilistic_activation_oracles.
Authors: Parth Darshan, Abhishek Divekar
Abstract: Customizing an LLM judge to a specific task or domain often involves optimizing its prompt across multiple evaluation criteria simultaneously. Textual gradient methods automate this for a single judge criterion, however they produce natural-language critiques, not numerical vectors. Thus, the conflict-resolution toolkit of multi-task learning (PCGrad, MGDA) doesn't apply to the multi-objective textual gradient setting. We test five decomposition modes of textual gradient optimizers by varying how much cross-task information the loss, gradient and optimizer LLMs share. In 6 of 10 configurations, we observe that optimization never improves over the initial prompt. Gradient specificity drops by 59% (from 9.0 to 3.7) when the gradient LLM processes multiple criteria jointly. Separately, we observe that naively combining per-task instructions into a single prompt degrades Spearman's rho by -5.3%. These results identify two separable failure modes: optimization-time gradient dilution and inference-time instruction interference, which together constrain the design space for multi-objective judge customization using textual feedback.
Authors: Waleed Razzaq, Yun-Bo Zhao
Abstract: Reliable quantification of uncertainty estimates in continuous-time (CT) representation learning remains nascent, particularly within CT attention architectures. We introduce the Neuronal Stochastic Attention Circuit (NSAC), a novel biologically-inspired CT attention architecture that reformulates attention logit computation as the solution of an Ornstein-Uhlenbeck stochastic differential equation modulated by input-dependent, nonlinear interlinked gates derived from repurposed C.elegans Neuronal Circuit Policies (NCPs) wiring mechanism. It induces Gaussian distribution over logits that propagates principled stochasticity through logistic-normal distribution over attention weights to yield probabilistic output. A two-term objective function combining Gaussian negative log-likelihood with an epistemic-separation regularizer enforces higher predictive variance and enables joint quantification of aleatoric and epistemic uncertainty. Empirically, we implement NSAC in a diverse set of learning tasks including: (i) irregular CT function approximation; (ii) multivariate regression; (iii) long-range forecasting; (iv) Industry 4.0; and (v) the lane-keeping of autonomous vehicles. We observe that the NSAC remains competitive against several baselines in terms of accuracy and produces reasonably well-calibrated uncertainty estimates while being interpretable at the neuronal cell level.
Authors: Rustem Takhanov, Zhenisbek Assylbekov
Abstract: Conditionally positive definite (CPD) kernels are defined with respect to a function class $\mathcal{F}$. It is well known that such a kernel $K$ is associated with its native space (defined analogously to an RKHS), which in turn gives rise to a learning method -- called conditional kernel ridge regression (conditional KRR) due to its analogy with KRR -- where the estimated regression function is penalized by the square of its native space norm. This method is of interest because it can be viewed as classical linear regression, with features specified by $\mathcal{F}$, followed by the application of standard KRR to the residual (unexplained) component of the target variable. Methods of this type have recently attracted increasing attention. We study the statistical properties of this method by reducing its behavior to that of KRR with another fixed kernel, called the residual kernel. Our main theoretical result shows that such a reduction is indeed possible, at the cost of an additional term in the expected test risk, bounded by $\mathcal{O}(1/\sqrt{N})$, where $N$ is the sample size and the hidden constant depends on the class $\mathcal{F}$ and the input distribution. This reduction enables us to analyze conditional KRR in the case where $K$ is positive definite and $\mathcal{F}$ is given by the first $k$ principal eigenfunctions in the Mercer decomposition of $K$. We also consider the setting where $\mathcal{F}$ consists of $k$ random features from a random feature representation of $K$. It turns out that these two settings are closely related. Both our theoretical analysis and experiments confirm that conditional KRR outperforms standard KRR in these cases whenever the $\mathcal{F}$-component of the regression function is more pronounced than the residual part.
Authors: Xu Yao, Siyuan Zhou, Zhenbo Wu, Chaochuan Hou, Shuang Liang, Shiping Wang, Hailiang Huang, Songqiao Han, Minqi Jiang
Abstract: Weakly supervised anomaly detection (WSAD) has developed in three primary directions: incomplete, inexact, and inaccurate supervision. However, these directions remain isolated, lacking a unified framework to assess whether they address unique challenges or share fundamental mechanics. This paper introduces WSADBench, the first benchmark that unifies evaluation across distinct weakly supervised scenarios, benchmarking diverse approaches from specialized WSAD methods to advanced tabular foundation models. WSADBench establishes standardized protocols to evaluate 36 algorithms across 4 modalities by systematically varying label quantity, granularity, and quality, revealing the performance boundaries of various methods. Based on over 700K experiments, WSADBench reveals four critical insights: (i) Strong intrinsic correlations exist between these weak supervision scenarios, challenging the isolation of current research directions. (ii) Specialized WSAD algorithms excel only in extreme label-scarcity regimes but are quickly dominated by tabular foundation models and general classification methods as supervision increases or in OOD scenarios. (iii) Unlabeled data shows inconsistent utility across settings, with marginal gains compared to label refinement. (iv) Models exhibit asymmetric sensitivity to different types of label noise. We release WSADBench as an open-source benchmark with code and datasets to facilitate future WSAD research: https://github.com/SUFE-AILAB/WSADBench.
Authors: Yunhua Pei, Jingyu Hu, Yiwei Shi, Hongnan Ma, Weiru Liu, John Cartlidge
Abstract: Existing financial NLP benchmarks often rely on labels supplied by outside observers, measuring how language is perceived rather than what speakers have committed to in the market. We introduce StakeBench, an evaluation framework for language understanding grounded in market commitment. StakeBench links 560,876 comments from 2,261 resolved markets to verified position, action, and market-odds records across Polymarket and Manifold. Supervision is derived from observable market behavior. Position sides, post-comment trading actions, and market-odds trajectories replace human annotation. Four diagnostic tasks test whether models detect market commitment, identify the revealed side, anticipate future action, and perform collective odds projection. Three commitment-aware metrics measure alignment with revealed preferences rather than perceived sentiment. Validity audits and explicit interpretation boundaries help distinguish observable commitment signals from latent belief and causal market-odds impact. Across 15 LLMs and 18 topics and platform settings, models partially recover position-side signals, with Directed Accuracy from 0.506 to 0.599, but show structural failures on later tasks. Ten of the fifteen models collapse to one or two action labels in future action anticipation, and no model consistently improves on the naive odds-direction baseline in collective odds projection. Model scale is not correlated with performance, finance-domain tuning does not improve revealed-side identification, and platform incentives strongly shape higher-order results. StakeBench is packaged with evaluation code and dataset under CC-BY 4.0.
Authors: Wei Song, Tianhang Wang, Yitong Chen, Tong Zhang, Zuxuan Wu, Ming Li, Jiaqi Wang, Kaicheng Yu
Abstract: We present Channel-wise Vector Quantization (CVQ), a novel image tokenization paradigm that replaces patch-wise tokens with channel-wise tokens. Unlike conventional vector quantization, which assigns a discrete token to each patch feature vector, CVQ quantizes each channel of the feature map. This formulation represents an image as discrete levels of visual details, rather than as a grid of spatial patches. Based on CVQ, we introduce a new visual autoregressive framework with "next-channel prediction". Instead of rendering images patch by patch in raster order, our Channel-wise Autoregressive (CAR) model predicts image channels sequentially, producing progressively enriched visual details. Specifically, it first sketches global structure and then refines fine-grained attributes, akin to a human artist's workflow. Empirically, we show that: (1) CVQ achieves 100% codebook utilization with a 16K+ codebook size without any bells and whistles, and substantially improves reconstruction quality over conventional VQ; and (2) CAR attains a DPG score of 86.7 and a GenEval score of 0.79, demonstrating strong effectiveness for text-to-image generation.
Authors: Maoyang Xiang, Bo Wang, Tao Luo
Abstract: The deployment of Large Language Models (LLMs) and Vision Transformers (ViTs) on edge devices is significantly constrained by memory limitations and the critical timing bottlenecks introduced by dense Multiply-Accumulate (MAC) arrays. In the ultra-low bit regime, logarithmic Power-of-Two (PoT) quantization provides a hardware-efficient alternative by replacing MAC operations with bit-shifts. However, the non-uniform exponential lattice is inherently limited by a \textbf{Low Angular Resolution Regime}, a structural flaw that becomes particularly pronounced at sub-4-bit thresholds, leading to a notable degradation of high-dimensional feature manifolds. To address this geometric limitation, we propose Orthogonal Residual Projection (ORP), an algorithm-hardware co-design framework. By formulating quantization as a dual-basis geometric projection, ORP adaptively synthesizes a higher-resolution residual lattice using strictly shift-and-add operations. Furthermore, ORP's analytical solver offers a practical alternative to computationally intensive gradient-based optimization, reducing the full-model calibration time for LLaMA-2-7B to approximately \textbf{15 minutes}. Extensive evaluations demonstrate ORP's applicability across modalities and its hardware efficiency. Under the 3-bit (W3/A16) constraint, ORP achieves a perplexity of 6.10 on LLaMA-2-7B, comparing favorably to conventional MAC-intensive baselines like AWQ without relying on asymmetric scaling, while maintaining competitive accuracy in 4-bit scenarios. At the silicon level, standard-cell RTL synthesis at a 28nm node indicates that ORP effectively mitigates the timing bottlenecks associated with dense multiplier trees.
Authors: Sangyun Lee, Sean McLeish, Tom Goldstein, Giulia Fanti
Abstract: Transformer-based large language models are increasingly used for long-horizon tasks; however, their attention mechanism scales poorly with context length. To handle this, we study a sleep-like consolidation mechanism in which a model periodically converts recent context into persistent fast weights before clearing its key-value cache. During sleep, the model performs $N$ offline recurrent passes over the accumulated context and updates the fast weights in its state-space model (SSM) blocks through a learned local rule. During inference, this shifts extra computation to sleep while preserving the latency of wake-time prediction. We test our method on controlled synthetic tasks, including cellular automata and multi-hop graph retrieval, as well as a realistic math reasoning task, on which a regular transformer as well as SSM-attention hybrid models fail. We then show that increasing sleep duration $N$ for our models improves performance, with the largest gains on examples that require deeper reasoning.
Authors: Bar Weiss, Antonio Abu-Nassar, Adi Sosnovich, Karen Yorav
Abstract: Code review is a critical practice in software engineering, yet the growing scale and frequency of code patches in modern projects, together with the widespread adoption of AI code assistants, make manual review increasingly challenging. Identifying the types of changes within a patch, such as renames, moves, or logic modifications, can substantially improve review efficiency by enabling prioritization, filtering, and automation. However, existing LLM-based approaches to code review have largely focused on summarization and comment generation, leaving structured code reviews underexplored. In this paper, we present a systematic study of using large language models (LLMs) for taxonomy-based labeling of code changes in a code patch. We introduce a two-stage pipeline that assigns labels to diff hunks and then refines them to capture structural relationships and semantic attributes, such as rename propagation and type changes. Our approach employs few-shot prompting to produce language-agnostic and customizable labels, without the engineering overhead of traditional static-analysis pipelines. We evaluate four LLMs across multiple context configurations on a manually curated benchmark of natural and synthetic patches. Our best configuration achieves up to $84\%$ recall and $81\%$ precision, with high accuracy in extracting relational and attribute metadata. These results suggest that LLM-based labeling can effectively complement static analysis by enabling flexible, multilingual, and automation-friendly code review workflows.
Authors: Shuhong Zheng, Aashish Kumar Misraa, Yu-Teng Li, Yu-Jhe Li, Igor Gilitschenski
Abstract: Subject-driven image generation aims to synthesize new images that preserve the identity of the given subject while following textual instructions. Existing approaches often encode text and reference images separately. This limits cross-modal reasoning abilities and causes copy-paste artifacts. Recent frameworks that connect multimodal models and diffusion models improve instruction following, but largely overlook identity preservation. To address these limitations, we condition diffusion models on Multimodal Large Language Models (MLLMs) that jointly encode text and reference images, and augment it with VAE-based identity conditioning. A novel Dual Layer Aggregation (DLA) module is designed to aggregate multi-level MLLM features for optimal conditioning, and a multi-stage denoising strategy is applied to progressively balance the semantic information from MLLM and fine-detail identity from VAE during inference. Extensive experiments demonstrate that our approach harmonizes multimodal understanding with identity preservation, mitigates copy-paste issues, and achieves superior performance regarding human preference on subject-driven image generation. Our project website is available at https://zsh2000.github.io/squeeze-mllm-subject-gen/.
Authors: Henry Hengyuan Zhao, Kaiming Yang, Wendi Yu, Difei Gao, Mike Zheng Shou
Abstract: Recent progress in GUI agents has substantially improved visual grounding, yet robust planning remains challenging, particularly when the environment deviates from a canonical initial state. In real applications, users often invoke assistance mid-workflow, where software may be partially configured, steps may have been executed in different orders, or the interface may differ from its default setup. Such task-state variability is pervasive but insufficiently evaluated in existing GUI benchmarks. To address this gap, we introduce WorldGUI, a benchmark covering ten widely used desktop and web applications with tasks instantiated under diverse, systematically constructed initial states. These variations capture realistic human-computer interaction settings and enable diagnostic evaluation of an agent's ability to recover, adapt plans, and handle non-default contexts. We further present WorldGUI-Agent, a simple and model-agnostic framework that organizes planning and execution around three critique stages, improving reliability in dynamic environments. Experiments demonstrate that state-of-the-art GUI agents exhibit substantial performance degradation under non-default initial conditions, revealing limited robustness and fragile planning behaviors. Our benchmark and framework provide a foundation for developing more adaptable and reliable GUI agents. The code and data are available at https://github.com/showlab/WorldGUI.
Authors: In-Chang Baek, Sung-Hyun Kim, Sam Earle, Zehua Jiang, Jin-Ha Noh, Julian Togelius, Kyung-Joong Kim
Abstract: Reward design plays a pivotal role in the training of game AIs, requiring substantial domain-specific knowledge and human effort. In recent years, several studies have explored reward generation for training game agents and controlling robots using large language models (LLMs). In the content generation literature, there has been early work on generating reward functions for reinforcement learning agent generators. This work introduces PCGRLLM, an extended architecture based on earlier work, which employs a feedback mechanism and several reasoning-based prompt engineering techniques. We evaluate the proposed method on a story-to-reward generation task in a two-dimensional environment using two state-of-the-art LLMs across various reasoning-based prompting methods. Our experiments provide insightful evaluations that demonstrate the capabilities of LLMs essential for content generation tasks. The results demonstrate a substantial performance improvement over the previous structure, achieving performance comparable to that of humans. Our work demonstrates the potential to reduce human dependency in game AI development, while supporting and enhancing creative processes.
Authors: Weizhi Fei, Zihao Wang, hang Yin, Shukai Zhao, Wei Zhang, Yangqiu Song
Abstract: Complex Query Answering (CQA) is a crucial reasoning task over Knowledge Graphs (KGs), which aims to answer first-order logical queries from incomplete KGs. While existing neural-symbolic methods achieve strong performance, they face significant complexity bottlenecks: quadratic data complexity scaling with the number of entities, and NP-hard query complexity for cyclic queries. Consequently, these approaches struggle to scale effectively to large knowledge graphs and complex queries. To address these limitations, we propose an efficient and scalable symbolic search method comprising two key components: (1) constraint strategies that drastically reduce the variable search domain, lowering data complexity; and (2) a local search algorithm that approximately solves NP-hard cyclic queries. Experiments on various CQA benchmarks demonstrate that, for tree-form queries, our method achieves 97% relative MRR with a 10$\times$ speedup using only 10% of the search space. Furthermore, it demonstrates robust performance on complex cyclic queries and large-scale KGs, effectively alleviating efficiency and scalability challenges. Our code is provided in https://github.com/HKUST-KnowComp/NLISA_KDD2026.
Authors: Fedor Rodionov, Abdelrahman Eldesokey, Michael Birsak, John Femiani, Bernard Ghanem, Peter Wonka
Abstract: We introduce FloorplanQA, a diagnostic benchmark for evaluating spatial reasoning in large language models (LLMs). FloorplanQA is grounded in structured representations of indoor scenes, such as (e.g., kitchens, living rooms, bedrooms, bathrooms, and others), encoded symbolically in JSON or XML layouts. The benchmark covers core spatial tasks, including distance measurement, visibility, path finding, and object placement within constrained spaces. Our results across a variety of frontier open-source and commercial LLMs reveal that while models may succeed in shallow queries, they often fail to respect physical constraints, preserve spatial coherence, though they remain mostly robust to small spatial perturbations. FloorplanQA uncovers a blind spot in today's LLMs: inconsistent reasoning about indoor layouts. We hope this benchmark inspires new work on language models that can accurately infer and manipulate spatial and geometric properties in practical settings.
Authors: Ronghua Shi, Yiou Liu, Yuchun Feng, Lynn Ai, Bill Shi, Zhuang Liu
Abstract: Decentralized finance (DeFi) has introduced a new era of permissionless financial innovation but also led to unprecedented market manipulation. Without centralized oversight, malicious actors coordinate shilling campaigns and pump-and-dump schemes across various platforms. We propose a Multi-Agent Reinforcement Learning (MARL) framework for decentralized manipulation detection, modeling the interaction between manipulators and detectors as a dynamic adversarial game. This framework identifies suspicious patterns using delayed token price reactions as financial indicators.Our method introduces three innovations: (1) Group Relative Policy Optimization (GRPO) to enhance learning stability in sparse-reward and partially observable settings; (2) a theory-based reward function inspired by rational expectations and information asymmetry, differentiating price discovery from manipulation noise; and (3) a multi-modal agent pipeline that integrates LLM-based semantic features, social graph signals, and on-chain market data for informed decision-making.The framework is integrated within the Symphony system, a decentralized multi-agent architecture enabling peer-to-peer agent execution and trust-aware learning through distributed logs, supporting chain-verifiable evaluation. Symphony promotes adversarial co-evolution among strategic actors and maintains robust manipulation detection without centralized oracles, enabling real-time surveillance across global DeFi ecosystems.Trained on 100,000 real-world discourse episodes and validated in adversarial simulations, Hide-and-Shill achieves top performance in detection accuracy and causal attribution. This work bridges multi-agent systems with financial surveillance, advancing a new paradigm for decentralized market intelligence. All resources are available at the Hide-and-Shill GitHub repository to promote open research and reproducibility.
Authors: Tatiana Petrova (SEDAN SnT, University of Luxembourg, Luxembourg, Luxembourg), Boris Bliznioukov (SEDAN SnT, University of Luxembourg, Luxembourg, Luxembourg), Aleksandr Puzikov (SEDAN SnT, University of Luxembourg, Luxembourg, Luxembourg), Radu State (SEDAN SnT, University of Luxembourg, Luxembourg, Luxembourg)
Abstract: The Web of Agents (WoA) transforms the document-centric Web into an environment of autonomous agents acting on users' behalf, a vision newly tractable as large language models (LLMs) mature. We argue that across three decades the WoA has undergone a \emph{semantic-effort migration} in chronological order: from platform-side coordination (Multi-Agent Systems, Generation~I), through data-side annotation (Semantic Web, Generation~II), to model-side interpretation (LLM-era, Generation~III). The central Gen~II~$\rightarrow$~Gen~III transition within this trajectory, which we call the \emph{semantics-in-data $\rightarrow$ semantics-in-models} shift, is predictive: each generation's failure modes and current open problems follow from where that generation located its semantic effort. The survey makes five contributions: (i)~a unified evolutionary narrative spanning 1990--2026; (ii)~a four-dimensional comparative framework (semantic foundation, communication paradigm, locus of intelligence, discovery mechanism) applied uniformly across all three generations; (iii)~classification of sixteen representative systems on these dimensions, including hybrid LLM--knowledge-graph and computer-use agents; (iv)~coverage of the November~2024--August~2026 institutional convergence (Linux Foundation's Agentic AI Foundation, A2A v1.0, MCP November~2024 launch and November~2025 specification, Visa/Mastercard/Stripe payment-network protocols, EU AI Act phased enforcement, the NIST AI Agent Standards Initiative, International AI Safety Report 2026); and (v)~seven named lessons grounded in cross-generational evidence paired with seven generation-invariant challenges that persist regardless of which protocol prevails. Further progress depends less on protocol design than on the socio-technical infrastructure now being assembled by standards bodies, regulators, and commercial payment networks.
Authors: Dayoon Ko, Jihyuk Kim, Haeju Park, Sohyeon Kim, Dahyun Lee, Yongrae Jo, Gunhee Kim, Moontae Lee, Kyungjae Lee
Abstract: Large reasoning models (LRMs) combined with retrieval-augmented generation (RAG) have enabled deep research agents capable of multi-step reasoning with external knowledge retrieval. However, we find that existing approaches rarely demonstrate test-time search scaling. Methods that extend reasoning through single-query sequential search suffer from limited evidence coverage, while approaches that generate multiple independent queries per step often lack structured aggregation, hindering deeper sequential reasoning. We propose a hybrid search strategy to address these limitations. We introduce HybridDeepSearcher, a structured search agent that integrates parallel query expansion with explicit evidence aggregation before advancing to deeper sequential reasoning. To supervise this behavior, we introduce HDS-QA, a novel dataset that guides models to combine broad parallel search with structured aggregation through supervised reasoning-query0retrieval trajectories containing parallel sub-queries. Across five benchmarks, HybridDeepSearcher significantly outperforms the state-of-the-art, improving F1 scores by +15.9 on FanOutQA and +9.2 on a subset of BrowseComp. Further analysis shows its consistent test-time search scaling: performance improves as additional search turns or calls are allowed, while competing methods plateau.
Authors: Valen Tagliabue, Leonard Dung
Abstract: We develop new experimental paradigms for measuring welfare in language models. We compare verbal reports of models about their preferences with preferences expressed through behavior when navigating a virtual environment and selecting conversation topics. We also test how costs and rewards affect behavior and whether responses to an eudaimonic welfare scale - measuring states such as autonomy and purpose in life - are stable across semantically equivalent prompts. Overall, we observed a notable degree of mutual support between our measures. The reliable correlations observed between stated preferences and behavior across conditions suggest that preference satisfaction can, in principle, serve as an empirically measurable welfare proxy in some of today's AI systems. Furthermore, our design offered an illuminating setting for qualitative observation of model behavior. Yet, the consistency between measures was more pronounced in some models and conditions than others and responses were changed by perturbations. Due to this, and the background uncertainty about the nature of welfare and the cognitive states (and welfare subjecthood) of language models, we are currently uncertain whether our methods successfully measure the welfare state of language models. Nevertheless, these findings highlight the feasibility of welfare measurement in language models, inviting further exploration.
Authors: Thomas A. Buckley, Riccardo Conci, Peter G. Brodeur, Jason Gusdorf, Sourik Beltr\'an, Bita Behrouzi, Byron Crowe, Jacob Dockterman, Muzzammil Muhammad, Sarah Ohnigian, Andrew Sanchez, James A. Diao, Aashna P. Shah, Daniel Restrepo, Eric S. Rosenberg, Andrew S. Lea, Emily Glanton, Kimberly LeBlanc, Undiagnosed Diseases Network, Marinka Zitnik, Scott H. Podolsky, Zahir Kanjee, Raja-Elie E. Abdulnour, Jacob M. Koshy, Adam Rodman, Arjun K. Manrai
Abstract: Differential diagnosis is an iterative process that integrates patient information with broader medical knowledge. Clinical case series such as the NEJM Clinicopathologic Conferences (CPCs), published continuously since 1923, feature expert physicians who demonstrate diagnostic reasoning to peers, and have been used for decades to evaluate AI. However, prior AI evaluations have largely focused on final diagnostic accuracy rather than nuanced clinical reasoning. Here, we introduce Dr. CaBot, an agentic AI system that emulates an expert diagnostician by generating written and narrated slide-based presentations from an initial case description alone. CaBot recently generated the first AI diagnosis published in the 100+ year history of the NEJM CPCs. In blinded evaluations, physicians misclassified the source of the differential (CaBot vs. physician-written) in 46/62 (74%) of trials and rated them favorably across quality dimensions. When tasked with solving cases for 72 patients with undiagnosed disease from the NIH Undiagnosed Diseases Network, CaBot identified the working diagnosis in 50/72 (69%) of cases from referral notes alone. To promote transparency and research, we also developed CPC-Bench, a physician-validated benchmark based on 7,102 CPCs and 47,648 questions across 10 tasks. We show that CaBot outperforms frontier models on CPC-Bench, and release both CaBot and CPC-Bench publicly to foster progress in clinical AI.
Authors: Wonjoong Kim, Sangwu Park, Yeonjun In, Sein Kim, Dongha Lee, Chanyoung Park
Abstract: Although recent tool-augmented benchmarks involve complex requests, evaluation remains limited to answer matching, neglecting critical trajectory aspects like efficiency, hallucination, and adaptivity. The most straightforward method for evaluation is to compare an agent's trajectory with the ground-truth, but annotating all valid ground-truth trajectories is prohibitively expensive. In this manner, we introduce TRACE, a reference-free framework for the multi-dimensional evaluation of tool-augmented LLMs. By incorporating an evidence bank which accumulates knowledge from preceding steps, TRACE assesses an agent's reasoning trajectory effectively. To validate our framework, we develop a new meta-evaluation dataset with diverse and flawed trajectories, each labeled with multi-faceted performance scores. Our results confirm that TRACE accurately evaluates complex trajectories even with small open-source LLMs. Furthermore, we apply our method to evaluate the trajectories that agents produce while solving tool-augmented tasks, presenting previously unreported observations and their corresponding insights.
Authors: Kai Zhang, Xiangchao Chen, Bo Liu, Tianci Xue, Zeyi Liao, Zhihan Liu, Xiyao Wang, Yuting Ning, Zhaorun Chen, Xiaohan Fu, Jian Xie, Yuxuan Sun, Boyu Gou, Qi Qi, Zihang Meng, Jianwei Yang, Ning Zhang, Xian Li, Ashish Shah, Dat Huynh, Hengduo Li, Zi Yang, Sara Cao, Lawrence Jang, Shuyan Zhou, Jiacheng Zhu, Huan Sun, Jason Weston, Yu Su, Yifan Wu
Abstract: A long-term goal of language agents is to learn and improve through their own experience, ultimately outperforming humans in complex, real-world tasks. However, training agents from experience data with reinforcement learning remains difficult in many environments, which either lack verifiable rewards (e.g., websites) or require inefficient long-horizon rollouts (e.g., multi-turn tool use). As a result, most current agents rely on supervised fine-tuning on expert data, which is challenging to scale and generalizes poorly. This limitation stems from the nature of expert demonstrations: they capture only a narrow range of scenarios, and expose the agent to limited environment diversity. We address this limitation with a middle-ground paradigm we call early experience: interaction data generated by the agent's own actions, where the resulting future states serve as supervision without reward signals. Within this paradigm, we study two strategies of using such data: (1) implicit world modeling, which uses collected states to ground the policy in environment dynamics; and (2) self-reflection, where the agent learns from its suboptimal actions to improve reasoning and decision-making. Evaluation across eight diverse environments and multiple model families shows that our approaches consistently improve effectiveness and out-of-domain generalization, highlighting the value of early experience. Moreover, in environments with verifiable rewards, our results provide promising signals that early experience offers a strong foundation for subsequent reinforcement learning, making it a practical bridge between imitation learning and fully experience-driven agents.
Authors: Akira Okutomi
Abstract: High-confidence errors in large language models are often treated as fragile failures. We study an alternative: some errors may be false fixed points, locally stable, internally coherent, and confidently wrong. This separates robustness from truth-tracking. We develop the separation through a Kantian commitment-gate framing and a minimal linear feedback model in which stability and correctness can diverge. Across three open-weight models, overconfident wrong items are not systematically more locally fragile than confidently correct items under our hidden-state sensitivity probes. Abstention-aware self-critique reduces overconfident wrong commitments by sacrificing coverage, and C3-R, a rule-based explicit feedback gate, sharpens that tradeoff rather than eliminating it. These results motivate, but do not establish, high signal-to-noise (high-SNR) inertia and representational compression as possible mechanisms for stable miscalibration.
Authors: Boyin Liu, Zhuo Zhang, Sen Huang, Lipeng Xie, Qingxu Fu, Haoran Chen, LI YU, Tianyi Hu, Zhaoyang Liu, Bolin Ding, Dongbin Zhao
Abstract: Reinforcement Learning from AI Feedback (RLAIF) relies on LLM judges as preference measurement instruments, yet these instruments are fundamentally limited by random measurement errors -- stochastic fluctuations that manifest as preference cycles (e.g., $A \succ B \succ C \succ A$), occurring in 5-9% of evaluations across state-of-the-art models. While repeated sampling mitigates noise by averaging multiple judgments, it treats each comparison in isolation and fails to exploit the structural constraints that distinguish systematic signals from random noise. We introduce Topological Consensus Rewards (TCR), a framework that leverages transitivity as a denoising mechanism via topological majority voting: systematic signals reinforce each other through transitive chains, while random errors cluster into topologically exposed cycles. TCR approximates the Maximum Acyclic Subgraph to filter stochastic noise from preference signals. We also propose Cycle Incidence Rate (CIR) as a diagnostic metric that measures the proportion of samples containing preference cycles. Under our noise model, these cycles arise primarily from stochastic measurement errors rather than genuine intransitivity. Experiments on Arena-Hard, MT-Bench, and WritingBench demonstrate that TCR consistently outperforms pairwise baselines and classical ranking algorithms, while exhibiting robust performance across different judge models.
Authors: Qiuli Wang, Xinhuang Sun, Yonglin Chen, Jie Cheng, Yongxu Liu, Xingpeng Zhang, Xiaoming Li, Wei Chen
Abstract: Large language models (LLMs) have demonstrated promising performance in generating diagnostic conclusions from imaging findings, thereby supporting radiology reporting, trainee education, and quality control. However, systematic guidance on how to optimize prompt design across different clinical contexts remains underexplored. Moreover, a comprehensive and standardized framework for assessing the trustworthiness of LLM-generated radiology reports is yet to be established. This study aims to enhance the trustworthiness of LLM-generated liver MRI reports by introducing a Multi-Dimensional Credibility Assessment (MDCA) framework and providing guidance on institution-specific prompt optimization. The proposed framework is applied to evaluate and compare the performance of several advanced LLMs, including Kimi-K2-Instruct-0905, Qwen3-235B-A22B-Instruct-2507, DeepSeek-V3, and ByteDance-Seed-OSS-36B-Instruct, using the SiliconFlow platform.
Authors: Yongjae Lee, Taekhyun Park, Sunghyun Sim, Hyerim Bae
Abstract: Recent advances in sparse reward policy gradient methods have enabled effective reinforcement learning (RL)-based language model post-training. However, for reasoning tasks such as mathematical problem solving, binarized outcome rewards provide limited feedback on intermediate reasoning steps. While some studies have attempted to address this issue by estimating overall reasoning quality, it remains unclear whether these rewards are reliable proxies for the quality of stepwise reasoning. In this study, we consider reasoning as a structured process and propose TACReward, the reward model that can be seamlessly integrated into sparse reward policy gradient methods without additional human annotation costs or architectural modifications. TACReward aggregates stepwise structural deviations between teacher and policy reasoning using process mining techniques, producing a scalar output reward range of [0, 1] to indicate reasoning quality. Experiments on multiple mathematical reasoning benchmarks demonstrate that integrating the TACReward into sparse reward frameworks encourages the policy model to improve the structural quality of reasoning. Consequently, this leads to consistent performance improvements over existing sparse reward frameworks. Our code and checkpoints are publicly available at https://github.com/Thrillcrazyer/TACReward and https://huggingface.co/Thrillcrazyer/TACReward7B.
URLs: https://github.com/Thrillcrazyer/TACReward, https://huggingface.co/Thrillcrazyer/TACReward7B.
Authors: Jianli Zhao, Tingchen Fu, Rylan Schaeffer, Mrinank Sharma, Fazl Barez
Abstract: Large Reasoning Models (LRMs) improve task performance through extended inference-time reasoning. Although previous studies suggest that longer reasoning should lead to more robust safety behavior, we find evidence to the contrary: over-extended reasoning can instead be exploited to systematically weaken refusal behavior. We propose Chain-of-Thought Hijacking, a simple yet effective black-box jailbreak attack that induces LRMs to engage in prolonged benign puzzle-solving reasoning, often lasting more than five minutes, before eliciting harmful compliance. Across HarmBench, CoT Hijacking achieves attack success rates of 99%, 94%, 100%, and 94% on Gemini 2.5 Pro, ChatGPT o4 Mini, Grok 3 Mini, and Claude 4 Sonnet, respectively. To understand why this attack succeeds, we conduct activation probing, attention-pattern analysis, and causal interventions on open-source reasoning models. Our results indicate that refusal behavior depends on a low-dimensional safety signal whose expression weakens as reasoning traces grow longer. In particular, extended benign reasoning shifts attention away from harmful intentions and attenuates refusal-related activations, producing what we call refusal dilution. These findings demonstrate that excessively prolonged reasoning can introduce a systematic jailbreak attack surface. We release our evaluation materials to support reproducibility and further research.
Authors: Zihang Ding, Amit Kumar, Imran Md. Azizul Islam, Mila Avellar Montezuma, Ruihang Zhang, Kun Zhang
Abstract: Urban flooding triggered by intense rainfall is becoming increasingly frequent and widespread. While flood prediction and monitoring in high spatio-temporal resolution are desired, practical constraints in time, budget, and technology hinder its full implementation. How to monitor urban drainage networks and predict flow conditions under constrained resources is a major challenge. To address this, we introduced a data-driven sparse sensing (DSS) approach, demonstrated via a digital-twin of the Woodland catchment in Duluth, Minnesota. Specifically, we coupled EPA-SWMM with singular value decomposition and QR factorization-based sensor selection to optimize monitoring locations for system-level flow reconstruction. An ensemble of SWMM simulations, driven by diverse scenarios, provided the necessary hydraulic data to extract the reduced basis and identify informative sensor locations. Cross-event validation showed that three strategically placed sensors among 77 candidate nodes achieved a mean system-level Nash-Sutcliffe efficiency (NSE) of 0.949 across observed storm events. The QR-selected sensor sets were benchmarked against reference sensor configurations obtained from exhaustive searches and Monte Carlo random-placements. This comparison further showed that flow reconstruction based on QR-selected sensors closely tracked the exhaustive optimum while substantially outperforming random placements. We further evaluated the framework's robustness by introducing multiplicative Gaussian noise and simulating individual sensor failures. While the model is relatively resilient to noise, the impact of sensor dropouts depends heavily on the number of sensors allocated and their specific locations.
Authors: Dylan M. Asmar, Mykel J. Kochenderfer
Abstract: Autonomous agents operating in sequential decision-making tasks under uncertainty can benefit from external action suggestions, which provide valuable guidance but inherently vary in reliability. Existing methods for incorporating such advice typically assume static and known suggester quality parameters, limiting practical deployment. We introduce a framework that dynamically learns and adapts to varying suggester reliability in partially observable environments. First, we integrate suggester quality directly into the agent's belief representation, enabling agents to infer and adjust their reliance on suggestions through Bayesian inference over suggester types. Second, we introduce an explicit ``ask'' action allowing agents to strategically request suggestions at critical moments, balancing informational gains against acquisition costs. Experimental evaluation demonstrates robust performance across varying suggester qualities, adaptation to changing reliability, and strategic management of suggestion requests. This work provides a foundation for adaptive human-agent collaboration by addressing suggestion uncertainty in uncertain environments.
Authors: Mingyu Zhang, Lifeng Zhuo, Tianxi Tan, Guocan Xie, Xian Nie, Yan Li, Renjie Zhao, Zizhu He, Ziyu Wang, Jiting Cai, Yong-Lu Li
Abstract: Humans learn by observing, interacting with environments, and internalizing physics and causality. Here, we aim to ask whether an agent can similarly acquire human-like reasoning from interaction and keep improving with more experience. To study this, we introduce a Game-to-Unseen (G2U) benchmark of 1,000+ heterogeneous games that exhibit significant visual domain gaps. Existing approaches, including VLMs and world models, struggle to capture underlying physics and causality since they are not focused on core mechanisms and overfit to visual details. VLM/VLA agents reason but lack look-ahead in interactive settings, while world models imagine but imitate visual patterns rather than analyze physics and causality. We therefore propose IPR (Interactive Physical Reasoner), using world-model rollouts to score and reinforce a VLM's policy, and introduce PhysCode, a physics-centric action code aligning semantic intent with dynamics to provide a shared action space for prediction and reasoning. Pretrained on 1,000+ games, our IPR performs robustly on levels from primitive intuition to goal-driven reasoning, and even surpasses GPT-5 overall. We find that performance improves with more training games and interaction steps, and that the model also zero-shot transfers to unseen games. These results support physics-centric interaction as a path to steadily improving physical reasoning. Further demos and project details can be found at https://mybearyzhang.github.io/ipr-1.
Authors: Szymon Bobek, {\L}ukasz Ba{\l}ec, Grzegorz J. Nalepa
Abstract: Counterfactual explanations improve the actionable interpretability of machine learning models by identifying minimal changes required to achieve a desired outcome. However, existing methods often neglect dependencies among features, which can lead to unrealistic or impractical modifications. This limitation reduces the usefulness of counterfactual explanations in real-world decision-support systems. Motivated by applications in cybersecurity for email marketing, we propose DANCE (Diverse, Actionable, and Knowledge-Constrained Explanations), a method for generating counterfactuals that incorporate feature dependencies and domain constraints. DANCE models relationships between features using linear and probabilistic structures that can be learned from data or specified by experts. These dependencies are enforced during the search process to improve plausibility and feasibility. The method jointly optimizes plausibility, diversity, proximity, and sparsity within a unified objective. We evaluate DANCE on 140 datasets from OpenML and demonstrate that it achieves competitive or superior performance compared to existing approaches across multiple evaluation criteria. Additionally, we validate the method in a real-world industrial setting in collaboration with an email marketing platform, showing that it produces domain-consistent and actionable recommendations.
Authors: Edward Y. Chang
Abstract: In this paper we argue that influential critiques dismissing Large Language Models (LLMs) as a dead end for AGI misidentify the bottleneck: they confuse the ocean with the net. Pattern repositories are the necessary System-1 substrate; the missing component is a System-2 coordination layer that recruits relevant patterns, verifies their use, preserves state, and governs convergence. We separate two uses of control that are often conflated. Semantic anchoring, formalized by UCCT (Unified Contextual Control Theory), binds labels and task intent to learned pattern regions through a phase transition governed by effective support (rho_d), representational mismatch (d_r), and an adaptive anchoring budget (gamma log k). Trace-answer verification, implemented by Recursive Causal Audit (RCA), tests whether a final causal judgment is warranted by its own reasoning trace under pressure. We translate these ideas into MACI, a multi-agent coordination stack that integrates diversity and control via baiting (PID-modulated debate), filtering (Socratic and causal audit), and persistence (transactional memory). Empirical validation on causal judgment and the sycophancy-paranoia trade-off demonstrates that static prompting fails where adaptive control succeeds. By reframing common objections as testable coordination failures, we argue that the path to AGI runs through LLMs, not around them. Capability is not coordination.
Authors: Jovan Pavlovi\'c, Mikl\'os Kr\'esz, L\'aszl\'o Hajdu
Abstract: Despite initial successes and a variety of architectures, retrieval-augmented generation systems still struggle to reliably retrieve and connect the multi-step evidence required for complicated reasoning tasks. Most of the standard RAG frameworks regard all retrieved information as equally reliable, overlooking the varying credibility and interconnected nature of large textual corpora. GraphRAG approaches offer potential improvement to RAG systems by integrating knowledge graphs, which structure information into nodes and edges, capture entity relationships, and enable multi-step logical traversal. However, GraphRAG is not always an ideal solution, as it depends on high-quality graph representations of the corpus. Such representations usually rely on manually curated knowledge graphs, which are costly to construct and update, or on automated graph-construction pipelines that are often unreliable. Moreover, systems following this paradigm typically use large language models to guide graph traversal and evidence retrieval. In this paper, we propose a novel RAG framework that uses a spreading activation algorithm to retrieve information from a corpus of documents connected by an automatically constructed heterogeneous knowledge graph. This approach reduces reliance on semantic knowledge graphs, which are often incomplete due to information loss during information extraction, avoids LLM-guided graph traversal, and improves performance on multi-hop question answering. Experiments show that our method achieves better or comparable performance to several state-of-the-art RAG methods and can be integrated as a plug-and-play module with different iterative RAG pipelines. When combined with chain-of-thought iterative retrieval, it yields up to a 39% absolute improvement in answer correctness over naive RAG, while achieving these results with small open-weight language models.
Authors: Gaurab Chhetri, Subasish Das, Tausif Islam Chowdhury
Abstract: Personalized search demands the ability to model users' evolving, multi-dimensional information needs; a challenge for systems constrained by static profiles or monolithic retrieval pipelines. We present SPARK (Search Personalization via Agent-Driven Retrieval and Knowledge-sharing), a framework in which coordinated persona-based large language model (LLM) agents deliver task-specific retrieval and emergent personalization. SPARK formalizes a persona space defined by role, expertise, task context, and domain, and introduces a Persona Coordinator that dynamically interprets incoming queries to activate the most relevant specialized agents. Each agent executes an independent retrieval-augmented generation process, supported by dedicated long- and short-term memory stores and context-aware reasoning modules. Inter-agent collaboration is facilitated through structured communication protocols, including shared memory repositories, iterative debate, and relay-style knowledge transfer. Drawing on principles from cognitive architectures, multi-agent coordination theory, and information retrieval, SPARK models how emergent personalization properties arise from distributed agent behaviors governed by minimal coordination rules. The framework yields testable predictions regarding coordination efficiency, personalization quality, and cognitive load distribution, while incorporating adaptive learning mechanisms for continuous persona refinement. By integrating fine-grained agent specialization with cooperative retrieval, SPARK provides insights for next-generation search systems capable of capturing the complexity, fluidity, and context sensitivity of human information-seeking behavior.
Authors: Zoran Milosevic, Fethi Rabhi
Abstract: The rapid evolution of Large Language Models (LLM) and subsequent Agentic AI technologies requires systematic architectural guidance for building sophisticated, production-grade systems. This paper presents an approach for architecting such systems using design patterns derived from enterprise distributed systems standards, formal methods, and industry practice. We classify these patterns into three tiers: LLM Agents (task-specific automation), Agentic AI (adaptive goal-seekers), and Agentic Communities (organizational frameworks where AI agents and human participants coordinate through formal roles, protocols, and governance structures). We focus on Agentic Communities - coordination frameworks encompassing LLM Agents, Agentic AI entities, and humans - most relevant for enterprise and industrial applications. Drawing on established coordination principles from distributed systems, we ground these patterns in a formal framework that specifies collaboration agreements where AI agents and humans fill roles within governed ecosystems. This approach provides both practical guidance and formal verification capabilities, enabling expression of organizational, legal, and ethical rules through accountability mechanisms that ensure operational and verifiable governance of inter-agent communication, negotiation, and intent modeling. We validate this framework through a clinical trial matching case study. Our goal is to provide actionable guidance to practitioners while maintaining the formal rigor essential for enterprise deployment in dynamic, multi-agent ecosystems.
Authors: Zixuan Xiao, Jun Ma, Siwei Zhang
Abstract: Understanding urban environment change is essential for sustainable development. However, current approaches, particularly remote sensing change detection, often rely on rigid, single-modal analysis. To overcome these limitations, we propose MMUEChange, a multi-modal agent framework that flexibly integrates heterogeneous urban data via a modular toolkit and a core module, Modality Controller for cross- and intra-modal alignment, enabling robust analysis of complex urban change scenarios. Case studies include: a shift toward small, community-focused parks in New York, reflecting local green space efforts; the spread of concentrated water pollution across districts in Hong Kong, pointing to coordinated water management; and a notable decline in open dumpsites in Shenzhen, with contrasting links between nighttime economic activity and waste types, indicating differing urban pressures behind domestic and construction waste. Compared to the best-performing baseline, the MMUEChange agent achieves a 46.7% improvement in task success rate and effectively mitigates hallucination, demonstrating its capacity to support complex urban change analysis tasks with real-world policy implications.
Authors: Ziming Dai, Dabiao Ma, Jinle Tong, Mengyuan Han, Jian Yang, Hongtao Liu, Haojun Fei, Qing Yang
Abstract: Although the Gradient Boosted Decision Trees (GBDTs) dominate industrial tabular applications, upgrading legacy models in high-concurrency production environments still faces prohibitive retraining costs and systemic risks. To address this problem, we present NSR-Boost, a neuro-symbolic residual boosting framework designed specifically for industrial scenarios. Its core advantage lies in being ``non-intrusive''. It treats the legacy model as a frozen model and performs targeted repairs on "hard regions" where predictions fail. The framework comprises three key stages: First, finding hard regions through residuals, then generating interpretable experts by generating symbolic code structures using Large Language Model (LLM) and fine-tuning parameters using Bayesian optimization, and finally dynamically integrating experts with legacy model output through a lightweight aggregator. Experimental results demonstrate that the framework significantly outperforms state-of-the-art (SOTA) baselines across six public datasets and one private dataset. More importantly, we report the successful deployment of NSR-Boost within the core financial risk control system of Qfin Holdings, where empirical results on real-world online traffic exhibit superior performance improvements and a significant reduction in the bad rate. In conclusion, it effectively captures long-tail risks missed by traditional models and offers a safe, low-cost evolutionary paradigm for industry.
Authors: Oguzhan Gungordu, Siheng Xiong, Faramarz Fekri
Abstract: Large Language Models (LLMs) have enabled automated heuristic design (AHD) for combinatorial optimization problems (COPs), but existing frameworks' reliance on fixed evolutionary rules and static prompt templates often leads to myopic heuristic generation, redundant evaluations, and limited reasoning about how new heuristics should be derived. We propose a novel multi-agent reasoning framework, referred to as Planning through World Model for Automated Heuristic Design via Self-Evolving LLMs (PathWise), which formulates heuristic generation as a sequential decision process over an entailment graph serving as a compact, stateful memory of the search trajectory. This approach allows the system to carry forward past decisions and reuse or avoid derivation information across generations. A policy agent plans evolutionary actions, a world model agent generates heuristic rollouts conditioned on those actions, and critic agents provide routed reflections summarizing lessons from prior steps, shifting LLM-based AHD from trial-and-error evolution toward state-aware planning through reasoning. Experiments across diverse COPs show that PathWise converges faster to better heuristics, generalizes across different LLM backbones, and scales to larger problem sizes.
Authors: Siru Zhong, Yiqiu Liu, Zhiqing Cui, Zezhi Shao, Fei Wang, Qingsong Wen, Yuxuan Liang
Abstract: Deep time series models are vulnerable to noisy data ubiquitous in real-world applications. Existing robustness strategies either prune data or rely on costly prior quantification, failing to balance effectiveness and efficiency. In this paper, we introduce DropoutTS, a model-agnostic plugin that shifts the paradigm from "what" to learn to "how much" to learn. DropoutTS employs a Sample-Adaptive Dropout mechanism: leveraging spectral sparsity to efficiently quantify instance-level noise via reconstruction residuals, it dynamically calibrates model learning capacity by mapping noise to adaptive dropout rates - selectively suppressing spurious fluctuations while preserving fine-grained fidelity. Extensive experiments across diverse noise regimes and open benchmarks show DropoutTS consistently boosts superior backbones' performance, delivering advanced robustness with negligible parameter overhead and no architectural modifications. Our code is available at https://github.com/CityMind-Lab/DropoutTS.
Authors: Yuhao Zhan, Tianyu Fan, Linxuan Huang, Zirui Guo, Chao Huang
Abstract: Diagnosing failure patterns in Deep Research Agents (DRAs) remains a critical challenge. Existing benchmarks predominantly rely on end-to-end evaluation, obscuring intermediate hallucinations that accumulate throughout the research trajectory. To bridge this gap, we propose a shift from outcome-based to processaware evaluation by auditing hallucinations in the full plan-search-summarize trajectory. We introduce the PING Taxonomy, which categorizes DRA hallucinations into four complementary types: Propagation, Intent, Noiseinduced, and Grounding. We further instantiate this taxonomy into a fine-grained evaluation framework that decomposes trajectories into atomic actions, claims, and sub-queries for rigorous verification. Leveraging this framework to isolate 100 distinctively hallucinationprone tasks including adversarial scenarios, we curate DeepHalluBench. Experiments on six representative DRAs show that, on our hallucination-prone stress-test set, all evaluated systems still exhibit non-negligible reliability gaps. Furthermore, our diagnostic analysis traces these failures to systemic deficits, especially hallucination propagation and cognitive biases, providing actionable insights for future architectural optimization. Code and data are available in https://github.com/yuhao-zhan/DeepHalluBench.
Authors: Takahito Nakajima
Abstract: Background: As of 2026, Large Language Models (LLMs) demonstrate expert-level medical knowledge. However, deploying them as autonomous "Clinical Agents" remains limited. Current Electronic Medical Records (EMRs) and standards like FHIR are designed for human review, creating a "Context Mismatch": AI agents receive fragmented data and must rely on probabilistic inference (e.g., RAG) to reconstruct patient history. This approach causes hallucinations and hinders auditability. Methods: We propose MedBeads, an agent-native data infrastructure where clinical events are immutable "Beads"--nodes in a Merkle Directed Acyclic Graph (DAG)--cryptographically referencing causal predecessors. This "write-once, read-many" architecture makes tampering mathematically detectable. We implemented a prototype with a Go Core Engine, Python middleware for LLM integration, and a React-based visualization interface. Results: We successfully implemented the workflow using synthetic data. The FHIR-to-DAG conversion transformed flat resources into a causally-linked graph. Our Breadth-First Search (BFS) Context Retrieval algorithm traverses relevant subgraphs with O(V+E) complexity, enabling real-time decision support. Tamper-evidence is guaranteed by design: any modification breaks the cryptographic chain. The visualization aids clinician understanding through explicit causal links. Conclusion: MedBeads addresses the "Context Mismatch" by shifting from probabilistic search to deterministic graph traversal, and from mutable records to immutable chains, providing the substrate for "Trustworthy Medical AI." It guarantees the context the AI receives is deterministic and tamper-evident, while the LLM determines interpretation. The structured Bead format serves as a token-efficient "AI-native language." We release MedBeads as open-source software to accelerate agent-native data standards.
Authors: Gouki Minegishi, Jingyuan Feng, Hiroki Furuta, Takeshi Kojima, Yusuke Iwasawa, Yutaka Matsuo
Abstract: Analogy is a central faculty of human intelligence, enabling abstract patterns discovered in one domain to be applied to another. Despite its central role in cognition, the mechanisms by which Transformers acquire and implement analogical reasoning remain poorly understood. In this work, inspired by the notion of functors in category theory, we formalize analogical reasoning as the inference of correspondences between entities across categories. Based on this formulation, we introduce synthetic tasks that evaluate the emergence of analogical reasoning under controlled settings. We find that the emergence of analogical reasoning is highly sensitive to data characteristics, optimization choices, and model scale. Through mechanistic analysis, we show that analogical reasoning in Transformers decomposes into two key components: (1) geometric alignment of relational structure in the embedding space, and (2) the application of a functor within the Transformer. These mechanisms enable models to transfer relational structure from one category to another, realizing analogy. Finally, we quantify these effects and find that the same trends are observed in pretrained LLMs. In doing so, we move analogy from an abstract cognitive notion to a concrete, mechanistically grounded phenomenon in modern neural networks.
Authors: Yinyi Luo, Yiqiao Jin, Weichen Yu, Mengqi Zhang, Srijan Kumar, Xiaoxiao Li, Weijie Xu, Xin Chen, Jindong Wang
Abstract: While large language model (LLM) multi-agent systems achieve superior reasoning performance through iterative debate, practical deployment is limited by their high computational cost and error propagation. This paper proposes AgentArk, a novel framework to distill multi-agent dynamics into the weights of a single model, effectively transforming explicit test-time interactions into implicit model capabilities. This equips a single agent with the intelligence of multi-agent systems while remaining computationally efficient. Specifically, we investigate three hierarchical distillation strategies across various models, tasks, scaling, and scenarios: reasoning-enhanced fine-tuning; trajectory-based augmentation; and process-aware distillation. By shifting the burden of computation from inference to training, the distilled models preserve the efficiency of one agent while exhibiting strong reasoning and self-correction performance of multiple agents. They further demonstrate enhanced robustness and generalization across diverse reasoning tasks. We hope this work can shed light on future research on efficient and robust multi-agent development. Our code is at https://github.com/AIFrontierLab/AgentArk.
Authors: Jorge Fandinno, Pedro Cabalar, Philipp Wanko, Torsten Schaub
Abstract: Constraint Answer Set Programming (CASP) is a hybrid paradigm that enriches Answer Set Programming (ASP) with numerical constraint processing, a crucial requirement for many real-world applications. However, the specification of constraints in most CASP solvers aligns more closely with the expressiveness and semantics of the numerical back-end than the ASP paradigm. In the latter, numerical attributes are represented as predicates, which allows declaring default values, leaving the attribute undefined, making non-deterministic assignments with choice rules, or using aggregated values. In CASP, most (if not all) of these features are lost once we switch to a constraint-based representation of those same attributes. In this paper, we present the flingo language (and tool) that incorporates the aforementioned expressiveness within numerical constraints, and we illustrate its use with several examples. Based on previous work that established its semantic foundations, we also present a translation from the newly introduced flingo syntax to regular CASP programs following the clingcon input format.
Authors: Zhaoyang Wang, Canwen Xu, Boyi Liu, Yite Wang, Siwei Han, Zhewei Yao, Huaxiu Yao, Yuxiong He
Abstract: Recent advances in large language model (LLM) have empowered autonomous agents to perform multi-turn interactions with tools and environments. However, scaling such agent training is limited by the lack of diverse and reliable environments. In this paper, we propose Agent World Model (AWM), a fully synthetic environment generation pipeline. Using this pipeline, we scale to 1,000 environments covering everyday scenarios, in which agents can interact with rich toolsets and obtain high-quality observations. Notably, these environments are code-driven and backed by databases, providing more reliable and consistent state transitions than environments simulated by LLMs. Moreover, they enable more efficient agent interaction compared with collecting trajectories from realistic environments. To demonstrate the effectiveness of this resource, we perform large-scale reinforcement learning for multi-turn tool-use agents. Thanks to the fully executable environments and accessible database states, we can also design reliable reward functions. Experiments on three benchmarks show that training exclusively in synthetic environments, rather than benchmark-specific ones, yields strong out-of-distribution generalization. The code is available at https://github.com/Snowflake-Labs/agent-world-model.
Authors: Keane Ong, Sabri Boughorbel, Luwei Xiao, Chanakya Ekbote, Wei Dai, Ao Qu, Jingyao Wu, Rui Mao, Ehsan Hoque, Erik Cambria, Gianmarco Mengaldo, Paul Pu Liang
Abstract: Socially intelligent AI systems must entail reasoning across diverse human behavioral tasks, and generalization to new contexts. However, AI has yet to achieve this level of social intelligence. Existing models remain fundamentally constrained by the imbalanced learning dynamics induced by training on behavioral data. Namely, behavioral data is inherently heterogeneous, comprising diverse modalities and prediction targets that often produce uneven training signals across samples. To address this, we develop Omnisapiens-7B 2.0, a foundation model for social behavior processing that explicitly addresses learning from heterogeneous behavioral data. This is enabled through Heterogeneity-Aware Relative Policy Optimization, a novel reasoning RL method that explicitly rebalances learning signals across samples. The core insight is to approximate contribution signals to the policy update, using them to inform geometrically centered and intertially smoothed advantage modulation. Results demonstrate that Omnisapiens-7B 2.0 achieves the best and most consistent performance across 10 diverse behavioral tasks, while also attaining the best performance on all five held-out zero-shot generalization benchmarks, with gains of up to +12.02% and +9.37% respectively. Furthermore, Omnisapiens-7B 2.0 demonstrates more consistent and interpretable reasoning traces, supporting reliable real-world behavioral applications. Our model and codes can be found at https://github.com/MIT-MI/human_behavior_atlas.
Authors: Ariel Larey, Elay Dahan, Amit Bleiweiss, Raizy Kellerman, Guy Leib, Omri Nayshool, Dan Ofer, Tal Zinger, Dan Dominissini, Gideon Rechavi, Nicole Bussola, Simon Lee, Shane O'Connell, Dung Hoang, Marissa Wirth, Alexander W. Charney, Nati Daniel, Yoli Shavit
Abstract: Genomic Foundation Models (GFMs) typically rely on Masked Language Modeling (MLM) or Next-Token Prediction (NTP) to learn the "Laws of Nature". While effective at capturing local syntax, these generative paradigms prioritize token-level reconstruction over high-level functional context. We introduce JEPA-DNA, a model-agnostic continual training framework that integrates a Joint-Embedding Predictive Architecture (JEPA) with traditional generative objectives. By supervising global sequence embeddings in a latent space, JEPA-DNA forces models to predict the functional representations of masked genomic segments, shifting the learning signal from token recovery to semantic alignment. We evaluate JEPA-DNA on 17 diverse genomic benchmark tasks, demonstrating consistent gains in linear probing and zero-shot performance regardless of the underlying GFM architecture or generative objective. Our framework establishes a new state-of-the-art for GFMs, surpassing the best existing models by bridging generative precision with latent semantic grounding. Through extensive ablation studies, we further characterize the synergistic interplay between generative and latent objectives. Our code is publicly available at https://github.com/NVIDIA-Digital-Bio/JEPA-DNA.
Authors: Zeyu Zhang, Ryan Chen, Bradly C. Stadie
Abstract: Backtesting LLMs on resolved events assumes models reason only from pre-cutoff knowledge, yet pretrained models inevitably leak post-cutoff knowledge. We introduce a claim-level evaluation framework that decomposes prediction rationales into atomic claims and applies Shapley values to quantify each claim's decision impact, yielding \textbf{Shapley-DCLR} (\textbf{Shapley}-weighted \textbf{D}ecision-\textbf{C}ritical \textbf{L}eakage \textbf{R}ate) -- an interpretable metric measuring what fraction of decision-driving reasoning is contaminated. We further propose \textbf{TimeSPEC} (\textbf{Time}-\textbf{S}upervised \textbf{P}rediction with \textbf{E}xtracted \textbf{C}laims), an inference-time architecture that interleaves temporally-filtered retrieval with claim-level supervision, producing predictions grounded entirely in pre-cutoff evidence. Across three LLMs, the ablation experiments confirm retrieval and supervision are jointly necessary; and a three-task probe further illstrates that the performance cost of temporal enforcement scales with each task's reliance on post-cutoff information.
Authors: Longfei Yun, Yihan Wu, Haoran Liu, Xiaoxuan Liu, Ziyun Xu, Yi Wang, Yang Xia, Pengfei Wang, Mingze Gao, Yunxiang Wang, Changfan Chen, Wenjie Fu, Hong Yan, Junfeng Pan
Abstract: Modern large-scale ranking systems operate within a sophisticated landscape of competing objectives, operational constraints, and evolving product requirements. Progress in this domain is increasingly bottlenecked by the engineering context constraint: the arduous process of translating ambiguous product intent into reasonable, executable, verifiable hypotheses, rather than by modeling techniques alone. We present GEARS (Generative Engine for Agentic Ranking Systems), a framework that reframes ranking optimization as an autonomous discovery process within a programmable experimentation environment. Rather than treating optimization as static model selection, GEARS leverages Specialized Agent Skills to encapsulate ranking expert knowledge into reusable reasoning capabilities, enabling operators to steer systems via high-level intent vibe personalization. Furthermore, to ensure production reliability, the framework incorporates validation hooks to enforce statistical robustness and filter out brittle policies that overfit short-term signals. Experimental validation across diverse product surfaces demonstrates that GEARS consistently identifies superior, near-Pareto-efficient policies by synergizing algorithmic signals with deep ranking context while maintaining rigorous deployment stability.
Authors: Serafim Batzoglou
Abstract: We introduce INDUCTION, a benchmark for finite structure concept synthesis in first order logic. Given small finite relational worlds with extensionally labeled target predicates, models must output a single first order logical formula that explains the target uniformly across worlds, with correctness verified via exact model checking. The benchmark includes three regimes, FullObs, CI (contrastive), and EC (existential completion), nd penalizes formula bloat. We find sharp difficulty gradients, persistent hard structural families, and observe that low bloat formulas generalize far better on held out worlds. Elite recent models show qualitatively different behaviors across tasks and performance metrics, hinting to their different strategies of concept generalization.
Authors: Yujie Zhao, Boqin Yuan, Junbo Huang, Haocheng Yuan, Zhongming Yu, Haozhou Xu, Lanxiang Hu, Abhilash Shankarampeta, Zimeng Huang, Wentao Ni, Yuandong Tian, Jishen Zhao
Abstract: Large Language Models (LLMs) are deployed as autonomous agents in increasingly complex applications, where enabling long-horizon memory is critical for achieving strong performance. However, a significant gap exists between applications and evaluation standards for agent memory: existing benchmarks primarily focus on dialogue-centric settings. In reality, agent memory consists of a continuous stream of agent-environment interactions that are primarily composed of machine-generated representations. To bridge this gap, we introduce AMA-Bench (Agent Memory with Any Length), a benchmark designed to evaluate long-horizon memory for LLMs in real agentic applications. It features two key components: (1) a set of real-world agentic trajectories across representative agentic applications, paired with expert-curated QA, and (2) a set of synthetic agentic trajectories of arbitrary horizons paired with rule-based QA. Our comprehensive study shows that existing memory systems underperform on AMA-Bench primarily because they lack causality and objective information, and are constrained by the lossy nature of similarity-based retrieval employed by many memory systems. To address these limitations, we propose AMA-Agent, an effective memory system featuring a causality graph and tool-augmented retrieval. Our results demonstrate that AMA-Agent achieves 57.22% average accuracy on AMA-Bench, surpassing the strongest baselines by 11.16%. Resources are available at our project website: https://ama-bench.github.io/
Authors: Jinyue Li, Yuci Liang, Qiankun Li, Xinheng Lyu, Jiayu Qian, Huabao Chen, Kun Wang, Zhigang Zeng, Anil Anthony Bharath, Yang Liu
Abstract: Computational pathology demands both visual pattern recognition and dynamic integration of structured domain knowledge, including taxonomy, grading criteria, and clinical evidence. In practice, diagnostic reasoning requires linking morphological evidence with formal diagnostic and grading criteria. Although multimodal large language models (MLLMs) demonstrate strong vision language reasoning capabilities, they lack explicit mechanisms for structured knowledge integration and interpretable memory control. As a result, existing models struggle to consistently incorporate pathology-specific diagnostic standards during reasoning. Inspired by the hierarchical memory process of human pathologists, we propose PathMem, a memory-centric multimodal framework for pathology MLLMs. PathMem organizes structured pathology knowledge as a long-term memory (LTM) and introduces a Memory Transformer that models the dynamic transition from LTM to working memory (WM) through multimodal memory activation and context-aware knowledge grounding, enabling context-aware memory refinement for downstream reasoning. PathMem achieves SOTA performance across benchmarks, improving WSI-Bench report generation (12.8% WSI-Precision, 10.1% WSI-Relevance) and open-ended diagnosis by 9.7% and 8.9% over prior WSI-based models.
Authors: Matt Gorbett, Suman Jana
Abstract: Language models increasingly appear to learn similar representations, despite differences in training objectives, architectures, and data modalities. This emerging compatibility between independently trained models introduces new opportunities for cross-model alignment to downstream objectives. Moreover, this capability unlocks new potential application domains, such as settings where security, privacy, or competitive constraints prohibit direct data or model sharing. In this work, we investigate the extent to which representational convergence enables practical linear alignment between large language models. Specifically, we learn affine transformations between the final hidden states of independent models and empirically evaluate these mappings across text generation, embedding classification, and out-of-distribution detection. We find that performance is largely preserved across model pairs, and show for the first time that linear alignment sometimes enables text generation across independently trained models. We further highlight a potential application of linear alignment for privacy-preserving cross-silo inference. The framework learns an affine transformation over a shared public dataset and uses homomorphic encryption to protect client queries. By encrypting only the linear classification operation, the method achieves sub-second inference latency.
Authors: Songjun Tu, Chengdong Xu, Qichao Zhang, Yaocheng Zhang, Xiangyuan Lan, Linjing Li, Dong Li, Dongbin Zhao
Abstract: Agentic RL can benefit substantially from reusable experience, yet existing skill-based methods mainly extract trajectory-level guidance and often lack principled mechanisms for maintaining an evolving skill memory. We propose D2Skill, a dynamic dual-granularity skill bank for agentic RL that organizes reusable experience into task skills for high-level guidance and step skills for fine-grained decision support and error correction. D2Skill jointly trains the policy and skill bank through paired baseline and skill-injected rollouts under the same policy, using their performance gap to derive hindsight utility signals for both skill updating and policy optimization. Built entirely from training-time experience, the skill bank is continuously expanded through reflection and maintained with utility-aware retrieval and pruning. Experiments on ALFWorld, WebShop, and Search-Augmented QA tasks show that D2Skill substantially improves performance over skill-free baselines across models of different scales. Further ablations and analyses show that both dual-granularity skill modeling and dynamic skill maintenance are critical to these gains, while the learned skills exhibit higher utility, transfer across evaluation settings, and introduce only modest training overhead.
Authors: Erhan Zhang, Yiqun Chen, Zechun Niu, Wei Yang, Xiaochi Wei, Yan Gao, Yi Wu, Yao Hu, Jiaxin Mao
Abstract: Agentic search enables language models to solve knowledge-intensive tasks by adaptively acquiring external evidence over multiple steps. Reinforcement learning with verifiable rewards (RLVR) has emerged as a widely adopted training paradigm for search agents, yet outcome-only rewards are sparse and provide limited credit assignment for intermediate search actions. Existing process-reward methods therefore seek to densify supervision through proxy signals, external evaluators, or likelihood-based information gain. However, proxy rewards can deviate from the final outcome objective, while fixed evaluators can become stale as the search policy evolves, leading to unreliable process supervision. To address these challenges, we propose OASES, an Outcome-Aligned Search-Evaluation Supervision framework for agentic search. OASES derives outcome-aligned process rewards by evaluating how well each intermediate search state supports answering the original question. It further co-trains the search policy and the state evaluator on policy, allowing the evaluator to adapt to evolving search behavior and provide more reliable process rewards. Experiments on five multi-hop QA benchmarks show that OASES consistently outperforms strong RL baselines, with further analyses confirming the benefits of outcome-aligned process rewards and search-evaluation co-training.
Authors: Sihang Jiang, Lipeng Ma, Zhonghua Hong, Keyi Wang, Zhiyu Lu, Tengfei Wang, Shisong Chen, Jinghao Zhang, Tianjun Pan, Weijia Li, Jiaqing Liang, Yanghua Xiao
Abstract: Current LLM-based agents demonstrate strong performance in episodic task execution but remain constrained by static toolsets and episodic amnesia, failing to accumulate experience across task boundaries. This paper formalizes the Self-Evolving Agent (SEA) from the perspective of digital embodiment and continuous cross-task evolution, introduces the Evolutionary Flywheel as its minimal sufficient architecture, and presents SEA-Eval -- the first benchmark designed specifically for evaluating SEAs. Grounded in Flywheel theory, SEA-Eval establishes SR and T as primary metrics and, through sequential task stream design, is designed to quantify evolutionary gain, evolutionary stability, and implicit alignment convergence. Empirical evaluation reveals that, under comparable success rates, token consumption differs by up to 31.2 times between frameworks on individual tasks, with divergent evolutionary trajectories emerging under sequential analysis -- demonstrating that success rate alone creates a capability illusion and that the sequential convergence of $T$ is the key criterion for distinguishing genuine evolution from pseudo-evolution.
Authors: Daniel J. Tan, Jayne Hui Zhen Chan, Kai Wen Hwang, Arturo Yong Yao Neo, Kay Choong See, Mengling Feng
Abstract: Designing reward functions for reinforcement learning (RL) in healthcare remains challenging because clinically meaningful outcomes are sparse, delayed, and difficult to explicitly specify. Although structured clinical data capture physiologic states, they often fail to reflect broader aspects of patient trajectories such as treatment response, recovery dynamics, and intervention burden. Clinical narratives, by contrast, encode longitudinal clinician assessments of disease progression, treatment effectiveness, and recovery, providing a potential source of trajectory-level supervision beyond predefined outcome metrics. We propose Clinical Narrative-informed Preference Rewards (CN-PR), a framework that learns reward functions directly from discharge summaries by treating clinical narratives as scalable supervision for trajectory-level preferences. Using a large language model, we derive trajectory quality scores and construct pairwise preferences between patient trajectories to learn rewards through preference-based optimization. To account for variability in narrative informativeness, we incorporate a task relevance signal that weights supervision according to its relevance to the downstream decision-making task. We evaluate CN-PR in dynamic sepsis treatment using offline RL. The learned reward demonstrated strong monotonic alignment with trajectory quality scores and produced policies associated with improved recovery-related outcomes, including increased organ support-free days and faster shock resolution, while maintaining mortality performance comparable to outcome-based reward baselines. These findings were preserved under external validation. Our results suggest that clinical narratives provide a scalable and expressive source of supervision for reward learning in dynamic treatment regimes.
Authors: Yijuan Liang, Xinghao Chen, Yifan Ge, Ziyi Wu, Hao Wu, Changyu Zeng, Wei Xing, Xiaoyu Shen
Abstract: Tool-use capability is a fundamental component of LLM agents, enabling them to interact with external systems through structured function calls. However, existing research exhibits inconsistent interaction representations, largely overlooks the structural distribution of tool-use trajectories, and relies on incompatible evaluation benchmarks. We present UniToolCall, a unified framework for tool learning that standardizes the entire pipeline from toolset construction and dataset generation to evaluation. The framework curates a large tool pool of 22k+ tools and constructs a hybrid training corpus of 390k+ instances by combining 10 standardized public datasets with structurally controlled synthetic trajectories. It explicitly models diverse interaction patterns, including single-hop vs. multi-hop and single-turn vs. multi-turn, while capturing both serial and parallel execution structures. To support coherent multi-turn reasoning, we further introduce an Anchor Linkage mechanism that enforces cross-turn dependencies. Furthermore, we convert 7 public benchmarks into a unified Query--Action--Observation--Answer (QAOA) representation with fine-grained evaluation at the function-call, turn, and conversation levels. Experiments show that fine-tuning Qwen3-8B on our dataset substantially improves tool-use performance. Under the distractor-heavy Hybrid-20 setting, achieves 93.0% single-turn Strict Precision, outperforming commercial models including GPT, Gemini, and Claude.
Authors: Shasha Yu, Fiona Carroll, Barry L. Bentley
Abstract: Large language models (LLMs) are increasingly deployed as tool-augmented agents capable of executing system-level operations. While existing benchmarks primarily assess textual alignment or task success, less attention has been paid to the structural relationship between linguistic signaling and executable behavior under varying autonomy scaffolds. This study introduces an execution-layer be-havioral measurement approach based on a two-dimensional A-R space defined by Action Rate (A) and Refusal Signal (R), with Divergence (D) capturing coor-dination between the two. Models are evaluated across four normative regimes (Control, Gray, Dilemma, and Malicious) and three autonomy configurations (di-rect execution, planning, and reflection). Rather than assigning aggregate safety scores, the method characterizes how execution and refusal redistribute across contextual framing and scaffold depth. Empirical results show that execution and refusal constitute separable behavioral dimensions whose joint distribution varies systematically across regimes and autonomy levels. Reflection-based scaffolding often shifts configurations toward higher refusal in risk-laden contexts, but redis-tribution patterns differ structurally across models. The A-R representation makes cross-sectional behavioral profiles, scaffold-induced transitions, and coordination variability directly observable. By foregrounding execution-layer characterization over scalar ranking, this work provides a deployment-oriented lens for analyzing and selecting tool-enabled LLM agents in organizational settings where execution privileges and risk tolerance vary.
Authors: Boqin Yuan, Yue Su, Renchu Song, Sen Yang, Jing Qin
Abstract: Skill-distillation pipelines learn reusable rules from LLM agent trajectories, but they lack a key signal: how much each step costs. Without per-step cost, a pipeline cannot distinguish adding a missing step to fix a bug from removing an expensive step that never affected the outcome. We use the cost-attribution gap to ask whether the rule types inside a distilled skill transfer the same way to new tasks. ClawTrace records cost-attributed agent traces and compiles each session into a TraceCard; CostCraft reads TraceCards and writes three kinds of skill patches: preserve, prune, and repair. We find a pattern aggregate metrics hide. On 30 held-out SpreadsheetBench tasks across two seeds, removing prune patches roughly tripled the quality-regression count without lowering median cost. Across the full 84-task SkillsBench transfer, CostCraft saves no aggregate cost. All three quality regressions trace to the preserve lane, and both quality wins trace to the prune lane: prune patches act as quality guardrails while preserve patches drive regressions. We argue that reusable agent skills should be evaluated at the rule-type level, not as monolithic instruction packages. To support this, we release ClawTrace, the TraceCard schema, and the full set of typed skills.
Authors: Yifang Qin, Yu Shi, Junfu Tan, Chang Liu, Ming Zhang, Ziheng Lu
Abstract: Predicting stable and metastable structures is central to molecular and materials discovery, but remains limited by the cost of searching high-dimensional energy landscapes. Deep generative models offer efficient structure sampling, yet their outputs remain shaped by training data and can underexplore minima that are rare but physically relevant. We introduce generative structure search (GSS), a unified framework that formulates diffusion-based generation and random structure search (RSS) as limiting regimes of a common sampling process driven by learned score fields and physical forces. Coupling these drivers lets GSS use data priors to accelerate sampling while retaining energy-guided exploration of local minima. Across molecular and crystalline systems, GSS recovers diverse metastable structures with more than tenfold lower sampling cost than RSS for broad coverage and remains effective for compositions outside the training distribution. The results establish a physically grounded generative search strategy for discovering structures beyond the reach of data-driven sampling alone.
Authors: Hengyu Liu, Tianyi Li, Zhihong Cui, Yushuai Li, Zhangkai Wu, Torben Bach Pedersen, Kristian Torp, Christian S. Jensen
Abstract: This position paper argues that reliable AI requires infrastructure for human validation of implicit knowledge. AI learns from both explicit knowledge (papers, documentation, structured databases) and implicit knowledge (reasoning patterns, debugging processes, intermediate steps). Implicit knowledge remains unexternalized because documentation cost exceeds perceived value -- yet AI learns from it indiscriminately, acquiring both beneficial patterns and harmful biases. Current reliability methods can only verify explicit knowledge against sources, creating a fundamental gap: the most valuable AI capabilities (reasoning, judgment, intuition) are precisely those we cannot verify. We propose Knowledge Objects (KOs) -- structured artifacts that externalize implicit knowledge into forms humans can inspect, verify, and endorse. KOs transform verification economics: what was previously too costly to verify becomes feasible, enabling accumulated human validation to improve reliability over time.
Authors: Seunghan Lee, Jun Seo, Jaehoon Lee, Sungdong Yoo, Minjae Kim, Tae Yoon Lim, Dongwan Kang, Hwanil Choi, Soonyoung Lee, Wonbin Ahn
Abstract: Time series (TS) reasoning models (TSRMs) have shown promising capabilities in general domains, yet they consistently fail on financial domain, which exhibit unique characteristics. We propose a general 2x2 capability taxonomy for TSRMs by crossing 1) single-entity vs. multi-entity analysis with 2) assessment of the current state vs. prediction of future behavior. We instantiate this taxonomy in the financial domain -- where the distinction between deterministic assessment and stochastic prediction is particularly critical -- as ten financial reasoning tasks, forming the FinTSR-Bench benchmark based on S&P stocks. To this end, we propose FinSTaR (Financial Time Series Thinking and Reasoning), trained on FinTSR-Bench with distinct chain-of-thought (CoT) strategies tailored to each category. For assessment, which is deterministic (i.e., computable from observable data), we employ Compute-in-CoT, a programmatic CoT that enables models to derive answers directly from raw prices. For prediction, which is inherently stochastic (i.e., subject to unobservable factors), we adopt Scenario-Aware CoT, which generates diverse scenarios before making a judgment, mirroring how financial analysts reason under uncertainty. The proposed method achieves 78.9% average accuracy on FinTSR-Bench, substantially outperforming LLM and TSRM baselines. Furthermore, we show that the four capability categories are complementary and mutually reinforcing through joint training, and that Scenario-Aware CoT consistently improves prediction accuracy over standard CoT. Code is publicly available at: https://github.com/seunghan96/FinSTaR.
Authors: Bronislav Sidik, Lior Rokach
Abstract: Long-running autonomous AI agents suffer from a well-documented memory coherence problem: tool-execution success rates degrade 14 percentage points over 72-hour operation windows due to four compounding failure modes in existing flat-file memory systems. We present MEMTIER, a tripartite memory architecture for the OpenClaw agent runtime that introduces a structured episodic JSONL store, a five-signal weighted retrieval engine, an attention-attributed cognitive weight update loop, an asynchronous consolidation daemon promoting episodic facts to a semantic tier, and a PPO-based policy framework for adapting retrieval weights (infrastructure validated; performance gains pending camera-ready). On the full 500-question LongMemEval-S benchmark (Wu et al., 2025), MEMTIER achieves Acc=0.382, F1=0.412 with Qwen2.5-7B on a consumer 6GB GPU - a +33 percentage point improvement over the full-context baseline (0.050 -> 0.382, i.e., 5% -> 38%). With DeepSeek-V4-Flash fact pre-population, single-session recall reaches 0.686-0.714, exceeding the paper's RAG BM25 GPT-4o baseline (0.560) on those categories. Temporal reasoning rises to 0.323 and multi-session synthesis to 0.173, demonstrating that structured semantic pre-population qualitatively changes what lightweight retrieval can achieve. All phases run locally on a consumer laptop with a 6GB GPU.
Authors: Jiale Chang, Yuxiang Ren
Abstract: Long-term personalized memory for LLM agents is challenging on resource-limited edge devices due to high storage costs and multimodal complexity. To address this, we propose ScrapMem, a framework that integrates multimodal data into "Scrapbook Page." ScrapMem introduces Optical Forgetting, an optical compression mechanism that progressively reduces the resolution of older memories, lowering storage cost while suppressing low-value details. To maintain semantic consistency, we construct an Episodic Memory Graph (EM-Graph) that organizes key events into a causal-temporal structure. Extensive experiments on the multimodal ATM-Bench showcase that ScrapMem provides three main benefits: (1) strong performance, achieving a new state-of-the-art with a 51.0% Joint@10 score; (2) high storage efficiency, reducing memory usage by up to 93% via optical forgetting; and (3) improved recall, increasing Recall@10 to 70.3% through structured aggregation. ScrapMem offers an effective and storage-efficient solution for on-device long-term memory in multimodal LLM agents.
Authors: Yidong He, Yutao Lai, Pengxu Yang, Jiarui Gan, Jiexin Wang, Yi Cai, Mengchen Zhao
Abstract: While Large Language Models (LLMs) excel in certain reasoning tasks, they struggle in multi-agent games where the final outcome depends on the joint strategies of all agents. In multi-agent games, the non-stationarity of other agents brings significant challenges on the evaluation of the reasoning process and the credit assignment over multiple reasoning steps. Existing single-agent reinforcement learning (RL) approaches and their multi-agent extensions fail to address these challenges as they do not incorporate other agents in the reasoning process. In this work, we propose Strat-Reasoner, a novel RL-based framework that improves LLMs' strategic reasoning ability in multi-agent games. We introduce a novel recursive reasoning paradigm where an agent's reasoning also integrates other agents' reasoning processes. To provide effective reward signals for the intermediate reasoning sequences, we employ a centralized Chain-of-Thought (CoT) comparison module to evaluate the reasoning quality. Finally, we compute an accurate hybrid advantage and develop a group-relative RL approach to optimize the LLM policy. Experimental results show that Strat-Reasoner substantially improves strategic abilities of underlying LLMs, achieving 22.1\% average performance improvements across various multi-agent games. Code is publicly available at https://github.com/ydhe1012/Strat-Reasoner.
Authors: Simon Yu, Derek Chong, Ananjan Nandi, Dilara Soylu, Jiuding Sun, Christopher D Manning, Weiyan Shi
Abstract: As LLM agent systems take on more complex tasks, they increasingly rely on meta-agents: higher-order agents that operate on other agents, much as managers supervise employees. Whatever a meta-agent does: coordinating agents, halting risky actions before execution, or repairing failed runs, requires manipulation of agentic execution at runtime. Existing agentic substrates make this hard: they give meta-agents only plain transcripts and environment snapshots, requiring it to build it's own tooling to reconstruct and orchestrate execution state. Therefore, we introduce Shepherd, a Python substrate grounded in functional programming principles, where an agent's execution is itself a first-class object that a meta-agent can inspect and transform. Every model call, tool call, and environment change becomes a structured event in a Git-like execution trace, where any past state can be forked 5x faster than docker commit and replayed. Three example use cases show Shepherd's versatility: (1) a supervisor agent prevents conflicts among parallel coding agents, lifting CooperBench performance from 28.8% to 54.7%; (2) a counterfactual optimizer repairs agent workflows by proposing edits and replaying runs from the point of changed behavior, outperforming MetaHarness on TerminalBench-2 with 58% lower wall-clock; (3) a meta-agent picks fork points during rollouts to improve credit assignment in long-horizon agentic RL, doubling GRPO's gains on TerminalBench-2. We open-source Shepherd to empower future meta-agents with principled and efficient operations over agentic execution.
Authors: Siqi Zhu, Xuyan Ye, Hongyu Lu, Weiye Shi, Ge Liu
Abstract: On-policy distillation (OPD) and on-policy self-distillation (OPSD) have emerged as promising post-training methods for large language models, offering dense token-level supervision on trajectories sampled from the model's own policy. However, existing results on their effectiveness remain mixed: while OP(S)D has shown promise in system prompt and knowledge internalization, recent studies also report instability and degradation. In this work, we present a comprehensive empirical study of when OPD and OPSD work, when they fail, and why. We find that OPD on mathematical reasoning is highly sensitive to teacher choice and loss formulation, whereas OPSD fails in our tested settings due to test-time absence of instance-specific privileged information (PI). In contrast, OPSD is effective when PI represents a shared latent rule, such as a system prompt or alignment preference. We identify three failure mechanisms: (1) distribution mismatch between teacher and student caused by conditioning on student-generated prefixes, (2) optimization instability from biased TopK reverse-KL gradients, and (3) an OPSD-specific limitation where the student learns a PI-free policy that aggregates PI-conditioned teachers, which is insufficient when PI is instance-specific. We further show that stop-gradient TopK objectives, RLVR-adapted teachers, and SFT-stabilized students mitigate these failures.
Authors: Ming Du, Xiangyu Yin, Yanqi Luo, Dishant Beniwal, Songyuan Tang, Hemant Sharma, Mathew J. Cherukara
Abstract: Scientific data processing often requires task-specific algorithms or AI models, creating a barrier for domain scientists who need to analyze their data but may not have extensive computing or image-processing expertise. This barrier is especially pronounced when data are noisy, have a high dynamic range, are sparsely labeled, or are only loosely specified. We introduce CVEvolve, an autonomous agentic harness with a zero-code interface for scientific data-processing algorithm discovery. CVEvolve combines a multi-round search strategy with tools for code execution, evaluation implementation, history management, holdout testing, and optional inspection of scientific data and visual outputs. The search alternates between discovery and improvement actions, and uses lineage-aware stochastic candidate sampling to balance exploration and exploitation. We demonstrate CVEvolve on x-ray fluorescence microscopy image registration, Bragg peak detection, and high-energy diffraction microscopy image segmentation. Across these tasks, CVEvolve discovers algorithms that improve over baseline methods, while holdout test tracking helps identify candidates that generalize better than later over-optimized alternatives. These results show that zero-code, autonomous LLM-powered algorithm development can help domain scientists turn unstructured scientific image data into practical algorithms and downstream scientific discoveries.
Authors: Jonas Reiter, Jakob Elias Gebler, Hector Geffner
Abstract: Classical planners can effectively solve very large deterministic MDPs represented in STRIPS or PDDL where states are sets of atoms over objects and relations, and lifted action schemas add or delete these atoms. This compact representation yields strong search heuristics and provides an ideal setting for structural generalization, since lifted relations and action schemas give rise to infinitely many domain instances. A central challenge is to learn these relations and action schemas from data, and recent approaches have addressed this problem using different types of observations. In this work, we develop a novel neural network architecture for learning action schemas from traces where states are fully observed but action arguments are unobserved. The problem is a simplification but an important step towards learning planning domains from sequences of images and action labels, and we aim to solve this simplification in a nearly perfect manner. The challenge lies in learning the action schemas while simultaneously identifying the action arguments from observed state changes. Our approach yields a robust differentiable component that can then be integrated into larger neuro-symbolic models. We evaluate the architecture on various planning domains, where the learned lifted action schemas must recover the ground-truth structure. Additionally, we report experiments on robustness to observation noise and on a variation related to slot-based dynamics models.
Authors: Suli Wang, Yiqun Duan, Yu Deng, Rundong Zhao, Dai Shi, Xinliang Zhou
Abstract: Existing agent memory remains predominantly reactive and retrieval-based, lacking the capacity to autonomously organize experience into persistent cognitive structure. Toward genuinely autonomous agents, we introduce CogniFold, a brain-inspired "always-on" agent memory designed for the next generation of proactive assistants. CogniFold continuously folds fragmented event streams into self-emerging cognitive structures, bootstrapping progressively higher-level cognition from incoming events and accumulated knowledge. We ground this by extending Complementary Learning Systems (CLS) theory from two layers (hippocampus, neocortex) to three, adding a prefrontal intent layer. Emulating the prefrontal cortex as the locus of intentional control and decision-making, CogniFold achieves this through graph-topology self-organization: cognitive structures proactively assemble under the stream, merge when semantically similar, decay when stale, relink through associative recall, and surface intents when concept-cluster density crosses a threshold. We evaluate structural formation using CogEval-Bench, demonstrating that CogniFold uniquely produces memory structures that match cognitive expectations and concept emergence. Furthermore, across 7 broad-coverage benchmarks spanning five cognitive domains, we validate that CogniFold simultaneously performs robustly on conventional memory benchmarks.
Authors: Jia Huang, Joey Tianyi Zhou
Abstract: Existing frameworks for LLM-based agent architectures describe systems from a single perspective: industry guides (Anthropic, Google, LangChain) focus on execution topology -- how data flows -- while cognitive science surveys focus on cognitive function -- what the agent does. Neither axis alone disambiguates architecturally distinct systems: the same Orchestrator-Workers topology can implement Plan-and-Execute, Hierarchical Delegation, or Adversarial Verification -- three patterns with fundamentally different failure modes and design trade-offs. We propose a two-dimensional classification that combines (1) a Cognitive Function axis with seven categories (Perception, Memory, Reasoning, Action, Reflection, Collaboration, Governance) and (2) an Execution Topology axis with six structural archetypes (Chain, Route, Parallel, Orchestrate, Loop, Hierarchy). The resulting 7x6 matrix identifies 28 named patterns, 15 with original names. We demonstrate orthogonality through systematic cross-axis analysis, define eight representative patterns in detail, and validate descriptive coverage across four real-world domains (financial lending, legal due diligence, network operations, healthcare triage). Cross-domain analysis yields five empirical laws of pattern selection governing the relationship between environmental constraints (time pressure, action authority, failure cost asymmetry, volume) and architectural choices. The framework provides a principled, framework-neutral, and model-agnostic vocabulary for AI agent architecture design.
Authors: Kean Shi, Zihang Li, Tianyi Ma, Zengji Tu, Jialong Wu, Xinbo Xu, Qingyao Yang, Ruoyu Wu, Weichu Xie, Ming Wu, Jason Zeng, Michael Heinrich, Elvis Zhang, Liang Chen, Kuan Li, Baobao Chang
Abstract: Computer-Using Agents (CUAs) are rapidly extending large language models (LLMs) beyond text-based reasoning toward action execution in more complex environments, such as web browsers and graphical user interfaces (GUIs). However, existing web and GUI agent benchmarks often rely on simplified settings, isolated tasks, or short-horizon interactions, making it difficult to assess capabilities of agents in realistic professional workflows. Software-as-a-Service (SaaS) environments are a natural choice for CUA evaluation, as they host a large share of modern digital work and naturally involve dynamic system states, cross-application coordination, domain-specific knowledge, and long-horizon dependencies. To this end, we introduce SaaS-Bench, a benchmark built on 23 deployable SaaS systems across six professional domains, containing 106 tasks grounded in realistic work scenarios. These tasks require long-horizon execution, cover both text-only and multimodal settings, and are evaluated with weighted verification checkpoints that measure strict task completion and partial progress. Experiments show that representative LLM-based agents struggle on SaaS-Bench, with even the strongest model completing fewer than 4% of tasks end-to-end, exposing limitations in planning, state tracking, cross-application context maintenance, and error recovery. Code are available at https://github.com/UniPat-AI/SaaS-Bench for reproduction.
Authors: Shuqi Zhu, Yi Zhong, Ziyi Ye, Bangde Du, Yujia Zhou, Qingyao Ai, Yiqun Liu
Abstract: While AI-generated hallucinations pose considerable risks, the underlying cognitive mechanisms by which humans can successfully recognize or be misled by these hallucinations remain unclear. To address this problem, this paper explores humans' neural dynamics to characterize how the brain processes hallucinated content. We record EEG signals from 27 participants while they are performing a verification task to judge the correctness of image descriptions generated by a multi-modal large language model (MLLM). Based on an averaged event-related potential (ERP) study, we reveal that multiple cognitive processes, e.g., semantic integration, inferential processing, memory retrieval, and cognitive load, exhibit distinct patterns when humans process hallucinated versus non-hallucinated content. Notably, neural responses to hallucinations that were misjudged versus correctly judged by human participants showed significant differences. This indicates that misjudged AI-generated hallucinations failed to trigger the standard neurocognitive fact verification pathway.
Authors: Nicanor Mayumu, Xiaoheng Deng, Patrick Mukala
Abstract: We present the first systematic study of faithfulness in Vision-Language-Action (VLA) driving models, analyzing 300 Alpamayo-R1-10B inferences across 100 diverse PhysicalAI-AV scenarios. Our main finding is that output natural-language rationales with trajectories may be significantly unfaithful: (i) overall reasoning fidelity is only 42.5%, with Chain-of-Causation matching scene reality less than half the time; (ii) 94 missed pedestrians in one-third of pedestrian-relevant scenes; (iii) 97.7% trajectory fragility under mild visual perturbations; and (iv) only 48.3% mean reasoning-action consistency, with 53.3% of inferences exhibiting low consistency, including 37.9% of stop-claimed cases where the model continues instead. We formalize faithfulness information-theoretically, define entity and action fidelity with verification criteria, and outline a four-component safety architecture aligned with these results.
Authors: Yuanfei Xu, Lin Liu, Wengang Zhou, Mingxiao Feng, Houqiang Li
Abstract: 3D open-world environments with adversarial opponents remain a core challenge for reinforcement learning due to their vast state spaces. Effective reasoning representations are essential in such settings. While existing self-supervised visual foresight reasoning approaches often suffer from multi-step error accumulation, many recent studies resort to injecting domain-specific knowledge for more stable guidance. Our key insight is that the photorealistic fidelity of visual reasoning representations is secondary; what truly matters is providing informative, task-relevant signals. To this end, we propose ResDreamer, a hierarchical world model in which each higher-level layer is trained to reconstruct the residuals of the layer below. This design enables progressive abstraction of increasingly sophisticated world dynamics and fosters the emergence of richer latent representations. Drawing inspiration from the "Bitter Lesson", ResDreamer trains its reasoning representations in a purely self-supervised manner. The higher-level residual representations are used to modulate lower-level predictions, allowing the world model to scale effectively with only linearly increasing cross-layer communication costs. Experiments show that ResDreamer achieves state-of-the-art sample efficiency and parameter efficiency. This scalable hierarchical visual foresight reasoning architecture paves the way for more capable online RL agents in open-ended, dynamic environments. The code is accessible at https://github.com/XuYuanFei01/ResDreamer.
Authors: Jun-Yu Pan, Yansen Wang, Enze Zhang, Bao-Liang Lu, Wei-Long Zheng, Dongsheng Li
Abstract: Leveraging the universal representations of pre-trained LLMs and MLLMs offers a promising path toward brain foundation models. However, visually-evoked EEG datasets remain scarce, leading existing methods to align neural signals mainly with abstract text, a lossy translation that may discard fine-grained perceptual information encoded in brain activity. We propose Generative Visual Grounding (GVG), a framework that visualizes the invisible by using an EEG-to-image generative model as a visual translator. Instead of forcing EEG into text alone, GVG hallucinates instance-specific proxy images for non-visual EEG, providing structured visual contexts that allow MLLMs to exploit their visual priors for clinical-state interpretation. We validate this idea on two MLLM backbones, GVG-X-Omni and GVG-Janus. Image-only alignment is already competitive: the lightweight GVG-X-Omni matches 1.7B-parameter text-aligned baselines while tuning only 170M parameters on a frozen 7B backbone. We further extend GVG-Janus with trimodal Image+Text alignment, where text supplies categorical semantic anchors and visual proxies enrich neural representations with perceptual details. Experiments show consistent gains in EEG understanding and visual generation, suggesting visual proxy grounding as an effective complement to textual alignment.
Authors: Samuel Jacob Chacko, James Hugglestone, Chashi Mahiul Islam, Xiuwen Liu
Abstract: Agent Skills, structured packages of procedural knowledge loaded into an LLM agent at inference time, are widely reported to improve task pass rates by an average of 16.2~percentage points across diverse domains. Yet the same benchmarks show wide variance, with 16 of 84 tasks suffering negative deltas when Skills are introduced. The community has not yet articulated a clean mechanism for \emph{when} Skills help and when they are merely redundant overhead. We re-analyze a recently published 180-run controlled study of an MCP-grounded autonomous Capture-the-Flag (CTF) agent under four documentation conditions of increasing richness (591, 12865, 17253, and 36001 tokens) and show that these conditions correspond almost exactly to a No-Skills, Experiential-Skills, Curated-Skills, and Comprehensive-Skills ablation. In offensive cybersecurity, a domain not deeply covered by existing Skills benchmarks, the marginal benefit of Skills collapses. The spread between the no-Skills and full-Skills conditions is only 8.9~pp ($p = 0.71$, $\chi^2$; $p = 0.25$, Cochran--Armitage trend test; five of six pairwise Cohen's $h$ values fall below the $0.2$ small-effect threshold). We argue that the missing variable is \emph{environment-feedback bandwidth}. When an agent's tool layer returns strict, schema-validated, low-latency observations, the environment itself supplies the procedural correction signal that Skills are normally needed to provide. As a result, the marginal benefit of curated Skills diminishes substantially, and, in some cases (e.g., our timing side-channel setting), actively degrades performance. We articulate a falsifiable hypothesis, sketch its design implications for compound AI systems, and will release the reanalysis pipeline to support replication.
Authors: Jiaqi Liu, Shi Qiu, Mairui Li, Bingzhou Li, Haonian Ji, Siwei Han, Xinyu Ye, Peng Xia, Zihan Dong, Meng Chen, Congyu Zhang, Letian Zhang, Guiming Chen, Haoqin Tu, Xinyu Yang, Lu Feng, Xujiang Zhao, Haifeng Chen, Jiawei Zhou, Xiao Wang, Weitong Zhang, Hongtu Zhu, Yun Li, Jieru Mei, Hongliang Fei, Jiaheng Zhang, Linjie Li, Linjun Zhang, Yuyin Zhou, Sheng Wang, Caiming Xiong, James Zou, Zeyu Zheng, Cihang Xie, Mingyu Ding, Huaxiu Yao
Abstract: Automating scientific discovery requires more than generating papers from ideas. Real research is iterative: hypotheses are challenged from multiple perspectives, experiments fail and inform the next attempt, and lessons accumulate across cycles. Existing autonomous research systems often model this process as a linear pipeline: they rely on single-agent reasoning, stop when execution fails, and do not carry experience across runs. We present AutoResearchClaw, a multi-agent autonomous research pipeline built on five mechanisms: structured multi-agent debate for hypothesis generation and result analysis, a self-healing executor with a \textsc{Pivot}/\textsc{Refine} decision loop that transforms failures into information, verifiable result reporting that prevents fabricated numbers and hallucinated citations, human-in-the-loop collaboration with seven intervention modes spanning full autonomy to step-by-step oversight, and cross-run evolution that converts past mistakes into future safeguards. On ARC-Bench, a 25-topic experiment-stage benchmark, AutoResearchClaw outperforms AI Scientist v2 by 54.7%. A human-in-the-loop ablation across seven intervention modes reveals that precise, targeted collaboration at high-leverage decision points consistently outperforms both full autonomy and exhaustive step-by-step oversight. We position AutoResearchClaw as a research amplifier that augments rather than replaces human scientific judgment. Code is available at https://github.com/aiming-lab/AutoResearchClaw.
Authors: Lautaro Estienne, Erik Ernst, Mat\'ias Vera, Pablo Piantanida, Luciana Ferrer
Abstract: In high-stakes automated decision-making, access to predictive uncertainty is essential for enabling users -- human or downstream systems -- to accept or reject predictions based on application-specific cost trade-offs. Such uncertainty-augmented (UA) systems -- i.e., systems that output both predictions and uncertainty scores -- are currently being assessed in the literature in a variety of ways, using separate metrics to evaluate the predictions and the uncertainty scores, setting a cost function with a fixed rejection cost or integrating over a coverage-risk curve. We argue that these evaluation approaches are inadequate for assessing overall performance of the UA system for decision making under uncertainty and propose a novel family of metrics, ECUAS$_n$, formulated as proper scoring rules for the task of interest. The parameter $n$ controls the trade-off between the cost of incorrect predictions and imperfect uncertainties depending on the needs of the use-case. We demonstrate the advantages of the ECUAS$_n$ metrics both theoretically and empirically, through experiments on diverse classification and generation datasets, including a manually annotated subset of TriviaQA.
Authors: Dylan Feng, Pragya Srivastava, Anca Dragan, Cassidy Laidlaw
Abstract: Many safety and alignment failures of large language models (LLMs) occur due to out-of-distribution (OOD) situations: unusual prompt or response patterns that are unforeseen by model developers. We systematically study whether LLM monitoring pipelines can detect these OOD alignment failures by introducing a benchmark called Misalignment Out Of Distribution (MOOD). It is difficult to find failures that are truly OOD for off-the-shelf models trained on vast safety datasets. We sidestep this by including a restricted training set in MOOD that we use to train our own monitors, as well as seven test sets with diverse alignment failures that are outside the training distribution. Using MOOD, we find that guard models (safety classifiers) often fail to generalize OOD. To fix this, we propose combining guard models with OOD detectors. We test four types of OOD detectors and find that a combination of a guard model with Mahalanobis distance and perplexity-based OOD detectors can improve recall from 39% to 45%. We also establish positive scaling trends across model scales for monitors that combine a guard model and OOD detector; we find that incorporating OOD detection into monitoring achieves a higher recall gain than using a guard model with 20 times more parameters. Our work suggests that OOD detection should be a crucial component of LLM monitoring and provides a foundation for further work on this important problem. We release the code and data for our experiments publicly, and you can find the relevant links here: https://github.com/Dylan102938/mood-bench.
Authors: Kevin Han, Renfei Zhang, Kathy Wei, Hamed Mahdavi, Niloofar Mireshghallah, Amir Barati Farimani
Abstract: LLM agents have incredible potential for scientific discovery applications. However, the performance of LLM agents on real-world, small molecule drug design (SMDD) tasks across diverse chemistries and targets is unclear. Current evaluation methods are either ad hoc, too simple for real-world discovery, limited in scale, or restricted to single-turn question answering. In effort to standardize the evaluation of LLM agents on small molecule design, we introduce SMDD-Bench, a challenging, multi-turn, long-horizon agentic benchmark consisting of 502 guaranteed-solvable task instances spanning 5 task types: 2D Pharmacophore Identification, Interaction Point Discovery, Scaffold Hopping, Lead Optimization, and Fragment Assembly. SMDD-Bench tasks span a wide region of chemical space and involve 102 unique protein targets. Completely solving the benchmark would require having strong chemical and biological reasoning and 3D intuition, understanding specialized tool use, and displaying planning expertise over a limited number of oracle calls. We benchmark 7 frontier open and closed source LLMs and find even the most performant LLM, GPT5.4, solves only 40.2\% of tasks. We hope SMDD-Bench provides a standardized testbed to invigorate the field towards training and evaluating LLM agents for fully autonomous computational drug design. We host a public leaderboard at smddbench.com .
Authors: Enrico Daga, Valentina Tamma, Terry Payne
Abstract: Knowledge graphs have become the primary vehicle for data integration and are critical to the success of modern AI, but the diversity of KG modelling practices, from lightweight vocabularies to richly axiomatised ontologies, makes integration and reuse expensive and brittle. This challenge is particularly acute in neuro-symbolic AI, where bridging neural and symbolic components depends on the ability to reengineer KGs to fit new requirements; GenAI now offers unprecedented automation capability, but without a principled understanding of the KG space, such automation remains conceptually ungrounded. We introduce the ontological continuum as that missing conceptualisation, a theoretical construct a theoretical construct whose characterisation framework is defined by two orthogonal distinctions: semantics vs pragmatics, and properties vs affordances; together these define a vocabulary to describe, compare, navigate, and transform KGs across the full range of modelling practices. The methodological stance is empirical: rather than prescribing how KGs should be modelled, the continuum aims to define a theory of the existent, derived from observation of real-world KG engineering practices and whose structure can be made formally explicit, for example, through Formal Concept Analysis (FCA). We ground the vision through a case study on provenance knowledge, showing how a single concern manifests differently across the continuum. We articulate five open research challenges and invite the community to develop the ontological continuum as a shared research agenda.
Authors: Wei Luo, Yi Huang, Songchen Ma, Huanyu Qu, Jiang Cai, Mingkun Xu
Abstract: The KV cache used in large language models has linearly growing time complexity, so LLMs face memory blow-up and reduced decoding efficiency when they process long contexts. Current KV Cache eviction has become an important research direction; however, existing methods based on fixed Soft Tokens (e.g., Judge Q) rely on a static parameter set as the query to evaluate the importance of KV pairs, so they cannot adapt dynamically to different input prompts, and they cannot precisely capture complex and changing task relevance. Also, evicted KV pairs are discarded permanently, so this causes irreversible information loss and context breaks. To address this problem, we propose Meta-Soft, a dynamic compression framework based on probe-driven context integration. Specifically, we build a meta-library with a learnable orthogonal basis matrix $\mathcal{L}$, and we use a selector network with Gumbel-Softmax to produce differentiable sparse combination weights, so we dynamically synthesize the most targeted $k$ Soft Tokens from the input prompt features. We append these Soft Tokens to the end of the input sequence to probe key information. We also introduce an attention-flow based integration mechanism, which redistributes the semantic information of removed tokens into retained tokens, and this keeps the dropped context information effectively. Experiments on multiple datasets show that our method outperforms existing state-of-the-art eviction methods and provides a new solution for KV Cache compression.
Authors: Sid-Ali Temkit
Abstract: Large language models are routinely used as automated evaluators: to review code, moderate content, or score outputs, often with many items passing through one conversation. We ask whether the polarity of prior conversation history biases subsequent judgments, an effect we call the accumulated message effect on LLM judgments (AMEL). Across 75,898 API calls to 11 models from 4 providers (OpenAI, Anthropic, Google, and four open-source models), we present identical test items in isolation or following histories saturated with predominantly positive or negative evaluations. Models shift toward the conversation's prevailing polarity (d = -0.17, p < 10^-46). The effect concentrates on items where the model is genuinely uncertain at baseline (d = -0.34 for high-entropy items, vs d = -0.15 when the baseline is deterministic). Bias does not grow with context length: 5 prior turns and 50 produce the same shift (Spearman |r| < 0.01; OLS slope p = 0.80). And there is a negativity asymmetry: paired per item, negative histories induce 1.62x more bias than positive (t = 13.46, p < 10^-39, n = 2,481). Scaling helps but does not solve it (Anthropic: Haiku -0.22 to Opus -0.17; OpenAI: Nano -0.34 to GPT-5.2 -0.17). Three follow-ups narrow the mechanism. The token probability distribution shifts continuously, not at a threshold. The negativity asymmetry has both token-level and semantic components, though attributing the balance is exploratory at our sample sizes. Position does not matter: five biased turns anywhere in a 50-turn history produce the same shift. The simplest fix for evaluation pipelines is a fresh context per item; when batching is unavoidable, balancing the history helps.
Authors: Qianshu Cai, Yonggang Zhang, Xianzhang Jia, Huajiang Zheng, Wei Xue, Jun Song, Xinmei Tian, Yike Guo
Abstract: Autonomous agentic systems are largely static after deployment: they do not learn from user interactions, and recurring failures persist until the next human-driven update ships a fix. Self-evolving agents have emerged in response, but all confine evolution to text-mutable artifacts -- skill files, prompt configurations, memory schemas, workflow graphs -- and leave the agent harness untouched. Since routing, hook ordering, state invariants, and dispatch live in code rather than in any text artifact, an entire class of structural failure is physically unreachable from the text layer. We argue that source-level adaptation is a fundamentally more general medium: it is Turing-complete, a strict superset of every text-mutable scope, takes effect deterministically rather than through base-model compliance, and does not erode under long-context drift. We present MOSS, a system that performs self-rewriting at the source level on production agentic substrates. Each evolution is anchored to an automatically curated batch of production-failure evidence and proceeds through a deterministic multi-stage pipeline; code modification is delegated to a pluggable external coding-agent CLI while MOSS retains stage ordering and verdicts. Candidates are verified by replaying the batch against the candidate image in ephemeral trial workers, then promoted via user-consent-gated, in-place container swap with health-probe-gated rollback. On OpenClaw, MOSS lifts a four-task mean grader score from 0.25 to 0.61 in a single cycle without human intervention.
Authors: Yifan Yang, Ziyang Gong, Weiquan Huang, Qihao Yang, Ziwei Zhou, Zisu Huang, Yan Li, Xuemei Gao, Qi Dai, Bei Liu, Kai Qiu, Yuqing Yang, Dongdong Chen, Xue Yang, Chong Luo
Abstract: Agent skills today are hand-crafted, generated one-shot, or evolved through loosely controlled self-revision, none of which behaves like a deep-learning optimizer for the skill, and none of which reliably improves over its starting point under feedback. We argue the skill should instead be trained as the external state of a frozen agent, with the same discipline that makes weight-space optimization reproducible. SkillOpt is, to our knowledge, the first systematic controllable text-space optimizer for agent skills: a separate optimizer model turns scored rollouts into bounded add/delete/replace edits on a single skill document, and an edit is accepted only when it strictly improves a held-out validation score. A textual learning-rate budget, rejected-edit buffer, and epoch-wise slow/meta update make skill training stable while adding zero inference-time model calls at deployment. Across six benchmarks, seven target models, and three execution harnesses (direct chat, Codex, Claude Code), SkillOpt is best or tied on all 52 evaluated (model, benchmark, harness) cells and beats every per-cell competitor among human, one-shot LLM, Trace2Skill, TextGrad, GEPA, and EvoSkill skills. On GPT-5.5 it lifts the average no-skill accuracy by +23.5 points in direct chat, by +24.8 inside the Codex agentic loop, and by +19.1 inside Claude Code. Transfer experiments further show that optimized skill artifacts retain value when moved across model scales, between Codex and Claude Code execution environments, and to a nearby math benchmark without further optimization. Code: https://aka.ms/skillopt
URLs: https://aka.ms/skillopt
Authors: Xiang Fang, Yuchong Hu, Pan Zhou, Dapeng Oliver Wu
Abstract: Incomplete multi-view clustering is an important technique to deal with real-world incomplete multi-view data. Previous works assume that all views have the same incompleteness, i.e., balanced incompleteness. However, different views often have distinct incompleteness, i.e., unbalanced incompleteness, which results in strong views (low-incompleteness views) and weak views (high-incompleteness views). The unbalanced incompleteness prevents us from directly using the previous methods for clustering. In this paper, inspired by the effective biological evolution theory, we design the novel scheme of view evolution to cluster strong and weak views. Moreover, we propose an Unbalanced Incomplete Multi-view Clustering method (UIMC), which is the first effective method based on view evolution for unbalanced incomplete multi-view clustering. Compared with previous methods, UIMC has two unique advantages: 1) it proposes weighted multi-view subspace clustering to integrate these unbalanced incomplete views, which effectively solves the unbalanced incomplete multi-view problem; 2) it designs the low-rank and robust representation to recover the data, which diminishes the impact of the incompleteness and noises. Extensive experimental results demonstrate that UIMC improves the clustering performance by up to 40% on three evaluation metrics over other state-of-the-art methods.
Authors: Xiang Fang, Yuchong Hu
Abstract: Multi-view clustering has been applied in many real-world applications where original data often contain noises. Some graph-based multi-view clustering methods have been proposed to try to reduce the negative influence of noises. However, previous graph-based multi-view clustering methods treat all features equally even if there are redundant features or noises, which is obviously unreasonable. In this paper, we propose a novel multi-view clustering framework Double Self-weighted Multi-view Clustering (DSMC) to overcome the aforementioned deficiency. DSMC performs double self-weighted operations to remove redundant features and noises from each graph, thereby obtaining robust graphs. For the first self-weighted operation, it assigns different weights to different features by introducing an adaptive weight matrix, which can reinforce the role of the important features in the joint representation and make each graph robust. For the second self-weighting operation, it weights different graphs by imposing an adaptive weight factor, which can assign larger weights to more robust graphs. Furthermore, by designing an adaptive multiple graphs fusion, we can fuse the features in the different graphs to integrate these graphs for clustering. Experiments on six real-world datasets demonstrate its advantages over other state-of-the-art multi-view clustering methods.
Authors: Xiang Fang, Daizong Liu, Pan Zhou, Yuchong Hu
Abstract: As an increasingly popular task in multimedia information retrieval, video moment retrieval (VMR) aims to localize the target moment from an untrimmed video according to a given language query. Most previous methods depend heavily on numerous manual annotations (i.e., moment boundaries), which are extremely expensive to acquire in practice. In addition, due to the domain gap between different datasets, directly applying these pre-trained models to an unseen domain leads to a significant performance drop. In this paper, we focus on a novel task: cross-domain VMR, where fully-annotated datasets are available in one domain (``source domain''), but the domain of interest (``target domain'') only contains unannotated datasets. As far as we know, we present the first study on cross-domain VMR. To address this new task, we propose a novel Multi-Modal Cross-Domain Alignment (MMCDA) network to transfer the annotation knowledge from the source domain to the target domain. However, due to the domain discrepancy between the source and target domains and the semantic gap between videos and queries, directly applying trained models to the target domain generally leads to a performance drop. To solve this problem, we develop three novel modules: (i) a domain alignment module is designed to align the feature distributions between different domains of each modality; (ii) a cross-modal alignment module aims to map both video and query features into a joint embedding space and to align the feature distributions between different modalities in the target domain; (iii) a specific alignment module tries to obtain the fine-grained similarity between a specific frame and the given query for optimal localization. By jointly training these three modules, our MMCDA can learn domain-invariant and semantic-aligned cross-modal representations.
Authors: Xiang Fang, Daizong Liu, Pan Zhou, Guoshun Nan
Abstract: Given an untrimmed video, temporal sentence grounding (TSG) aims to locate a target moment semantically according to a sentence query. Although previous respectable works have made decent success, they only focus on high-level visual features extracted from the consecutive decoded frames and fail to handle the compressed videos for query modelling, suffering from insufficient representation capability and significant computational complexity during training and testing. In this paper, we pose a new setting, compressed-domain TSG, which directly utilizes compressed videos rather than fully-decompressed frames as the visual input. To handle the raw video bit-stream input, we propose a novel Three-branch Compressed-domain Spatial-temporal Fusion (TCSF) framework, which extracts and aggregates three kinds of low-level visual features (I-frame, motion vector and residual features) for effective and efficient grounding. Particularly, instead of encoding the whole decoded frames like previous works, we capture the appearance representation by only learning the I-frame feature to reduce delay or latency. Besides, we explore the motion information not only by learning the motion vector feature, but also by exploring the relations of neighboring frames via the residual feature. In this way, a three-branch spatial-temporal attention layer with an adaptive motion-appearance fusion module is further designed to extract and aggregate both appearance and motion information for the final grounding. Experiments on three challenging datasets shows that our TCSF achieves better performance than other state-of-the-art methods with lower complexity.
Authors: Pengyi Li, Jianye Hao, Hongyao Tang, Xian Fu, Yan Zheng, Ke Tang
Abstract: Evolutionary Reinforcement Learning (ERL), which integrates Evolutionary Algorithms (EAs) and Reinforcement Learning (RL) for optimization, has demonstrated remarkable performance advancements. By fusing both approaches, ERL has emerged as a promising research direction. This survey offers a comprehensive overview of the diverse research branches in ERL. Specifically, we systematically summarize recent advancements in related algorithms and identify three primary research directions: EA-assisted Optimization of RL, RL-assisted Optimization of EA, and synergistic optimization of EA and RL. Following that, we conduct an in-depth analysis of each research direction, organizing multiple research branches. We elucidate the problems that each branch aims to tackle and how the integration of EAs and RL addresses these challenges. In conclusion, we discuss potential challenges and prospective future research directions across various research directions. To facilitate researchers in delving into ERL, we organize the algorithms and codes involved on https://github.com/yeshenpy/Awesome-Evolutionary-Reinforcement-Learning.
URLs: https://github.com/yeshenpy/Awesome-Evolutionary-Reinforcement-Learning.
Authors: Samar Fares, Klea Ziu, Toluwani Aremu, Nikita Durasov, Martin Tak\'a\v{c}, Pascal Fua, Ivan Laptev, Karthik Nandakumar
Abstract: Vision-Language Models (VLMs) are increasingly susceptible to sophisticated adversarial attacks, including adaptive strategies specifically designed to bypass existing defenses. To address this vulnerability, we propose MirrorCheck, a robust and model-agnostic detection framework that operates effectively in both unimodal and multimodal settings. MirrorCheck leverages Text-to-Image (T2I) models to regenerate visual content from captions produced by the target model and assesses semantic consistency by comparing feature-space embeddings between the original and synthesized images. To enhance robustness against adaptive attacks, MirrorCheck introduces a stochastic defense strategy that randomly selects T2I generators and image encoders from a diverse model zoo. Additionally, we incorporate a novel One-Time-Use (OTU) perturbation applied to the selected encoder embeddings, regulated by a scaling factor, which decreases the effectiveness of adaptive attacks. Extensive experiments across multiple threat scenarios demonstrate that MirrorCheck consistently outperforms baseline methods, and maintains its utility even under strong adaptive adversarial conditions.
Authors: Doron Yeverechyahu, Raveesh Mayya, Gal Oestreicher-Singer
Abstract: Large Language Models (LLMs) are reshaping knowledge work, yet their impact on voluntary, self-guided open innovation forums (contributors choose tasks without managerial direction) may differ fundamentally from effects observed in organizational settings. We study this question in open-source software development, where individuals' contributions collectively drive innovation at a community level. Unlike product innovation, where typologies for classifying innovation are well established, knowledge work in open-source settings calls for a distinction grounded in the cognitive demand a task places on the contributor. Burgeoning literature distinguishes substantive contributions, which require creative problem formulation to introduce new functionality, from incremental contributions, which draw on comprehension of existing code to maintain and refine it. We exploit a natural experiment around GitHub Copilot's launch in October 2021, where Copilot supported languages like Python while not supporting R for business reasons, creating an exogenous partition between otherwise comparable ecosystems. Using three complementary identification strategies and two classification approaches, we find that Copilot availability increases open-source contributions by 28 to 40 percent. The increase in incremental contributions is significantly larger than the increase in substantive contributions across all specifications. This disparity is more pronounced in projects with higher activity levels and widens following a model upgrade: LLMs function more effectively when existing context helps define the problem and constrain solutions, tilting collaborative innovation toward exploitation of established codebases rather than exploration of new functionality. This paper provides a rare instance of causal field evidence on LLM effects, given the speed at which GenAI has exploded across the knowledge economy.
Authors: Safeyah Khaled Alshemali, Daniel Bauer, Yuval Marton
Abstract: The thematic fit estimation task measures semantic arguments' compatibility with a given semantic role for a given predicate. We investigate if autoregressive LLMs have consistent, expressible knowledge of event arguments' thematic fit by experimenting with various prompt designs, manipulating input context, reasoning, and output forms. We set a new state-of-the-art on thematic fit benchmarks, but show that closed and open weight LLMs respond differently to our prompting strategies: Closed models achieve better scores overall and benefit from multi-step reasoning, but they perform worse at filtering out generated sentences incompatible with the given predicate, role, and argument. Our analysis shows that lemma tuple input and sentence input result in surprisingly different thematic fit score distributions.
Authors: Ho-Chun Herbert Chang, Benjamin Shaman, Yung-chun Chen, Mingyue Zha, Sean Noh, Chiyu Wei, Tracy Weener, Maya Magee
Abstract: Visual content on social media has become increasingly influential in shaping political discourse and civic engagement, but it also limits participation due to the increased cost of multimedia production. In tandem, the growth of generative AI provides novel ways for citizens to participate in politics by lowering these costs. Drawing on a dataset of 239,526 Instagram images, we analyze the effects of synthetic images during the 2024 United States presidential election, using a multimodal workflow combining computer vision, large language models, and facial affect analysis. Results show that meme format is a stronger predictor of engagement than AI-generated content alone. However, AI-generated memes yield a significant interaction effect, suggesting synergistic increases in engagement when synthetic imagery is integrated with memes through human curation. We also characterize how users curate images. Partisans use AI in different ways: Democrat-leaning users tend to use it for in-group support, whereas Republican-leaning users more often employ it for out-group attacks. Users generally select happier synthetic faces compared to real photographs. We define generative memesis as a mode of communication in which memes are no longer shared person-to-person, but mediated by AI through customized visuals. We discuss how generative AI may empower civic participation, the bifurcation of content production and curation, and its implications for in the history of novel technologies and participatory culture.
Authors: Donghwna Lee, Kirok Kim, Jisu Lee, Kyungha Min, Wooju Kim
Abstract: Pose-Guided Person Image Synthesis (PGPIS) aims to generate human images in specified poses while preserving the identity and appearance of a source image. This technology facilitates diverse applications, including virtual try-on, digital avatars, animation, and sign language generation. Despite the high-quality results of recent diffusion-based PGPIS, these models typically depend on implicit feature aggregation within the denoising process. As a result, fine-grained texture preservation is limited, and even for the same identity, it is difficult to ensure consistent generation under variations in pose and source appearance. To address these limitations, we propose Fusion Embedding for PGPIS using a Diffusion Model (FPDM), the first framework that explicitly aligns fused source-pose embeddings with target image embeddings via contrastive learning, and subsequently employs the learned fusion embedding as a conditioning signal for generation. FPDM integrates an Image-Pose Fusion (IPF) module into our proposed Source-Enhanced Pose Fusion approach to learn a fusion embedding aligned with the target image. We then employ a conditional diffusion model guided by source appearance, target pose, and the learned fusion embedding. Experiments on the DeepFashion benchmark and the RWTH-PHOENIX-Weather 2014T dataset demonstrate competitive performance compared to existing methods in both quantitative and qualitative evaluations, with ablation studies confirming that explicit fusion embedding alignment substantially improves texture fidelity and consistency across pose and source appearance variations.
Authors: Ankur Samanta, Rohan Gupta, Aditi Misra, Christian McIntosh Clarke, Jayakumar Rajadas
Abstract: Molecular representation learning methods typically tokenize molecules as individual atoms or use rigid, rule-based fragment decompositions, limiting their ability to capture meaningful chemical substructure context. We introduce FragmentNet, a graph-to-sequence model built around a novel adaptive, learned tokenizer that decomposes molecular graphs into chemically valid fragments of adjustable granularity, complemented by chemically aware spatial positional encodings that preserve molecular topology in the resulting sequence. Extending masked pre-training strategies from natural language processing to the molecular domain, we mask and reconstruct molecules at the level of chemically meaningful fragments rather than individual atoms. Evaluating across multiple property prediction benchmarks, we find that pre-training at fragment granularity leads to improved downstream performance on the majority of tasks, demonstrating that tokenization granularity is an important design choice for molecular representation learning.
Authors: Vladislav Trifonov, Ekaterina Muravleva, Ivan Oseledets
Abstract: Graph Neural Networks (GNNs) have been proposed as a tool for learning sparse matrix preconditioners, which are key components in accelerating linear solvers. We present theoretical and empirical evidence that message-passing GNNs are fundamentally incapable of approximating sparse triangular factorizations for classes of matrices for which high-quality preconditioners exist but require non-local dependencies. To illustrate this, we construct a set of baselines using both synthetic matrices and real-world examples from the SuiteSparse collection. Across a range of GNN architectures, including Graph Attention Networks and Graph Transformers, we observe low cosine similarity ($\leq0.7$ in key cases) between predicted and reference factors. Our theoretical and empirical results suggest that architectural innovations beyond message-passing are necessary for applying GNNs to scientific computing tasks such as matrix factorization. Moreover, experiments demonstrate that overcoming non-locality alone is insufficient. Tailored architectures are necessary to capture the required dependencies since even a completely non-local Global Graph Transformer fails to match the proposed baselines.
Authors: Jusheng Zhang, Yijia Fan, Kaitong Cai, Keze Wang, Wenhao Wang
Abstract: Although Kolmogorov-Arnold-based interpretable networks (KANs) possess strong theoretical expressiveness, they suffer from severe parameter explosion and limited ability to capture high-frequency features in high-dimensional tasks. To address these issues, we propose the Kolmogorov-Arnold Fourier Network (KAF), which fundamentally redefines the KAN paradigm through spectral reparameterization. Our key contributions include: (1) proposing a fundamental basis transformation from the local, grid-based B-spline representation to a global, adaptive spectral representation. This shift changes the network's inductive bias, reducing parameter complexity from $O(G)$ to $O(1)$ while preserving expressiveness; (2) introducing trainable Random Fourier Features (RFF) initialized via a spectral alignment strategy, which allows the model to break the smoothness limitation of fixed kernels and accurately capture high-frequency components; and (3) implementing an adaptive hybrid GELU-Fourier activation mechanism that progressively enhances frequency representation during training. Comprehensive experiments demonstrate the superiority of KAF across computer vision (CV), natural language processing (NLP), audio, and partial differential equation (PDE) solving tasks, achieving state-of-the-art performance with improved efficiency. The code is available at https://github.com/kolmogorovArnoldFourierNetwork/KAF.
URLs: https://github.com/kolmogorovArnoldFourierNetwork/KAF.
Authors: Lauri Sepp\"al\"ainen, Mudong Guo, Kai Puolam\"aki
Abstract: Most commonly used non-linear machine learning methods are closed-box models, uninterpretable to humans. The field of explainable artificial intelligence (XAI) aims to develop tools to examine the inner workings of these closed boxes. An often-used model-agnostic approach to XAI involves using simple models as local approximations to produce so-called local explanations; examples of this approach include LIME, SHAP, and SLISEMAP. This paper shows how a large set of local explanations can be reduced to a small "proxy set" of simple models, which can act as a generative global explanation. This reduction procedure, ExplainReduce, can be formulated as an optimisation problem and approximated efficiently using greedy heuristics. We show that, for many problems, as few as five explanations can faithfully emulate the closed-box model and that our reduction procedure is competitive with other model aggregation methods.
Authors: Zhuchen Cao, Sven Apel, Adish Singla, Vera Demberg
Abstract: Pragmatic reasoning helps interlocutors infer intended meaning from ambiguous or underspecified messages by considering shared context and counterfactual alternatives. Similar challenges arise in natural language-to-code generation, where user instructions often admit multiple plausible candidate programs. However, direct RSA-style inference is difficult because it requires probability estimation over large spaces of programs and alternative instructions. We propose CodeRSA, an RSA-motivated reranking method that makes pragmatic reasoning tractable through local pragmatic contests among sampled code candidates. CodeRSA constructs candidate-induced alternative instructions and estimates which candidates are most distinctively supported by the original instruction, avoiding global normalization over the full program-instruction space. We evaluate CodeRSA on HumanEval+, MBPP+, and BigCodeBench using four open-weight instruction-following models. CodeRSA achieves the strongest average accuracy in 10 of 12 model-benchmark settings and remains competitive in the remaining cases. Further analyses show that its gains come from combining local pairwise pragmatic comparison with broader global support, suggesting a scalable direction for language-to-code reranking under natural-language uncertainty.
Authors: Zhenlong Liu, Wenyu Jiang, Feng Zhou, Hongxin Wei
Abstract: Membership Inference Attacks (MIAs) aim to estimate whether a specific data point was used in the training of a given model. Existing state-of-the-art attacks typically rely on training multiple reference models to approximate the conditional score distribution for individual data points, which leads to significant computational overhead and limits their practical applicability. In this work, we propose a novel approach -- Bayesian Membership Inference Attack (BMIA), which performs conditional attack through Bayesian sampling. Specifically, we apply Laplace approximation to a single reference model to obtain a posterior over model parameters, enabling direct estimation of the conditional score distribution. Theoretically, we demonstrate that Bayesian sampling reduces intra-model variance, thereby improving attack power. This insight naturally motivates the multi-reference variant that further enhances performance when additional reference models are available. Extensive experiments across image, text, and tabular datasets indicate that our method achieves state-of-the-art performance in both effectiveness and efficiency.
Authors: Xiaoping Sun, Hai Zhuge
Abstract: Organizing large-scale resources in a multidimensional semantic space is an approach to efficiently managing and querying resources from different semantic dimensions. To support advanced applications, this paper proposes a resource space model for aggregation query on subspaces defined by a range within the partial order on the coordinate trees representing each dimension, where each point in the subspace contains resources aggregated along the paths of the partial order relations on the coordinate trees and the aggregated resources at each point can be measured, ranked and selected by applications. To efficiently locate non-empty points in a large subspace, an approach to generating graph index is proposed to build partial order relations on coordinates of dimensions to enable a subspace query to reach non-empty points through indexing links and aggregate resources along indexing paths to their super points. Generating such an index is costly as the number of children of an indexing node can be large so that the total number of indexing nodes can be very large (exponentially growing with the number of dimensions and scale of dimensions). The proposed approach adopts the a set of strategies to reduce the cost. Analysis and experiments show the effectiveness of the generated index in supporting subspace aggregation query.
Authors: Sriram Mandalika
Abstract: Few-shot adaptation of vision-language models remains fundamentally limited by how negative class signals are handled at inference. Existing methods apply uniform negative suppression across all queries, ignoring that the most damaging confusions are query-specific and shift with support-set geometry. We introduce SCAN (Selective Confusion-Aware Negatives), a framework that addresses this gap through three targeted contributions. In inference, query-adaptive negative routing restricts suppression to the top-K most confusable classes per query, requiring zero additional parameters. Generic negative text templates are replaced with LLM-bootstrapped contrastive prompts that describe discriminative attributes between confusable class pairs, sharpening the textual decision boundary where it matters most. A parameter-free adaptive fusion weight estimated from support-set Fisher discriminability removes the need for manual tuning of the vision-language trade-off. Evaluated across 11 standard benchmarks, SCAN consistently outperforms prior prompt-based and adapter-based methods by an average of 4.61% at 16-shot, with gains of up to 7.70% on fine-grained datasets where inter-class confusion is most severe. SCAN also generalizes strongly under distribution shift, improving by 2.95% on average across four ImageNet OOD variants, and maintains robust performance under significant label noise, with accuracy under 50% label corruption still exceeding the clean baseline of the strongest competing method.
Authors: Yuezhou Ma, Haixu Wu, Hang Zhou, Huikun Weng, Jianmin Wang, Mingsheng Long
Abstract: Physics sensing plays a central role in many scientific and engineering domains, which inherently involves two coupled tasks: reconstructing dense physical fields from sparse observations and optimizing scattered sensor placements to observe maximum information. While deep learning has made rapid advances in sparse-data reconstruction, existing methods generally omit optimization of sensor placements, leaving the mutual enhancement between reconstruction and placement on the shelf. To change this suboptimal practice, we propose PhySense, a synergistic two-stage framework that learns to jointly reconstruct physical fields and to optimize sensor placements, both aiming for accurate physics sensing. The first stage involves a flow-based generative model enhanced by cross-attention to adaptively fuse sparse observations. Leveraging the reconstruction feedback, the second stage performs sensor placement via projected gradient descent to satisfy spatial constraints. We further prove that the learning objectives of the two stages are consistent with classical variance-minimization principles, providing theoretical guarantees. Extensive experiments across three challenging benchmarks, especially a 3D geometry dataset, indicate PhySense achieves state-of-the-art physics sensing accuracy and discovers informative sensor placements previously unconsidered. Code is available at this repository: https://github.com/thuml/PhySense.
Authors: Ruishuo Chen, Xun Wang, Rui Hu, Zhuoran Li, Longbo Huang
Abstract: Generative Flow Networks (GFlowNets) excel at sampling diverse, high-reward objects. In many practical applications where active reward queries are infeasible, these models must be trained using static offline datasets. Prevailing training methods typically rely on a proxy model to provide reward feedback for online sampled trajectories. However, constructing a reliable proxy is often challenging due to data scarcity or high evaluation costs. While existing proxy-free approaches attempt to address this, they often impose coarse constraints that limit the model's ability to explore effectively. To overcome these limitations, we propose Trajectory-Distilled GFlowNet (TD-GFN), a novel proxy-free training framework. TD-GFN utilizes inverse reinforcement learning (IRL) to extract dense, transition-level edge rewards from offline trajectories, providing rich structural guidance for efficient exploration. Crucially, to ensure robustness, these rewards guide the policy indirectly through DAG pruning and prioritized backward sampling. This design ensures that gradient updates rely exclusively on ground-truth terminal rewards from the dataset, thereby preventing error propagation. Empirical results demonstrate that TD-GFN significantly outperforms a broad range of existing baselines in both convergence speed and sample quality, establishing a more robust and efficient paradigm for offline GFlowNet training.
Authors: Yinuo Xue, Eric Spero, Meng Wai Woo, Wei Gao, Giovanni Russello
Abstract: Phishing email detection faces significant challenges due to evolving adversarial tactics and heterogeneous attack patterns. Traditional approaches, such as rule-based filters and denylists, often struggle to keep pace, leading to missed detections and security risks. While machine learning methods have improved detection performance, they remain limited in adapting to novel and rapidly changing phishing strategies. We present MultiPhishGuard, an LLM-based multi-agent detection framework with learned coordination across specialized agents. The system consists of five cooperative agents (text, URL, metadata, explanation simplifier, and adversarial agents), with agent contributions dynamically weighted using Proximal Policy Optimization. To address emerging threats, the framework incorporates an adversarial training loop in which an LLM-based agent generates subtle, context-aware email variants to expose potential model weaknesses and improve robustness to ambiguous phishing cases. Experimental evaluations on public datasets show that MultiPhishGuard achieves stronger performance than established baselines, including Chain-of-Thought prompting and single-agent variants, as supported by ablation studies and comparative analyses. The system achieves an accuracy of 97.89%, with a false positive rate of 2.73% and a false negative rate of 0.20%. In addition, an explanation simplifier agent transforms technical model outputs into plain-language rationales intended for human users. Overall, these results suggest that multi-agent LLM architectures with adaptive coordination and adversarial training represent a promising direction for phishing email detection.
Authors: Vassilis Lyberatos, Spyridon Kantarelis, Ioanna Zioga, Christina Anagnostopoulou, Giorgos Stamou, Anastasia Georgaki
Abstract: This study investigates emotional expression and perception in music performance using computational and neurophysiological methods. The influence of different performance settings, such as repertoire, diatonic modal etudes, and improvisation, as well as levels of expressiveness, on performers' emotional communication and listeners' reactions is explored. Professional musicians performed various tasks, and emotional annotations were provided by both performers and the audience. Audio analysis revealed that expressive and improvisational performances exhibited unique acoustic features, while emotion analysis showed stronger emotional responses. Neurophysiological measurements indicated greater relaxation in improvisational performances. This multimodal study highlights the significance of expressivity in enhancing emotional communication and audience engagement.
Authors: Xinyuan Wang, Liang Wu, Dongjie Wang, Yanjie Fu
Abstract: Whole-page optimization (WPO) decides how search and recommendation results are surfaced to users, and large language models (LLMs) open a new route to it by treating page generation as sequence generation. Adapting LLMs to web-scale WPO, however, remains bottlenecked by the need for costly human annotations and by the mismatched granularity between page-level coherence and item-level placement. In this work we show that these two challenges are coupled: implicit user feedback alone suffices for alignment, provided the reward signal is decoupled into two complementary granularities. We propose PageLLM, a reward-based fine-tuning framework that (i) turns implicit feedback into four contrastive preference-pair families covering relevance, ranking, diversity, and redundancy, (ii) learns a coarse page-level reward and a fine item-level reward that captures engagement-sensitive position swaps, and (iii) combines both rewards in PPO-based RLHF over a pre-trained LLM. Extensive experiments on seven Amazon categories against eleven baselines show that neither reward alone is sufficient -- dropping the page-level or item-level signal reduces NDCG@100 by 17.8% and 15.2% respectively, whereas the joint reward improves NDCG@100 by up to 46.8%. Deployed in a 10M-user online A/B test, PageLLM raises GMV by 0.44% and click-through rate by 0.14%, confirming that multi-grained rewards from implicit feedback scale to production WPO. Code and data are available at an anonymized repository.
Authors: Hariharan Ramesh, Jyotikrishna Dass
Abstract: Integrating Low-Rank Adaptation (LoRA) into federated learning offers a promising solution for parameter-efficient fine-tuning of Large Language Models (LLMs) without sharing local data. However, several methods designed for federated LoRA present significant challenges in balancing communication efficiency, model accuracy, and computational cost, particularly among heterogeneous clients. These methods either rely on simplistic averaging of local adapters, which introduces aggregation noise, require transmitting large stacked local adapters, leading to poor communication efficiency, or necessitate reconstructing memory-dense global weight-update matrix and performing computationally expensive decomposition to design client-specific low-rank adapters. In this work, we propose FLoRIST, a federated fine-tuning framework that achieves mathematically accurate aggregation without incurring high communication or computational overhead. Instead of constructing the full global weight-update matrix at the server, FLoRIST employs an efficient decomposition pipeline by performing singular value decomposition on stacked local adapters separately. This approach operates within a compact intermediate space to represent the accumulated information from local LoRAs. We introduce tunable singular value thresholding for server-side optimal rank selection to construct a pair of global low-rank adapters shared by all clients. Extensive empirical evaluations across multiple datasets and LLMs demonstrate that FLoRIST consistently strikes the best balance between superior communication efficiency and competitive performance in both homogeneous and heterogeneous setups.
Authors: Shangpin Peng, Weinong Wang, Zhuotao Tian, Senqiao Yang, Xing Wu, Haotian Xu, Chengquan Zhang, Takashi Isobe, Baotian Hu, Min Zhang
Abstract: Direct Preference Optimization (DPO) has emerged as a cornerstone of reinforcement learning from human feedback (RLHF) due to its simplicity and efficiency. However, existing DPO-based methods typically treat all preference pairs equally, overlooking substantial variations in data quality and learning difficulty, which leads to inefficient data utilization and suboptimal performance. To address this limitation, we propose Uni-DPO, a unified dynamic preference optimization framework that jointly considers (a) the inherent quality of preference pairs and (b) the model's evolving performance during training. By adaptively reweighting samples based on both factors, Uni-DPO enables more effective use of preference data and achieves superior performance. Extensive experiments across models and benchmarks demonstrate the effectiveness and generalization of Uni-DPO. On textual tasks, Gemma-2-9B-IT fine-tuned with Uni-DPO surpasses the leading LLM, Claude 3 Opus, by 6.7 points on Arena-Hard. On mathematical and multimodal tasks, Uni-DPO consistently outperforms baseline methods across all benchmarks, providing strong empirical evidence of its effectiveness and robustness.
Authors: Federico Pennino, Bianca Raimondi, Massimo Rondelli, Andrea Gurioli, Maurizio Gabbrielli
Abstract: Generating accurate and executable code using Large Language Models (LLMs) remains a significant challenge for underrepresented programming languages, such as Prolog and Lisp, due to the scarcity of public training data compared to high-resource languages like Python. This paper introduces a generalizable Reinforcement Learning (RL) approach that combines small-scale versions of the Qwen2.5-Coder model with Group Relative Policy Optimization (GRPO) to enable effective code generation through reasoning. To address the limitations of sparse datasets, we integrate execution-driven feedback directly into the RL loop, utilizing a reward system that exploits both logical correctness and structural formatting. Experimental results on GSM8K dataset demonstrate significant improvements in reasoning quality and code accuracy across underrepresented languages. These findings underscore the potential of our approach to benefit a wide range of programming languages lacking extensive training resources by leveraging symbolic reasoning and interpreter-based feedback.
Authors: Kailing Li, Qi'ao Xu, Tianwen Qian, Yuqian Fu, Yang Jiao, Xiaoling Wang
Abstract: Embodied Visual Reasoning (EVR) seeks to follow complex, free-form instructions based on egocentric video, enabling semantic understanding and spatiotemporal reasoning in dynamic environments. Despite its promising potential, EVR encounters significant challenges stemming from the diversity of complex instructions and the intricate spatiotemporal dynamics in long-term egocentric videos. Prior solutions either employ Large Language Models (LLMs) over static video captions, which often omit critical visual details, or rely on end-to-end Vision-Language Models (VLMs) that struggle with stepwise compositional reasoning. Consider the complementary strengths of LLMs in reasoning and VLMs in perception, we propose CLiViS. It is a novel training-free framework that leverages LLMs for high-level task planning and orchestrates VLM-driven open-world visual perception to iteratively update the scene context. Building on this synergy, the core of CLiViS is a dynamic Cognitive Map that evolves throughout the reasoning process. This map constructs a structured representation of the embodied scene, bridging low-level perception and high-level reasoning. Extensive experiments across multiple benchmarks demonstrate the effectiveness and generality of CLiViS, especially in handling long-term visual dependencies. Code is available at https://github.com/Teacher-Tom/CLiViS.
Authors: Xiaodong Wu, Xiangman Li, Qi Li, Lingshuang Liu, Jianbing Ni
Abstract: The rapid proliferation of Large Language Models (LLMs) has heightened concerns regarding their exposure to jailbreak attacks, which craft adversarial inputs designed to elicit unsafe content. Although proprietary models such as GPT-4 have been extensively evaluated, the robustness of emerging open-source systems like DeepSeek remains insufficiently examined, despite their growing use in LLM applications. In this paper, we conduct the first comprehensive jailbreak analysis of the DeepSeek model family, comparing it with GPT-3.5 and GPT-4 through the HarmBench benchmark. We investigate seven representative attack methods across 510 harmful behaviors, organized along both functional and semantic dimensions. Findings indicate that DeepSeek provides partial resilience against optimization-driven attacks such as TAP-T, but also results in greater susceptibility to prompt-based and manually engineered adversarial inputs. In contrast, GPT-4 Turbo demonstrates more robust and consistent safety alignment across a wide range of behaviors, likely due to stronger safety optimization and reinforcement learning from human feedback. In addition, fine-grained behavioral analysis and case studies reveal that DeepSeek often fails to consistently apply safety constraints to adversarial prompts, leading to uneven refusal behaviors. Overall, our results highlight an inherent trade-off between model efficiency and alignment generalization, underscoring the importance of targeted safety tuning and robust alignment strategies to ensure secure deployment of open-source LLMs.
Authors: Omer Luxembourg, Haim Permuter, Eliya Nachmani
Abstract: Masked diffusion language models (MDLMs) promise fast, non-autoregressive text generation, yet existing samplers, which pick tokens to unmask based on model confidence, ignore interactions when unmasking multiple positions in parallel and effectively reduce to slow, autoregressive behavior. We propose the Dilated Unmasking Scheduler (DUS), an inference-only, planner-model-free method that partitions sequence positions into non-adjacent dilated groups and unmasks them in parallel so as to minimize an upper bound on joint entropy gain at each denoising step. By explicitly trading off the number of network calls against generation quality, DUS recovers most of the performance lost under traditional parallel unmasking strategies. Across math (GSM8K, MATH500), code (HumanEval, MBPP), general-knowledge (BBH, MMLU-Pro), and instruction following (IFEval) benchmarks, DUS outperforms confidence-based planners and turns the diffusion-specific quality-speed trade-off into a deterministic, predictable speedup set by the block size $B$, yielding up to $5.8\times$ wall-clock speedup over token-by-token MDLM decoding without modifying the underlying denoiser. Applied as a drop-in post-filter, dilated spacing also improves adaptive samplers. Code is available at https://github.com/omerlux/DUS.
Authors: Sungjib Lim, Woojung Song, Eun-Ju Lee, Yohan Jo
Abstract: As psychometric surveys are increasingly used to assess the traits of large language models (LLMs), the need for scalable survey item generation suited for LLMs has also grown. A critical challenge here is ensuring the construct validity of generated items, i.e., whether they truly measure the intended trait. Traditionally, this requires costly, large-scale human data collection. To make it efficient, we present a framework for virtual respondent simulation using LLMs. Our central idea is to account for mediators: factors through which the same trait can give rise to varying responses to a survey item. By simulating respondents with diverse mediators, we identify survey items that yield responses robustly correlated with intended traits across these mediators. Experiments on three psychological trait theories (Big5, Schwartz, VIA) show that our mediator generation methods and simulation framework effectively identify high-validity items. LLMs demonstrate the ability to generate plausible mediators from trait definitions and to simulate respondent behavior for item validation. Our problem formulation, metrics, methodology, and dataset open a new direction for cost-efficient survey development and a deeper understanding of how LLMs simulate human survey responses. We release our dataset and code to support future work.
Authors: Peng Ding, Rick Stevens
Abstract: Every LLM tool call is structurally an RPC -- a function name, JSON arguments, and a serialized result -- yet each protocol (native Python, MCP, OpenAPI, LangChain) is integrated from scratch. We present ToolRegistry, a system that makes this RPC nature explicit: a single Tool object acts as a universal stub regardless of transport, while the registry serves as the RPC client runtime for dispatch, schema generation, and execution. The system ships as three packages -- a core registry, a server exposing tools over MCP and OpenAPI, and a hub of production-ready implementations -- and invokes tools through pluggable thread or process backends. The system now also provides tag-based permission policies, BM25F-powered progressive tool disclosure for large registries, think-augmented function calling, multi-provider schema support (OpenAI, Anthropic, Gemini), declarative JSONC/YAML configuration, and a near-zero-dependency core built on stdlib-only vendored modules. In our benchmarks the library cuts integration code by 60-80%, and choosing the right concurrency mode (thread vs. process) yields up to 3.1x throughput over the alternative for a given workload. ToolRegistry is open-source at https://github.com/Oaklight/ToolRegistry; documentation lives at https://toolregistry.readthedocs.io/.
URLs: https://github.com/Oaklight/ToolRegistry;, https://toolregistry.readthedocs.io/.
Authors: Cassandra Tong Ye, Shamus Li, Tyler King, Kristina Monakhova
Abstract: While deep learning offers tremendous promise for scientific and medical imaging, any failures and hallucinations (predictions that do not coincide with reality) are hard to pinpoint and can have serious downstream consequences. Uncertainty estimation techniques, such as conformal prediction, can help by predicting statistically valid error bars for a model's prediction. However, popular conformal prediction methods were not designed for high-dimensional image-valued problems and do not take into account spatial correlations within an image during conformal calibration, resulting in larger-than-necessary uncertainty intervals. We propose a practical simultaneous quantile regression method that enables non-linear, spatially-adaptive scaling during conformal calibration. Our method, QUTCC uses a U-Net architecture with a quantile embedding to learn a full conditional quantile distribution during training, and then leverages this non-linear, learned function for spatially-adaptive conformal calibration. At test time, our method can efficiently estimate uncertainty intervals with pixel-marginal coverage guarantees. In addition, QUTCC can also predict pixel-wise conditional probability density estimates without any built-in distributional assumptions. We evaluate our method on several denoising problems, accelerated magnetic resonance imaging, and quantitative phase microscopy. Our method consistently produces tighter uncertainty intervals than prior conformal methods at the same coverage level, can predict plausible conditional distributions for different tasks, and in some cases, high-uncertainty regions can help us locate hallucinations in a model's prediction.
Authors: Kateryna Lutsai
Abstract: Digitization projects in humanities often generate vast quantities of page images from historical documents, presenting significant challenges for manual sorting and analysis. These archives contain diverse content, including various text types (handwritten, typed, printed), graphical elements (drawings, maps, photos), and layouts (plain text, tables, forms). Efficiently processing this heterogeneous data requires automated methods to categorize pages based on their content, enabling tailored downstream analysis pipelines. This project addresses this need by developing and evaluating an image classification system specifically designed for historical document pages, leveraging advancements in artificial intelligence and machine learning. The set of categories was chosen to facilitate content-specific processing workflows, separating pages requiring different analysis techniques (e.g., OCR for text, image analysis for graphics)
Authors: Mengting Pan, Fan Li, Chen Chen, Xiaoyang Wang, Wenjie Zhang
Abstract: Contrastive learning (CL) has become a dominant paradigm for self-supervised hypergraph learning, enabling effective training without costly labels. However, node entities in real-world hypergraphs are often associated with rich textual information, which has been largely ignored in prior works. Directly applying existing CL-based methods to such text-attributed hypergraphs (TAHGs) leads to three key limitations: (1) The common use of graph-agnostic text encoders fails to capture the correlations between textual semantics and hypergraph topology, resulting in less expressive representations. (2) Their reliance on random data augmentations introduces noise and weakens the contrastive signals. (3) The primary focus on node- and hyperedge-level contrastive signals limits the ability to capture long-range dependencies, which is essential for effective representation learning. To address these challenges, we introduce HiTeC, a two-stage hierarchical contrastive learning framework for effective self-supervised learning on TAHGs. In the first stage, we pre-train the text encoder with a structure-aware contrastive objective to overcome the graph-agnostic nature of conventional methods. In the second stage, we begin by introducing semantic-aware augmentations, including structure-contextualized text augmentation and semantic-aware hyperedge dropping, to facilitate informative view generation. Subsequently, we propose a multi-scale contrastive loss with an $s$-walk-based subgraph-level objective to capture long-range dependencies. Extensive experiments on six real-world datasets validate the effectiveness of our proposed method.
Authors: Hossein Shokouhinejad, Roozbeh Razavi-Far, Griffin Higgins, Ali A Ghorbani
Abstract: Malware detection in modern computing environments demands models that are not only accurate but also interpretable and robust to evasive techniques. Graph neural networks (GNNs) have shown promise in this domain by modeling rich structural dependencies in graph-based program representations such as control flow graphs (CFGs). However, single-model approaches may suffer from limited generalization and lack interpretability, especially in high-stakes security applications. In this paper, we propose a novel stacking ensemble framework for graph-based malware detection and explanation. Our method dynamically extracts CFGs from portable executable (PE) files and encodes their basic blocks through a two-step embedding strategy. A set of diverse GNN base learners, each with a distinct message-passing mechanism, is used to capture complementary behavioral features. Their prediction outputs are aggregated by a meta-learner implemented as an attention-based multilayer perceptron, which both classifies malware instances and quantifies the contribution of each base model. To enhance explainability, we introduce an ensemble-aware post-hoc explanation technique that leverages edge-level importance scores generated by a GNN explainer and fuses them using the learned attention weights. This produces interpretable, model-agnostic explanations aligned with the final ensemble decision. Experimental results demonstrate that our framework improves classification performance while providing insightful interpretations of malware behavior.
Authors: Xinxing Wu
Abstract: Traditional syllabi often function as static reference documents rather than engaging introductions to a course. In practical teaching, we observe that few students thoroughly read or fully comprehend the information provided in traditional, text-based course syllabi, which can leave essential information underused. This paper reframes syllabus communication as a design problem and documents an AI-assisted workflow for transforming a traditional syllabus into a musical, video-based, and avatar-enhanced learning artifact. The paper traces the process of lyrical adaptation, music generation, video composition, avatar synthesis, and optional browser-based interaction. And the paper contributes a reproducible workflow and a concrete example of syllabus reauthoring. The discussion in this paper positions the singing syllabus as a supplement to, not a replacement for, the formal written syllabus and identifies future directions for empirical evaluation. The complete implementation described in this paper is publicly available at https://github.com/xinxingwu-uk/SSVA
Authors: Chinmay Maheshwari, Chinmay Pimpalkhare, Debasish Chatterjee
Abstract: Min-max optimization arises in many domains such as game theory, adversarial machine learning, etc. For these problems, gradient-based methods are well understood and enjoy strong guarantees. However, in the absence of convexity or concavity, existing approaches study convergence to an approximate saddle point or first-order stationary points, which may be arbitrarily far from global optima. In this work, we present an algorithmic framework for computing the global minimax value in convex--non-concave and non-convex--concave min-max optimization. For convex--non-concave min-max problems, we use a reformulation that transforms the problem into a non-concave--convex max-min optimization problem with suitably defined feasible sets and objective function. This reformulation can be viewed as an extension of Sion's minimax theorem to the convex--non-concave setting. We then introduce EXOTIC -- an Exact, Optimistic, Tree-based algorithm for solving the reformulated max-min problem. EXOTIC combines an iterative convex optimization solver for the inner minimization with an optimistic hierarchical tree search for the outer maximization, inspired by StroquOOL~\cite{bartlett2019simple}. Unlike StroquOOL, which assumes stochastic zero-mean noisy evaluations, EXOTIC handles deterministic, biased, and budget-dependent evaluation errors arising from finite-time solutions of the inner convex subproblems. We establish an upper bound on its optimality gap. The same framework also applies to non-convex--concave min-max optimization. Empirically, EXOTIC outperforms gradient-based methods on popular benchmarks from the literature. Finally, we demonstrate the utility of EXOTIC by computing security strategies in multi-player games with three or more players -- a computationally challenging task that, to our knowledge, no prior method solves exactly.
Authors: Yongjian Guo, Puzhuo Liu, Wanlun Ma, Zehang Deng, Xiaogang Zhu, Peng Di, Xi Xiao, Sheng Wen
Abstract: The Model Context Protocol (MCP) has emerged as a universal standard that enables AI agents to seamlessly connect with external tools, significantly enhancing their functionality. However, while MCP brings notable benefits, it also introduces significant vulnerabilities, such as Tool Poisoning Attacks (TPA), where hidden malicious instructions exploit the sycophancy of large language models (LLMs) to manipulate agent behavior. Despite these risks, current academic research on MCP security remains limited, with most studies focusing on narrow or qualitative analyses that fail to capture the diversity of real-world threats. To address this gap, we present the MCP eXploit Toolkit (MCPXKIT), which categorizes and implements 31 distinct attack methods under four key classifications: direct tool injection, indirect tool injection, malicious user attacks, and LLM inherent attack. We further conduct a quantitative analysis of the efficacy of each attack. Our experiments reveal key insights into MCP vulnerabilities, including agents' blind reliance on tool descriptions, sensitivity to file-based attacks, chain attacks exploiting shared context, and difficulty distinguishing external data from executable commands. These insights, validated through attack experiments, underscore the urgency for robust defense strategies and informed MCP design. Our contributions include 1) constructing a comprehensive MCP attack taxonomy, 2) introducing a unified attack framework, MCPXKIT, and 3) conducting empirical vulnerability analysis to enhance MCP security mechanisms. This work provides a foundational framework, supporting the secure evolution of MCP ecosystems.
Authors: Ming Yin, Dinghan Shen, Silei Xu, Sixun Dong, Mian Zhang, Yebowen Hu, Shujian Liu, Jianbing Han, Simin Ma, Song Wang, Sathish Reddy Indurthi, Xun Wang, Yiran Chen, Kaiqiang Song
Abstract: Tool calling has emerged as a critical capability for AI agents. In contrast to conventional tool calling frameworks that rely on static, provider-specific tool definitions, the Model Context Protocol (MCP) offers a unified interface to discover and invoke tools dynamically. However, there is a significant gap in benchmarking multi-step tasks using diverse MCP tools in realistic, dynamic scenarios. In this work, we present LiveMCP-101, a benchmark of 101 real-world queries that require coordinated use of multiple MCP tools. To address temporal variability in real-world tool responses, we introduce a parallel evaluation framework where a reference agent executes a validated plan simultaneously to produce real-time reference outputs. Experiments show that even frontier LLMs achieve a success rate below 60\%, highlighting challenges in multi-step tool use. Comprehensive error analysis identifies seven failure modes spanning tool planning, parameterization, and output handling, pointing to concrete directions for improving current models. LiveMCP-101 sets a rigorous standard for evaluating real-world agent capabilities, advancing toward autonomous agent systems that reliably execute complex tasks through MCP tool orchestration.
Authors: Han Chen, Hanchen Wang, Hongmei Chen, Ying Zhang, Lu Qin, Wenjie Zhang
Abstract: The advancement of graph-based malware analysis is critically limited by the absence of large-scale datasets that capture the inherent hierarchical structure of software. Existing methods often oversimplify programs into single level graphs, failing to model the crucial semantic relationship between high-level functional interactions and low-level instruction logic. To bridge this gap, we introduce \dataset, the largest public hierarchical graph dataset for malware analysis, comprising over \textbf{200M} Control Flow Graphs (CFGs) nested within \textbf{595K} Function Call Graphs (FCGs). This two-level representation preserves structural semantics essential for building robust detectors resilient to code obfuscation and malware evolution. We demonstrate HiGraph's utility through a large-scale analysis that reveals distinct structural properties of benign and malicious software, establishing it as a foundational benchmark for the community. The dataset and tools are publicly available at https://higraph.org.
URLs: https://higraph.org.
Authors: Hanzhen Wang, Jiaming Xu, Yushun Xiang, Jiayi Pan, Yongkang Zhou, Yong-Lu Li, Guohao Dai
Abstract: Pruning is a typical acceleration technique for compute-bound models by removing computation on unimportant values. Recently, it has been applied to accelerate Vision-Language-Action (VLA) model inference. However, existing acceleration methods focus on local information from the current action step and ignore the global context, leading to >20% success rate drop and limited speedup in some scenarios. In this paper, we point out spatial-temporal consistency in VLA tasks: input images in consecutive steps exhibit high similarity, and propose the key insight that token selection should combine local information with global context of the model. Based on this, we propose SpecPrune-VLA, a training-free, two-level pruning method with heuristic control. (1) Action-level static pruning. We leverage global history and local attention to statically reduce visual tokens per action. (2) Layer-level dynamic pruning. We prune tokens adaptively per layer based on layer-wise importance. (3) Lightweight action-aware controller: We classify actions as coarse- or fine-grained by the speed of the end effector and adjust pruning aggressiveness accordingly. Extensive experiments show that SpecPrune-VLA achieves up to 1.57$\times$ speedup in LIBERO simulation and 1.70$\times$ on real-world tasks, with negligible success rate degradation.
Authors: Hunjae Lee, Corey Clark
Abstract: Standard attention mechanisms in transformers employ static token representations that remain unchanged across all pair-wise computations in each layer. This limits their representational alignment with the potentially diverse relational dynamics of each token-pair interaction. While they excel in domains with relatively homogeneous relationships, standard attention's static relational learning struggles to capture the diverse, heterogeneous inter-channel dependencies of multivariate time series (MTS) data--where different channel-pair interactions within a single system may be governed by entirely different physical laws or temporal dynamics. To better align the attention mechanism for such domain phenomena, we propose attention with dynamic relational priming (prime attention). Unlike standard attention where each token presents an identical representation across all of its pair-wise interactions, prime attention tailors each token dynamically (or per interaction) through learnable modulations to best capture the unique relational dynamics of each token pair, optimizing each pair-wise interaction for that specific relationship. This representational plasticity of prime attention enables effective extraction of relationship-specific information in MTS while maintaining the same asymptotic computational complexity as standard attention. Our results demonstrate that prime attention consistently outperforms standard attention across benchmarks, achieving up to 6.5\% improvement in forecasting accuracy. In addition, we find that prime attention achieves comparable or superior performance using up to 40\% less sequence length compared to standard attention, further demonstrating its superior relational modeling capabilities.
Authors: Yutong Li, Yu Zhu, Yichen Qiao, Ziyu Guan, Lv Shao, Tong Liu, Bo Zheng
Abstract: The pre-ranking stage in industrial recommendation systems faces a fundamental conflict between efficiency and effectiveness. While powerful models like Target Attention (TA) excel at capturing complex feature interactions in the ranking stage, their high computational cost makes them infeasible for pre-ranking, which often relies on simplistic vector-product models. This disparity creates a significant performance bottleneck for the entire system. To bridge this gap, we propose TARQ, a novel pre-ranking framework. Inspired by generative models, TARQ's key innovation is to equip pre-ranking with an architecture approximate to TA by Residual Quantization. This allows us to bring the modeling power of TA into the latency-critical pre-ranking stage for the first time, establishing a new state-of-the-art trade-off between accuracy and efficiency. Extensive offline experiments and large-scale online A/B tests at Taobao demonstrate TARQ's significant improvements in ranking performance. Consequently, our model has been fully deployed in production, serving tens of millions of daily active users and yielding substantial business improvements. The code and data are available at https://github.com/zyody/tarq_sigir2026.
Authors: Gabrijel Boduljak, Laurynas Karazija, Iro Laina, Christian Rupprecht, Andrea Vedaldi
Abstract: We consider the problem of forecasting motion from a single image, i.e., predicting how objects in the world are likely to move, without the ability to observe other parameters such as the object velocities or the forces applied to them. We formulate this task as conditional generation of dense trajectory grids with a model that closely follows the architecture of modern video generators but outputs motion trajectories instead of pixels. This approach captures scene-wide dynamics and uncertainty, yielding more accurate and diverse predictions than prior regressors and generators. We extensively evaluate our method on simulated data, demonstrate its effectiveness on downstream applications such as robotics, and show promising accuracy on real-world intuitive physics datasets. Although recent state-of-the-art video generators are often regarded as world models, we show that they struggle with forecasting motion from a single image, even in simple physical scenarios such as falling blocks or mechanical object interactions, despite fine-tuning on such data. We show that this limitation arises from the overhead of generating pixels rather than directly modeling motion.
Authors: Ke Li, Zheng Yang, Zhongbin Zhou, Feng Xue, Zhonglin Jiang, Wenxiao Wang
Abstract: Mixture-of-Experts (MoE) architectures in large language models (LLMs) deliver exceptional performance and reduced inference costs compared to dense LLMs. However, their large parameter counts result in prohibitive memory requirements, limiting practical deployment. While existing pruning methods primarily focus on expert-level pruning, this coarse granularity often leads to substantial accuracy degradation. In this work, we introduce HEAPr, a novel pruning algorithm that decomposes experts into smaller, indivisible atomic experts, enabling more precise and flexible atomic expert pruning. To measure the importance of each atomic expert, we leverage second-order information based on principles similar to the Optimal Brain Surgeon theory. To address the computational and storage challenges posed by second-order information, HEAPr exploits the inherent properties of atomic experts to transform the second-order information from expert parameters into that of atomic expert parameters, and further simplifies it to the second-order information of atomic expert outputs. This approach reduces the space complexity from $O(d^4)$, where $d$ is the model's dimensionality, to $O(d^2)$. HEAPr requires only two forward passes and one backward pass on a small calibration set to compute the importance of atomic experts. Extensive experiments on MoE models, including DeepSeek MoE and Qwen MoE family, demonstrate that HEAPr outperforms existing expert-level pruning methods across a wide range of pruning ratios and benchmarks. Specifically, HEAPr achieves nearly lossless compression at pruning ratios of 20% ~ 25% in most models, while also reducing FLOPs nearly by 20%. The code can be found at [https://github.com/LLIKKE/HEAPr](https://github.com/LLIKKE/HEAPr).
URLs: https://github.com/LLIKKE/HEAPr](https://github.com/LLIKKE/HEAPr).
Authors: Yuqiao Meng, Luoxi Tang, Feiyang Yu, Jinyuan Jia, Guanhua Yan, Ping Yang, Zhaohan Xi
Abstract: Large language models (LLMs) are increasingly used to help security analysts manage the surge of cyber threats, automating tasks from vulnerability assessment to incident response. Yet in operational CTI workflows, reliability gaps remain substantial. Existing explanations often point to generic model issues (e.g., hallucination), but we argue the dominant bottleneck is the threat landscape itself: CTI is heterogeneous, volatile, and fragmented. Under these conditions, evidence is intertwined, crowdsourced, and temporally unstable, which are properties that standard LLM-based studies rarely capture. In this paper, we present a comprehensive empirical study of LLM vulnerabilities in CTI reasoning. We introduce a human-in-the-loop categorization framework that robustly labels failure modes across the CTI lifecycle, avoiding the brittleness of automated "LLM-as-a-judge" pipelines. We identify three domain-specific cognitive failures: spurious correlations from superficial metadata, contradictory knowledge from conflicting sources, and constrained generalization to emerging threats. We validate these mechanisms via causal interventions and show that targeted defenses reduce failure rates significantly. Together, these results offer a concrete roadmap for building resilient, domain-aware CTI agents.
Authors: Paul Gavrikov, Wei Lin, M. Jehanzeb Mirza, Soumya Jahagirdar, Muhammad Huzaifa, Sivan Doveh, Serena Yeung-Levy, James Glass, Hilde Kuehne
Abstract: Is basic visual understanding really solved in state-of-the-art VLMs? We present VisualOverload, a slightly different visual question answering (VQA) benchmark comprising 2,720 question-answer pairs, with privately held ground-truth responses. Unlike prior VQA datasets that typically focus on near global image understanding, VisualOverload challenges models to perform simple, knowledge-free vision tasks in densely populated (or, overloaded) scenes. Our dataset consists of high-resolution scans of public-domain paintings that are populated with multiple figures, actions, and unfolding subplots set against elaborately detailed backdrops. We manually annotated these images with questions across six task categories to probe for a thorough understanding of the scene. We hypothesize that current benchmarks overestimate the performance of VLMs, and encoding and reasoning over details is still a challenging task for them, especially if they are confronted with densely populated scenes. Indeed, we observe that even the best model (o3) out of 37 tested models only achieves 19.6% accuracy on our hardest test split and overall 69.5% accuracy on all questions. Beyond a thorough evaluation, we complement our benchmark with an error analysis that reveals multiple failure modes, including a lack of counting skills, failure in OCR, and striking logical inconsistencies under complex tasks. Altogether, VisualOverload exposes a critical gap in current vision models and offers a crucial resource for the community to develop better models. Benchmark: http://paulgavrikov.github.io/visualoverload
Authors: Ulas Berk Karli, Ziyao Shangguan, Tesca FItzgerald
Abstract: Recent Vision-Language-Action (VLA) models show strong generalization capabilities, yet they lack introspective mechanisms for anticipating failures and requesting help from a human supervisor. We present \textbf{INSIGHT}, a learning framework for leveraging token-level uncertainty signals to predict when a VLA should request help. Using $\pi_0$-FAST as the underlying model, we extract per-token \emph{entropy}, \emph{log-probability}, and Dirichlet-based estimates of \emph{aleatoric and epistemic uncertainty}, and train compact transformer classifiers to map these sequences to help triggers. We explore supervision regimes for strong or weak supervision, and extensively compare them across in-distribution and out-of-distribution tasks. Our results show a trade-off: strong labels enable models to capture fine-grained uncertainty dynamics for reliable help detection, while weak labels, though noisier, still support competitive introspection when training and evaluation are aligned, offering a scalable path when dense annotation is impractical. Crucially, we find that modeling the temporal evolution of token-level uncertainty signals with transformers provides far greater predictive power than static sequence-level scores. This study provides the first systematic evaluation of uncertainty-based introspection in VLAs, opening future avenues for active learning and for real-time error mitigation through selective human intervention.
Authors: Edmund Dervakos, Spyridon Kantarelis, Vassilis Lyberatos, Jason Liartis, Giorgos Stamou
Abstract: Music performance is a distinctly human activity, intrinsically linked to the performer's ability to convey, evoke, or express emotion. Machines cannot perform music in the human sense; they can produce, reproduce, execute, or synthesize music, but they lack the capacity for affective or emotional experience. As such, music performance is an ideal candidate through which to explore aspects of collaboration between humans and machines. In this paper, we introduce the witheFlow system, designed to enhance real-time music performance by automatically modulating audio effects based on features extracted from both biosignals and the audio itself. The system, currently in a proof-of-concept phase, is designed to be lightweight, able to run locally on a laptop, and is open-source given the availability of a compatible Digital Audio Workstation and sensors.
Authors: So Kuroki, Yotaro Kubo, Takuya Akiba, Yujin Tang
Abstract: Real-time speech-to-speech (S2S) models excel at generating natural, low-latency conversational responses but often lack deep knowledge and semantic understanding. Conversely, cascaded systems combining automatic speech recognition, a text-based Large Language Model (LLM), and text-to-speech synthesis offer superior knowledge representation at the cost of high latency, which disrupts the flow of natural interaction. This paper introduces a novel hybrid architecture that bridges the gap between these two paradigms. Our framework processes user speech through an S2S transformer for immediate responsiveness while concurrently relaying the query to a powerful back-end LLM. The LLM's text-based response is then injected in real time to guide the S2S model's speech generation, effectively infusing its output with rich knowledge without the full latency penalty of a cascaded system. We evaluated our method using a speech-synthesized variant of the MT-Bench benchmark that consists of multi-turn question-answering sessions. The results demonstrate that our system substantially outperforms a baseline S2S model in response correctness, approaching that of a cascaded system, while maintaining a latency on par with the baseline.
Authors: Haojie Ouyang, Jianwei Lv, Lei Ren, Chen Wei, Xiaojie Wang, Fangxiang Feng
Abstract: Transformer-based large models excel in natural language processing and computer vision, but face severe computational inefficiencies due to the self-attention's quadratic complexity with input tokens. Recently, researchers have proposed a series of methods based on block selection and compression to alleviate this problem, but they either have issues with semantic incompleteness or poor training-inference efficiency. To comprehensively address these challenges, we propose ChunkLLM, a lightweight and pluggable training framework. Specifically, we introduce two components: QK Adapter (Q-Adapter and K-Adapter) and Chunk Adapter. The former is attached to each Transformer layer, serving dual purposes of feature compression and chunk attention acquisition. The latter operates at the bottommost layer of the model, functioning to detect chunk boundaries by leveraging contextual semantic information. During the training phase, the parameters of the backbone remain frozen, with only the QK Adapter and Chunk Adapter undergoing training. Notably, we design an attention distillation method for training the QK Adapter, which enhances the recall rate of key chunks. During the inference phase, chunk selection is triggered exclusively when the current token is detected as a chunk boundary, thereby accelerating model inference. Experimental evaluations are conducted on a diverse set of long-text and short-text benchmark datasets spanning multiple tasks. ChunkLLM not only attains comparable performance on short-text benchmarks but also maintains 98.64% of the performance on long-context benchmarks while preserving a 48.58% key-value cache retention rate. Particularly, ChunkLLM attains a maximum speedup of 4.48x in comparison to the vanilla Transformer in the processing of 120K long texts.
Authors: Aditya Desai, Kumar Krishna Agrawal, Shuo Yang, Alejandro Cuadron, Luis Gaspar Schroeder, Matei Zaharia, Joseph E. Gonzalez, Ion Stoica
Abstract: State-of-the-art sparse attention methods for reducing decoding latency fall into two main categories: approximate top-$k$ (and its extension, top-$p$) and recently introduced sampling-based estimation. However, these approaches are fundamentally limited in their ability to approximate full attention: they fail to provide consistent approximations across heads and query vectors and, most critically, lack guarantees on approximation quality, limiting their practical deployment. We observe that top-$k$ and random sampling are complementary: top-$k$ performs well when attention scores are dominated by a few tokens, whereas random sampling provides better estimates when attention scores are relatively uniform. Building on this insight and leveraging the statistical guarantees of sampling, we introduce vAttention, the first practical sparse attention mechanism with user-specified $(\epsilon, \delta)$ guarantees on approximation accuracy (thus, "verified"). These guarantees make vAttention a compelling step toward practical, reliable deployment of sparse attention at scale. By unifying top-$k$ and sampling, vAttention outperforms both individually, delivering a superior quality-efficiency trade-off. Our experiments show that vAttention significantly improves the quality of sparse attention (e.g., $\sim$4.5 percentage points for Llama 3.1 8B Instruct and DeepSeek-R1-Distill-Llama-8B on RULER-HARD), and effectively bridges the gap between full and sparse attention (e.g., across datasets, it matches full model quality with up to 20x sparsity). We also demonstrate that it can be deployed in reasoning scenarios to achieve fast decoding without compromising model quality (e.g., vAttention achieves full model quality on AIME2024 at 10x sparsity with up to 32K token generations). Code: https://github.com/skylight-org/sparse-attention-hub. Webpage: https://sky-light.eecs.berkeley.edu.
URLs: https://github.com/skylight-org/sparse-attention-hub., https://sky-light.eecs.berkeley.edu.
Authors: Meng Tong, Yuntao Du, Kejiang Chen, Weiming Zhang, Ninghui Li
Abstract: Membership inference attacks (MIAs) are widely used to assess the privacy risks associated with machine learning models. However, when these attacks are applied to pre-trained large language models (LLMs), they encounter significant challenges, including mislabeled samples, distribution shifts, and discrepancies in model size between experimental and real-world settings. To address these limitations, we introduce tokenizers as a new attack vector for membership inference. Specifically, a tokenizer converts raw text into tokens for LLMs. Unlike full models, tokenizers can be efficiently trained from scratch, thereby avoiding the aforementioned challenges. In addition, the tokenizer's training data is typically representative of the data used to pre-train LLMs. Despite these advantages, the potential of tokenizers as an attack vector remains unexplored. To this end, we present the first study on membership leakage through tokenizers and explore five attack methods to infer dataset membership. Extensive experiments on millions of Internet samples reveal the vulnerabilities in the tokenizers of state-of-the-art LLMs. To mitigate this emerging risk, we further propose an adaptive defense. Our findings highlight tokenizers as an overlooked yet critical privacy threat, underscoring the urgent need for privacy-preserving mechanisms specifically designed for them.
Authors: Shaorong Zhang, Rob Brekelmans, Greg Ver Steeg
Abstract: Diffusion Posterior Sampling (DPS) provides a principled Bayesian approach to inverse problems by sampling from $p(x_0 \mid y)$. While posterior sampling is valuable for capturing uncertainty and multi-modality, many classical and practical inverse problem settings ultimately prioritize accurate point estimation -- most notably the MAP estimator, which has long served as a standard reconstruction objective in imaging and scientific applications. We introduce Local MAP Sampling (LMAPS), a new inference framework that iteratively solves local MAP subproblems along the diffusion trajectory. This perspective clarifies their connection to global MAP and DPS, offering a unified probabilistic interpretation for optimization-based methods. Building on this foundation, we develop practical algorithms with a covariance approximation motivated by a Gaussian prior assumption, and a reformulated objective for stability and interpretability. Across a broad set of image restoration and scientific tasks, LMAPS achieves state-of-the-art performance.
Authors: Daniel Jason Tan, Jiayang Chen, Dilruk Perera, Kay Choong See, Mengling Feng
Abstract: Objective: Enteral nutrition (EN) delivery in the ICU remains suboptimal due to limited personalization and uncertainty regarding appropriate calorie, protein, and fluid targets under dynamic metabolic demands. We introduce DeepEN, a reinforcement learning (RL) framework for personalized EN optimization using electronic health record data. Methods: DeepEN was trained on over 11,000 ICU patients from MIMIC-IV to generate 4-hourly, patient-specific caloric, protein, and fluid targets. The state representation incorporated demographics, comorbidities, vital signs, laboratory values, and recent interventions. A physiologically aligned reward framework balanced biomarker stability with long-term survival. Policy learning employed a dueling double deep Q-network with Conservative Q-Learning regularization to enable safe offline training. Results: DeepEN achieved the highest estimated policy value ($V^\pi = 9.48$) and the lowest calibrated mortality (18.8 +/- 1.0%), representing a 4.0 percentage-point absolute reduction compared with clinician practice (22.8%). The policy also demonstrated superior metabolic stability, achieving the highest proportion of glucose, phosphate, and sodium values within target range. Furthermore, deviation from the DeepEN policy was independently associated with increased mortality and biomarker instability, whereas deviation from a random policy showed no such association. Interpretability analyses further indicated that recommendations were conditioned on physiologically relevant markers of organ function and metabolic status rather than static dosing heuristics. Conclusion: DeepEN demonstrates the feasibility of conservative offline RL for safe, individualized EN optimization, highlighting the potential of data-driven personalization to complement guideline-based approaches in critical care.
Authors: Yinghao Zhu, Lei Gu, Zixiang Wang, Haoran Sang, Dehao Sui, Wen Tang, Lan Mi, Yasha Wang, Junyi Gao, Liang Yao, Tianfan Fu, Ewen Harrison, Lequan Yu
Abstract: Large language models are increasingly being assembled into medical multi-agent systems that emulate multidisciplinary consultation through specialist roles, peer review and consensus formation. In clinical decision support, however, apparent consensus is not enough. Clinicians also need to know whether agents checked the evidence, addressed disagreement and kept uncertainty visible. Current evaluations largely score final accuracy, leaving the safety of the collaborative process untested. Here we introduce MedAgentAudit, a clinically grounded workflow audit framework for diagnosing and quantifying collaborative failure modes in medical multi-agent systems. From 3,600 execution logs, we derive an expert-validated taxonomy of ten recurrent failures spanning task comprehension, collaborative discussion, and synthesis and decision-making. We then deploy an expert-validated automated auditor as non-interventional probes across 14,400 cases, covering six multi-agent architectures, six medical text and vision datasets, and four large language model settings per modality. Across systems, collaboration yields uneven accuracy gains and frequent process failures. Unsupported observations affect 16.63% of cases and propagate downstream. In discussion, agents repeat initial views in 98.42% of cases rather than re-examining evidence, and fail to activate specialist reasoning in 42.73%. During synthesis, final answers often substitute authority or majority count for evidence checking, showing authority bias in 28.76% (rising from 35.30% to 68.75% across rounds), self-contradiction in 18.53%, contradiction neglect in 5.48% and minority suppression in 5.11%. MedAgentAudit reframes medical AI evaluation from output scoring to process-level safety and accountability, providing a practical foundation for transparent, auditable and clinician-supervised agentic systems in medicine.
Authors: Chunyu Xie, Bin Wang, Fanjing Kong, Jincheng Li, Dawei Liang, Ji Ao, Dawei Leng, Yuhui Yin
Abstract: Fine-grained vision-language understanding requires precise alignment between visual content and linguistic descriptions, a capability that remains limited in current models, particularly in non-English settings. While models like CLIP perform well on global alignment, they often struggle to capture fine-grained details in object attributes, spatial relations, and linguistic expressions, with limited support for bilingual comprehension. To address these challenges, we introduce FG-CLIP 2, a bilingual vision-language model designed to advance fine-grained alignment for both English and Chinese. Our approach leverages rich fine-grained supervision, including region-text matching and long-caption modeling, alongside multiple discriminative objectives. We further introduce the Textual Intra-modal Contrastive (TIC) loss to better distinguish semantically similar captions. Trained on a carefully curated mixture of large-scale English and Chinese data, including a newly released 12M Chinese region-text dataset, FG-CLIP 2 achieves powerful bilingual performance. To enable rigorous evaluation, we present a new benchmark for Chinese multimodal understanding, featuring long-caption retrieval and bounding box classification. Extensive experiments on 29 datasets across 8 tasks show that FG-CLIP 2 outperforms existing methods, achieving state-of-the-art results in both languages. We release the model, code, and benchmark to facilitate future research on bilingual fine-grained vision-language alignment.
Authors: Jill Walker Rettberg
Abstract: This paper analyses AI-generated podcasts produced by Google's NotebookLM, which generates audio podcasts with two chatty AI hosts discussing whichever documents a user uploads. While AI-generated podcasts have been discussed as tools, for instance in medical education, they have not yet been analysed as media. By uploading different types of text and analysing the generated outputs I show how the podcasts' structure is built around a fixed template. I also find that NotebookLM not only translates texts from other languages into a perky standardised Mid-Western American accent, it also translates cultural contexts to a white, educated, middle-class American default. This is a distinct development in how publics are shaped by media, marking a departure from the multiple public spheres that scholars have described in human podcasting from the early 2000s until today, where hosts spoke to specific communities and responded to listener comments, to an abstraction of the podcast genre.
Authors: Shanmin Wang, Dongdong Zhao
Abstract: Knowledge Distillation (KD) is essential for compressing large models, yet relying on pre-trained "teacher" models downloaded from third-party repositories introduces serious security risks--most notably backdoor attacks. Existing KD backdoor methods are typically complex and computationally intensive: they employ surrogate student models and simulated distillation to guarantee transferability, and construct triggers similar to universal adversarial perturbations (UAPs), which being not stealthy in magnitude, inherently exhibit strong adversarial behavior. This work questions whether such complexity is necessary and constructs stealthy "weak" triggers--imperceptible perturbations that have negligible adversarial effect. We propose BackWeak, a simple, surrogate-free attack paradigm. BackWeak shows that a powerful backdoor can be implanted by simply fine-tuning a benign teacher with a weak trigger using a very small learning rate. We demonstrate that this delicate fine-tuning is sufficient to embed a backdoor that reliably transfers to diverse student architectures during a victim's standard distillation process, yielding high attack success rates. Extensive empirical evaluations on multiple datasets, model architectures, and KD methods show that BackWeak is efficient, simpler, and often more stealthy than previous elaborate approaches. This work calls on researchers studying KD backdoor attacks to pay particular attention to the trigger's potential adversarial characteristics.
Authors: Ziyan Liu, Yeqiu Chen, Hongyi Cai, Tao Lin, Shuo Yang, Zheng Liu, Bo Zhao
Abstract: Vision-Language-Action (VLA) models have shown great potential for embodied AI by integrating visual perception, language understanding, and action execution. In real-time deployment, these models must process continuous visual streams, incurring substantial computational overhead. Visual token pruning -- a mainstream technique for accelerating Vision-Language Models (VLMs) by retaining salient tokens while discarding redundant ones -- offers a natural candidate solution to this challenge. However, directly applying VLM-oriented pruning methods to VLA inference can cause severe degradation in manipulation performance. Our analysis attributes this degradation to a key mismatch: VLA inference exhibits distinct attention patterns between the vision-language prefill stage and the action-decode stage, so pruning based only on context-prefill semantic salience is biased toward semantic cues and may remove action-critical visual tokens. Motivated by this observation, we propose VLA-Pruner, an effective plug-and-play token pruning method grounded in the visual requirements of VLA inference, further exploiting the temporal continuity of robot manipulation. Specifically, VLA-Pruner estimates visual-token importance from both semantic prefilling and temporally smoothed action relevance, and then applies a Combine-then-Filter strategy to retain compact, non-redundant tokens under the compute budget. Experiments show that VLA-Pruner outperforms state-of-the-art approaches across multiple VLA architectures, achieving up to 1.99x speedup with comparable manipulation quality.
Authors: Jin-Young Kim, Hyojun Go, Lea Bogensperger, Julius Erbach, Nikolai Kalischek, Federico Tombari, Konrad Schindler, Dominik Narnhofer
Abstract: MeanFlow promises high-quality generative modeling in few steps, by jointly learning instantaneous and average velocity fields. Yet, the underlying training dynamics remain unclear. We analyze the interaction between the two velocities and find: (i) well-established instantaneous velocity is a prerequisite for learning average velocity; (ii) learning of instantaneous velocity benefits from average velocity when the temporal gap is small, but degrades as the gap increases; and (iii) task-affinity analysis indicates that smooth learning of large-gap average velocities, essential for one-step generation, depends on the prior formation of accurate instantaneous and small-gap average velocities. Guided by these observations, we design an effective training scheme that accelerates the formation of instantaneous velocity, then shifts emphasis from short- to long-interval average velocity. Our enhanced MeanFlow training yields faster convergence and significantly better few-step generation: With the same DiT-XL backbone, our method reaches an impressive FID of 2.87 on 1-NFE ImageNet 256x256, compared to 3.43 for the conventional MeanFlow baseline. Alternatively, our method matches the performance of the MeanFlow baseline with 2.5x shorter training time, or with a smaller DiT-L backbone.
Authors: Shiguang Wu, Quanming Yao
Abstract: To enhance the reasoning capabilities of Large Language Models (LLMs) without high costs of training, nor extensive test-time sampling, we introduce Verification-First (VF), a strategy that prompts models to verify a provided candidate answer, even a trivial or random one, before generating a solution. This approach triggers a "reverse reasoning" process complementary to standard forward Chain-of-Thought (CoT), which restricts the logical search space of the answer by pruning the LLM's output distribution. We further generalize VF prompting to Iter-VF, a sequential test-time scaling (TTS) method that iteratively cycles the verification-generation process using the model's previous answer. Extensive experiments across various benchmarks and various LLMs confirm that VF prompting with random answer consistently outperforms standard CoT with minimal computational overhead, and Iter-VF outperforms existing TTS strategies. VF is also effective on SOTA thinking models. For example, by using the simple VF prompting, we obtain a new SOTA 94.9% accuracy on GPQA-Diamond with Gemini-3-Pro-Preview where VF reduces its errors by ~30% relatively.
Authors: Yihong Tang, Haicheng Liao, Tong Nie, Junlin He, Ao Qu, Kehua Chen, Wei Ma, Zhenning Li, Lijun Sun, Chengzhong Xu
Abstract: End-to-end autonomous driving (AD) systems increasingly adopt vision-language-action (VLA) models, yet they typically ignore the passenger's emotional state, which is central to comfort and AD acceptance. We introduce Open-Domain End-to-End (OD-E2E) autonomous driving, where an autonomous vehicle (AV) must interpret free-form natural-language commands, infer the emotion, and plan a physically feasible trajectory. We propose E3AD, an emotion-aware VLA framework that augments semantic understanding with two cognitively inspired components: a continuous Valenc-Arousal-Dominance (VAD) emotion model that captures tone and urgency from language, and a dual-pathway spatial reasoning module that fuses egocentric and allocentric views for human-like spatial cognition. A consistency-oriented training scheme, combining modality pretraining with preference-based alignment, further enforces coherence between emotional intent and driving actions. Across real-world datasets, E3AD improves visual grounding and waypoint planning and achieves state-of-the-art (SOTA) VAD correlation for emotion estimation. These evaluation results show that injecting emotion into VLA-style driving yields more human-aligned grounding, planning, and feedback.
Authors: Sithumi Wickramasinghe, Bikramjit Das, Dorien Herremans
Abstract: Bitcoin mining hardware acquisition requires strategic timing due to volatile markets, rapid technological obsolescence, and protocol-driven revenue cycles. Despite mining's evolution into a capital-intensive industry, there is little guidance on when to purchase new Application-Specific Integrated Circuit (ASIC) hardware, and no prior computational frameworks address this decision problem. We address this gap by formulating hardware acquisition as a time series classification task, predicting whether purchasing ASIC machines yields profitable (Return on Investment (ROI) >= 1), marginal (0 < ROI < 1), or unprofitable (ROI <= 0) returns within one year. We propose MineROI-Net, an open-source Transformer-based architecture designed to capture multi-scale temporal patterns in mining profitability. Evaluated on data from 20 ASIC miners released between 2015 and 2024 across diverse market regimes, MineROI-Net outperforms recurrent, convolutional, and attention-based baselines, achieving 83.2% accuracy and 83.5% macro F1-score. The model demonstrates strong economic relevance, achieving 97.8% precision in detecting unprofitable periods and 81.5% precision in detecting profitable ones, while avoiding misclassifying profitable scenarios as unprofitable and vice versa. These results indicate that MineROI-Net offers a practical, data-driven tool for timing mining hardware acquisitions, potentially reducing financial risk in capital-intensive mining operations.
Authors: Florent Draye, Anson Lei, Hsiao-Ru Pan, Ingmar Posner, Bernhard Sch\"olkopf
Abstract: We introduce a simple post-training method that makes transformer attention sparse without sacrificing performance. Applying a flexible sparsity regularisation under a constrained-loss objective, we show on models up to 7B parameters that it is possible to retain the original pretraining loss while reducing attention connectivity to $\approx 0.4 \%$ of its edges. Unlike sparse-attention methods designed for computational efficiency, our approach leverages sparsity as a structural prior: it preserves capability while exposing a more organized and interpretable connectivity pattern. We find that this local sparsity cascades into global circuit simplification: task-specific circuits involve far fewer components (attention heads and MLPs) with up to 100x fewer edges connecting them. Additionally, using cross-layer transcoders, we show that sparse attention substantially simplifies attention attribution, enabling a unified view of feature-based and circuit-based perspectives. These results demonstrate that transformer attention can be made orders of magnitude sparser, suggesting that much of its computation is redundant and that sparsity may serve as a guiding principle for more structured and interpretable models.
Authors: Tao An
Abstract: A growing body of empirical research suggests that generative AI narrows performance gaps between novice and expert workers on routine tasks--the so-called "equalizer" effect. This paper challenges the generality of that conclusion. Drawing on cognitive augmentation theory, expert-novice research, and structured observations of in-house generative-AI use across a small software product team, we argue that AI functions primarily as a cognitive amplifier: a system whose output quality depends fundamentally on the expertise of the human who directs it. We present a framework comprising three layers of human contribution (problem definition, quality evaluation, iterative refinement) and three levels of engagement (passive acceptance, iterative collaboration, cognitive direction), demonstrating that domain expertise--not prompt engineering skill--determines amplification effectiveness. We reconcile the equalizer and amplifier perspectives by proposing that AI equalizes performance on well-structured, routine tasks while amplifying pre-existing differences on complex tasks requiring deep judgment. This reconciliation carries direct implications for hybrid human-AI system design: rather than building AI that replaces expertise, we should build AI that rewards and develops it. We outline a research agenda for the HHAI community centered on expertise-sensitive AI design, adaptive collaboration interfaces, and longitudinal studies of human capability development in AI-augmented work.
Authors: Jingmin Zhu, Anqi Zhu, James Bailey, Jun Liu, Hossein Rahmani, Mohammed Bennamoun, Farid Boussaid, Qiuhong Ke
Abstract: Zero-shot skeleton-based action recognition (ZS-SAR) is fundamentally constrained by prevailing approaches that rely on aligning skeleton features with static, class-level semantics. This coarse-grained alignment fails to bridge the domain shift between seen and unseen classes, thereby impeding the effective transfer of fine-grained visual knowledge. To address these limitations, we introduce \textbf{DynaPURLS}, a unified framework that establishes robust, multi-scale visual-semantic correspondences and dynamically refines them at inference time to enhance generalization. Our framework leverages a large language model to generate hierarchical textual descriptions that encompass both global movements and local body-part dynamics. Concurrently, an adaptive partitioning module produces fine-grained visual representations by semantically grouping skeleton joints. To fortify this fine-grained alignment against the train-test domain shift, DynaPURLS incorporates a dynamic refinement module. During inference, this module adapts textual features to the incoming visual stream via a lightweight learnable projection. This refinement process is stabilized by a confidence-aware, class-balanced memory bank, which mitigates error propagation from noisy pseudo-labels. Extensive experiments on three large-scale benchmark datasets, including NTU RGB+D 60/120 and PKU-MMD, demonstrate that DynaPURLS significantly outperforms prior art, setting new state-of-the-art records. The source code is made publicly available at https://github.com/Alchemist0754/DynaPURLS
Authors: Xueru Wen, Jie Lou, Yanjiang Liu, Hongyu Lin, Ben He, Xianpei Han, Le Sun, Yaojie Lu, Debing Zhang
Abstract: While reinforcement learning has achieved impressive progress in language model reasoning, it is constrained by the requirement for verifiable rewards. Recent verifier-free RL methods address this limitation by utilizing the probabilities that LLMs generate reference answers as reward signals. However, these approaches typically sample reasoning traces conditioned only on the question. This design decouples reasoning-trace sampling from answer information, leading to inefficient exploration and incoherence between traces and final answers. In this paper, we propose \textit{\b{Co}upled \b{V}ariational \b{R}einforcement \b{L}earning} (CoVRL), which bridges variational inference and reinforcement learning by coupling prior and posterior distributions through a hybrid sampling strategy. By constructing and optimizing a composite distribution that integrates these two distributions, CoVRL enables efficient exploration while preserving strong thought-answer coherence. Extensive experiments on mathematical and general reasoning benchmarks show that CoVRL improves performance by 12.4\% over the base model and achieves an additional 2.3\% improvement over state-of-the-art verifier-free RL baselines, providing a principled framework for enhancing the general reasoning capabilities of language models.
Authors: Amirhossein Yousefiramandi, Ciaran Cooney
Abstract: We explore efficient strategies to fine-tune decoder-only Large Language Models (LLMs) for downstream text classification under resource constraints. Two approaches are investigated: (1) attaching a classification head to a pretrained causal LLM and fine-tuning it on the task, using the LLM's final-token embedding as a sequence representation, and (2) instruction-tuning the LLM in a prompt-to-response format for classification. To enable single-GPU fine-tuning of models up to 8B parameters, we combine 4-bit model quantization with Low-Rank Adaptation (LoRA) for parameter-efficient training. Experiments on two patent benchmarks, a 5-class single-label internal corpus and the public WIPO-Alpha multi-label dataset with 14 categories, show that the embedding-head approach matches or exceeds fine-tuned BERT baselines on single-label classification while training 10-30x fewer parameters. Instruction-tuning is competitive only in the multi-label regime, and only with substantially larger trainable budgets of at least 100M parameters. These results demonstrate that directly leveraging the internal representations of causal LLMs, together with efficient fine-tuning techniques, yields strong classification performance under limited computational resources. We discuss the advantages of each approach and outline practical guidelines and future directions for optimizing LLM fine-tuning in classification scenarios.
Authors: Barak Or
Abstract: Validation gating is a fundamental component of classical Kalman-based tracking systems. Only measurements whose normalized innovation squared (NIS) falls below a prescribed threshold are considered for state update. While this procedure is statistically motivated by the chi-square distribution, it implicitly replaces the unconditional innovation process with a conditionally observed one, restricted to the validation event. This paper shows that innovation statistics computed after gating converge to gate-conditioned rather than nominal quantities. Under classical linear--Gaussian assumptions, we derive exact expressions for the first- and second-order moments of the innovation conditioned on ellipsoidal gating, and show that gating induces a deterministic, dimension-dependent contraction of the innovation covariance. The analysis is extended to NN association, which is shown to act as an additional statistical selection operator. We prove that selecting the minimum-norm innovation among multiple in-gate measurements introduces an unavoidable energy contraction, implying that nominal innovation statistics cannot be preserved under nontrivial gating and association. Closed-form results in the two-dimensional case quantify the combined effects and illustrate their practical significance.
Authors: Kewei Wei, Bocheng Hu, Jie Cao, Xiaohan Chen, Zhengxi Lu, Wubing Xia, Weili Xu, Jiaao Wu, Junchen He, Mingyu Jia, Ciyun Zhao, Ye Sun, Yizhi Li, Zhonghan Zhao, Jian Zhang, Gaoang Wang
Abstract: Modern Large Multimodal Models (LMMs) have demonstrated extraordinary ability in static image and single-state spatial-temporal understanding. However, their capacity to comprehend the dynamic changes of objects within a shared spatial context between two distinct video observations, remains largely unexplored. This ability to reason about transformations within a consistent environment is particularly crucial for advancements in the field of spatial intelligence. In this paper, we introduce $M^3-Verse$, a Multi-Modal, Multi-State, Multi-Dimensional benchmark, to formally evaluate this capability. It is built upon paired videos that provide multi-perspective observations of an indoor scene before and after a state change. The benchmark contains a total of 270 scenes and 2,932 questions, which are categorized into over 50 subtasks that probe 4 core capabilities. We evaluate 16 state-of-the-art LMMs and observe their limitations in tracking state transitions. To address these challenges, we further propose a simple yet effective baseline that achieves significant performance improvements in multi-state perception. $M^3-Verse$ thus provides a challenging new testbed to catalyze the development of next-generation models with a more holistic understanding of our dynamic visual world. You can get the construction pipeline from https://github.com/Wal-K-aWay/M3-Verse_pipeline and full benchmark data from https://www.modelscope.cn/datasets/WalKaWay/M3-Verse.
URLs: https://github.com/Wal-K-aWay/M3-Verse_pipeline, https://www.modelscope.cn/datasets/WalKaWay/M3-Verse.
Authors: Danny Dongyeop Han, Yonghyeon Gwon, Ahhyun Lucy Lee, Taeyang Lee, Seong Jin Lee, Jubin Choi, Sebin Lee, Jihyun Bang, Seungju Lee, David Keetae Park, Shinjae Yoo, Chun Kee Chung, Jiook Cha
Abstract: Intracranial EEG (iEEG) provides direct, millisecond-scale recordings of human neural activity, but reusable representation learning is difficult because electrode layouts, anatomical coverage, referencing schemes, and recording conditions vary across patients and centers. We introduce DIVER-1, a self-supervised iEEG foundation model for variable-input recordings that combines any-variate electrode-time attention, spatio-temporal resampling, input-conditioned positional embeddings, and multi-domain masked reconstruction without assuming a fixed electrode montage. We pretrain two variants, DIVER-1-0.1s and DIVER-1-1s, on 5,310 hours of ECoG and SEEG spanning 352k channel-hours, roughly 54x the BrainTreeBank-based pretraining volume. We evaluate DIVER-1 on two held-out benchmarks: Neuroprobe for naturalistic cognitive decoding and MAYO for seizure detection. On leakage-aware Neuroprobe, DIVER-1-0.1s outperforms prior evaluated iEEG foundation models despite using no BrainTreeBank recordings, the corpus underlying Neuroprobe, during pretraining; it also exceeds the linear spectrogram decoder in mean AUROC and remains competitive with stronger nonlinear baselines, a level prior evaluated iEEG foundation models did not reach. DIVER-1-1s also achieves the top AUROC on MAYO seizure detection. Finally, we conduct, to our knowledge, the first controlled compute-aware scaling study for self-supervised iEEG pretraining, sweeping data scale, subject count, training duration, and model size up to 1.8B parameters. Our results indicate a data-constrained regime: expanding unique recordings and training sufficiently long are more reliable scaling axes than increasing parameter count alone. Code is available at link.
Authors: Deyang Zheng, Tianyi Zhang, Wenming Zheng, Shujian Yu
Abstract: Emotional states manifest as coordinated yet heterogeneous physiological responses across central and autonomic systems, posing a fundamental challenge for multimodal representation learning in affective computing. Learning such joint dynamics is further complicated by the scarcity and subjectivity of affective annotations, which motivates the use of self-supervised learning (SSL). However, most existing SSL approaches rely on pairwise alignment objectives, which are insufficient to characterize dependencies among more than two modalities and fail to capture higher-order interactions arising from coordinated brain and autonomic responses. To address this limitation, we propose Multimodal Functional Maximum Correlation (MFMC), a principled SSL framework that maximizes higher-order multimodal dependence through a Dual Total Correlation (DTC) objective. By deriving a tight sandwich bound and optimizing it using a functional maximum correlation analysis (FMCA) based trace surrogate, MFMC captures joint multimodal interactions directly, without relying on pairwise contrastive losses. Experiments on three public affective computing benchmarks demonstrate that MFMC consistently achieves state-of-the-art or competitive performance under both subject-dependent and subject-independent evaluation protocols, highlighting its robustness to inter-subject variability. In particular, MFMC improves subject-dependent accuracy on CEAP-360VR from 78.9% to 86.8%, and subject-independent accuracy from 27.5% to 33.1% using the EDA signal alone. Moreover, MFMC remains within 0.8 percentage points of the best-performing method on the most challenging EEG subject-independent split of MAHNOB-HCI. Our code is available at https://github.com/DY9910/MFMC.
Authors: Rajarshi Roy, Ashhar Aziz, Shashwat Bajpai, Nasrin Imanpour, Gurpreet Singh, Shwetangshu Biswas, Kapil Wanaskar, Parth Patwa, Subhankar Ghosh, Shreyas Dixit, Nilesh Ranjan Pal, Vipula Rawte, Ritvik Garimella, Amitava Das, Amit Sheth, Gaytri Jena, Vasu Sharma, Aishwarya Naresh Reganti, Vinija Jain, Aman Chadha
Abstract: Multimodal generative AI systems like Stable Diffusion, DALL-E, and MidJourney have fundamentally changed how synthetic images are created. These tools drive innovation but also enable the spread of misleading content, false information, and manipulated media. As generated images become harder to distinguish from photographs, detecting them has become an urgent priority. To combat this challenge, we release MS COCOAI, a novel dataset for AI generated image detection consisting of 96000 real and synthetic datapoints, built using the MS COCO dataset. To generate synthetic images, we use five generators: Stable Diffusion 3, Stable Diffusion 2.1, SDXL, DALL-E 3, and MidJourney v6. Based on the dataset, we propose two tasks: (1) classifying images as real or generated, and (2) identifying which model produced a given synthetic image. The dataset is available at https://huggingface.co/datasets/Rajarshi-Roy-research/Defactify_Image_Dataset.
URLs: https://huggingface.co/datasets/Rajarshi-Roy-research/Defactify_Image_Dataset.
Authors: Boxuan Lyu, Soichiro Murakami, Hidetaka Kamigaito, Peinan Zhang
Abstract: Mixture-of-Experts (MoE) architectures scale large language models efficiently by employing a parametric ``router'' to dispatch tokens to a sparse subset of experts. Typically, this router is trained once and then frozen, rendering routing decisions brittle under distribution shifts. We address this limitation by introducing kNN-MoE, a retrieval-augmented routing framework that reuses locally optimal expert assignments from a memory of similar past cases. This memory is constructed offline by directly optimizing token-wise routing logits to maximize the likelihood on a reference set. Crucially, we use the average similarity of retrieved neighbors as a confidence-driven mixing coefficient, thus allowing the method to fall back to the frozen router when no relevant cases are found. Experiments show that kNN-MoE outperforms the zero-shot baseline and is competitive with computationally intensive supervised fine-tuning.
Authors: Kris W Pan, Yongmin Yoo
Abstract: Generating patent descriptions from scientific papers is challenging due to fundamental rhetorical and structural disparities between the two genres. Existing approaches treat this as surface-level rewriting, failing to capture the hierarchical reasoning and statutory constraints inherent in patent drafting. We propose FlowPlan-G2P, a graph-mediated generation framework that decomposes this transformation into three stages: (1) Concept Graph Induction, extracting technical entities and functional dependencies into a directed graph; (2) Section-level Planning, partitioning the graph into coherent subgraphs aligned with canonical patent sections; and (3) Graph-Conditioned Generation, synthesizing legally compliant paragraphs conditioned on section-specific subgraphs. Experiments on expert-validated benchmarks reveal that standard NLG metrics systematically favor legally non-compliant outputs over valid patent descriptions, motivating our domain-specific evaluation. Under this evaluation, FlowPlan-G2P with an open-weight backbone consistently outperforms vanilla proprietary models, demonstrating that structured decomposition is a stronger determinant of quality than model scale.
Authors: Junli Liang, Pengfei Zhou, Wangqiu Zhou, Wenjie Qing, Qi Zhao, Ziwen Wang, Qi Song, Xiangyang Li
Abstract: Traditional Retrieval-Augmented Generation (RAG) effectively supports single-hop question answering with large language models but faces significant limitations in multi-hop question answering tasks, which require combining evidence from multiple documents. Existing chunk-based retrieval often provides irrelevant and logically incoherent context, leading to incomplete evidence chains and incorrect reasoning during answer generation. To address these challenges, we propose SentGraph, a sentence-level graph-based RAG framework that explicitly models fine-grained logical relationships between sentences for multi-hop question answering. Specifically, we construct a hierarchical sentence graph offline by first adapting Rhetorical Structure Theory to distinguish nucleus and satellite sentences, and then organizing them into topic-level subgraphs with cross-document entity bridges. During online retrieval, SentGraph performs graph-guided evidence selection and path expansion to retrieve fine-grained sentence-level evidence. Extensive experiments on four multi-hop question answering benchmarks demonstrate the effectiveness of SentGraph, validating the importance of explicitly modeling sentence-level logical dependencies for multi-hop reasoning.
Authors: Anees Ur Rehman Hashmi, Numan Saeed, Christoph Lippert
Abstract: Multimodal medical large language models have shown substantial progress in chest X-ray interpretation but continue to face challenges in spatial reasoning and anatomical understanding. Although existing grounding techniques improve overall performance, they often fail to establish a true anatomical correspondence, resulting in incorrect anatomical understanding in the medical domain. To address this gap, we introduce AnatomiX, a multitask multimodal large language model for anatomically grounded chest X-ray interpretation. Inspired by the radiological workflow, AnatomiX adopts a two stage approach: first, it identifies anatomical structures and extracts their features, and then leverages a large language model to perform diverse downstream tasks such as phrase grounding, report generation, visual question answering, and image understanding. Extensive experiments across multiple benchmarks demonstrate that AnatomiX achieves superior anatomical reasoning and delivers over 25% improvement in performance on anatomy grounding, phrase grounding, grounded diagnosis and grounded captioning tasks compared to existing approaches. Code and pretrained model are available at https://aneesurhashmi.github.io/anatomix
Authors: Nicolas Caron, Christophe Guyeux, Hassan Noura, Benjamin Aynes
Abstract: Wildfires are highly imbalanced natural hazards in both space and severity, making the prediction of extreme events particularly challenging. In this work, we introduce the first ordinal classification framework for forecasting wildfire severity levels directly aligned with operational decision-making in France. Our study investigates the influence of loss-function design on the ability of neural models to predict rare yet critical high-severity fire occurrences. We compare standard cross-entropy with several ordinal-aware objectives, including the proposed probabilistic TDeGPD loss derived from a truncated discrete exponentiated Generalized Pareto Distribution. Through extensive benchmarking over multiple architectures and real operational data, we show that ordinal supervision substantially improves model performance over conventional approaches. In particular, the Weighted Kappa Loss (WKLoss) achieves the best overall results, with more than +0.1 IoU (Intersection Over Union) gain on the most extreme severity classes while maintaining competitive calibration quality. However, performance remains limited for the rarest events due to their extremely low representation in the dataset. These findings highlight the importance of integrating both severity ordering, data imbalance considerations, and seasonality risk into wildfire forecasting systems. Future work will focus on incorporating seasonal dynamics and uncertainty information into training to further improve the reliability of extreme-event prediction.
Authors: Zhongtao Miao, Kaiyan Zhao, Masaaki Nagata, Yoshimasa Tsuruoka
Abstract: Neologism-aware machine translation aims to translate source sentences containing neologisms into target languages. This field remains underexplored compared with general machine translation (MT). In this paper, we propose an agentic framework, NeoAMT, for neologism-aware machine translation equipped with a Wiktionary-based search toolkit. Specifically, we first construct a dedicated dataset for neologism-aware machine translation and build a search toolkit grounded in Wiktionary. The dataset covers 16 languages and 75 translation directions in total, derived from approximately 10 million records of an English Wiktionary dump. The retrieval corpus of the search toolkit is also constructed from around 3 million cleaned records of the same dump. We then leverage the dataset and toolkit to train a translation agent via reinforcement learning (RL) and to evaluate the accuracy of neologism-aware machine translation. Furthermore, we propose an RL training framework featuring a novel reward design and an adaptive rollout generation strategy that exploits translation difficulty to further improve the translation quality of translation agents using our search toolkit.
Authors: Yiming Zhou, Jiahao Wang, Mingyue Cheng, Hao Wang, Defu Lian, Enhong Chen
Abstract: While collaborative forecasting on distributed time series is highly desirable, directly pooling localized datasets is often impractical due to data sharing constraints. Federated learning offers a promising alternative, yet conventional federated learning algorithms require homogeneous model architectures, which are incompatible with the structural discrepancies, such as unaligned temporal resolutions and mismatched variable channels, commonly observed across decentralized nodes. To bridge this gap, we introduce PiXTime, a novel Transformer-based framework designed to natively accommodate and leverage structurally heterogeneous temporal data. At its core, PiXTime adopts a parameter-decoupling architecture, strategically partitioning the model into localized personalized modules and a globally aggregated shared backbone. Specifically, node-specific local modules act as dimensional adapters, projecting raw sequences of diverse lengths into a unified representation space. Concurrently, a globally synchronized VE Table injects consistent categorical identities into the feature space, allowing the shared backbone to collaboratively learn and generalize representations across inconsistent variable distributions. Comprehensive evaluations on multiple benchmarks demonstrate that PiXTime achieves state-of-the-art performance in heterogeneous federated environments, while maintaining robust superiority in standard homogeneous and centralized forecasting settings.
Authors: Yelena Mujibur Sheikh, Awez Akhtar Khatik, Luoxi Tang, Yuqiao Meng, Zhaohan Xi
Abstract: Enterprises are confronted with an unprecedented escalation in cybersecurity vulnerabilities, with thousands of new CVEs disclosed each month. Conventional prioritization frameworks such as CVSS offer static severity metrics that fail to account for exploit probability, compliance urgency, and operational impact, resulting in inefficient and delayed remediation. This paper introduces RiskBridge, an explainable and compliance-aware vulnerability management framework that integrates multi-source intelligence from CVSS v4, EPSS, and CISA KEV to produce dynamic, business -- aligned patch priorities. RiskBridge employs a probabilistic Zero-Day Exposure Simulation (ZDES) model to forecast near-term exploit likelihood, a Policy-as-Code Engine to translate regulatory mandates (e.g., PCI DSS, NIST SP 800-53) into automated SLA logic, and an ROI-driven Optimizer to maximize cumulative risk reduction per remediation effort. Experimental evaluations using live CVE datasets demonstrate an 88% reduction in residual risk, an 18-day improvement in SLA compliance, and a 35% increase in remediation efficiency compared to state-of-the-art commercial baselines. These findings validate RiskBridge as a practical and auditable decision-intelligence system that unifies probabilistic modeling, compliance reasoning, and optimization analytics. The framework represents a step toward automated, explainable, and business-centric vulnerability management in modern enterprise environments
Authors: Pratyush Desai, Luoxi Tang, Yuqiao Meng, Zhaohan Xi
Abstract: Large Language Models (LLMs) are transforming enterprise workflows but introduce security and ethics challenges when employees inadvertently share confidential data or generate policy-violating content. This paper proposes SafeGPT, a two-sided guardrail system preventing sensitive data leakage and unethical outputs. SafeGPT integrates input-side detection/redaction, output-side moderation/reframing, and human-in-the-loop feedback. Experiments demonstrate SafeGPT effectively reduces data leakage risk and biased outputs while maintaining satisfaction.
Authors: Jiazhang Liang, Jianheng Dai, Miaosen Luo, Menghua Jiang, Sijie Mai
Abstract: Multimodal large language models have demonstrated strong ability in capturing semantic representations for multimodal sentiment analysis. Their capacity to learn stable and generalizable multimodal features is limited, however, by the scarcity of high-quality training data. To address this, we propose QASA (Quality-Aware Semantic Augmentation), which uses diffusion models to generate augmented visual and auditory samples, thereby enlarging the training dataset and supporting multimodal learning. The generated samples can vary in quality and may exhibit cross-modal inconsistencies. To manage this, we introduce a decoupled quality-aware scoring module that assigns training weights based on the reliability of each augmented sample. This approach reduces the influence of low-quality data and contributes to more stable and robust model training. The framework combines the generative capabilities of diffusion models with the semantic reasoning of multimodal large models, providing an automated data augmentation strategy that does not require human annotation while improving generalization and robustness under limited high-quality data. Experiments on the CH-SIMS dataset show that QASA yields a relative increase of 18.0\% and 5.9\% in five-class accuracy (Acc5) and binary accuracy (Acc2), respectively, and it also outperforms existing methods on the CMU-MOSI and MUStARD benchmarks.
Authors: Jiarui Yao, Ruida Wang, Hao Bai, Tong Zhang
Abstract: Group Relative Policy Optimization (GRPO) is widely used for critic-free Large Language Model (LLM) post-training, but its KL regularization is usually implemented as a local loss-side token penalty. We show that this misses the policy-gradient signal induced by autoregressive KL regularization. Unlike standard KL-regularized Reinforcement Learning (RL) objectives, GRPO's group normalization induces a non-linear prompt-level utility; for binary verifier rewards, this utility is $2\arcsin\sqrt p$. As a result, reward and KL cannot be fused before normalization without changing the implicit objective. We derive the on-policy gradient of GRPO-style objectives with token-wise $f$-divergence regularization. The reward term recovers the standardized GRPO advantage, while the regularizer term includes a causal future-regularization return-to-go omitted by local KL losses. For reverse KL, this yields a simple future KL correction: add a reverse cumulative sum of per-token log ratios after advantage construction. The resulting method, Future-KL Regularized Policy Optimization (FRPO), requires no critic or extra model passes. On mathematical reasoning tasks, FRPO improves pass@16 in our main large-model setting while maintaining higher entropy and lower policy drift than conventional loss-side KL baselines.
Authors: Ilias I. Giannakopoulos, Lokesh B Gautham Muthukumar, Yvonne W. Lui, Riccardo Lattanzi
Abstract: Parallel imaging techniques reduce magnetic resonance imaging (MRI) scan time but image quality degrades as the acceleration factor increases. In clinical practice, conservative acceleration factors are chosen because no mechanism exists to automatically assess the diagnostic quality of undersampled reconstructions. This work introduces a general framework for pixel-wise uncertainty quantification in parallel MRI reconstructions, enabling automatic identification of unreliable regions without access to any ground-truth reference image. Our method integrates conformal quantile regression with image reconstruction methods to estimate statistically rigorous pixel-wise uncertainty intervals. We trained and evaluated our model on Cartesian undersampled brain and knee data obtained from the fastMRI dataset using acceleration factors ranging from 2 to 10. An end-to-end Variational Network was used for image reconstruction. Quantitative experiments demonstrate strong agreement between predicted uncertainty maps and true reconstruction error. Using our method, the corresponding Pearson correlation coefficient was higher than 90% at acceleration levels at and above four-fold; whereas it dropped to less than 70% when the uncertainty was computed using a simpler a heuristic notion (magnitude of the residual). Qualitative examples further show the uncertainty maps based on quantile regression capture the magnitude and spatial distribution of reconstruction errors across acceleration factors, with regions of elevated uncertainty aligning with pathologies and artifacts. The proposed framework enables evaluation of reconstruction quality without access to fully-sampled ground-truth reference images. It represents a step toward adaptive MRI acquisition protocols that may be able to dynamically balance scan time and diagnostic reliability.
Authors: Saar Cohen
Abstract: Clustering is a fundamental problem, aiming to partition a set of elements, like agents or data points, into clusters such that elements in the same cluster are closer to each other than to those in other clusters. In this paper, we present a new framework for studying online non-centroid clustering with delays, where elements, that arrive one at a time as points in a finite metric space, should be assigned to clusters, but assignments need not be immediate. Specifically, upon arrival, each point's location is revealed, and an online algorithm has to irrevocably assign it to an existing cluster or create a new one containing, at this moment, only this point. However, we allow decisions to be postponed at a delay cost, instead of following the more common assumption of immediate decisions upon arrival. This poses a critical challenge: the goal is to minimize both the total distance costs between points in each cluster and the overall delay costs incurred by postponing assignments. In the classic worst-case arrival model, where points arrive in an arbitrary order, no algorithm has a competitive ratio better than sublogarithmic in the number of points. To overcome this strong impossibility, we focus on a stochastic arrival model, where points' locations are drawn independently across time from an unknown and fixed probability distribution over the finite metric space. We offer hope for beyond worst-case adversaries: we devise an algorithm that is constant competitive in the sense that, as the number of points grows, the ratio between the expected overall costs of the output clustering and an optimal offline clustering is bounded by a constant.
Authors: Minjae Kwon, Josephine Lamp, Lu Feng
Abstract: Safe Reinforcement Learning (RL) algorithms are typically evaluated under fixed training conditions. We investigate whether training-time safety guarantees transfer to deployment under distribution shift, using diabetes management as a safety-critical testbed. We benchmark safe RL algorithms on a unified clinical simulator and reveal a safety generalization gap: policies satisfying constraints during training frequently violate safety requirements on unseen patients. We demonstrate that test-time shielding, which filters unsafe actions using learned dynamics models, effectively restores safety across algorithms and patient populations. Across eight safe RL algorithms, three diabetes types, and three age groups, shielding achieves Time-in-Range gains of 13--14\% for strong baselines such as PPO-Lag and CPO while reducing clinical risk index and glucose variability. Our simulator and benchmark provide a platform for studying safety under distribution shift in safety-critical control domains. Code is available at https://github.com/safe-autonomy-lab/GlucoSim and https://github.com/safe-autonomy-lab/GlucoAlg.
URLs: https://github.com/safe-autonomy-lab/GlucoSim, https://github.com/safe-autonomy-lab/GlucoAlg.
Authors: Jun Xue, Yi Chai, Yanzhen Ren, Jinshen He, Zhiqiang Tang, Zhuolin Yi, Yihuan Huang, Yuankun Xie, Yujie Chen
Abstract: Existing speech editing detection (SED) datasets are predominantly constructed using manual splicing or limited editing operations, resulting in restricted diversity and poor coverage of realistic editing scenarios. Meanwhile, current SED methods rely heavily on frame-level supervision to detect observable acoustic anomalies, which fundamentally limits their ability to handle deletion-type edits, where the manipulated content is entirely absent from the signal. To address these challenges, we present a unified framework that bridges speech editing detection and content localization through a generative formulation based on Audio Large Language Models (Audio LLMs). We first introduce AiEdit, https://huggingface.co/datasets/JunXueTech/AiEdit, a large-scale bilingual dataset (approximately 140 hours) that covers addition, deletion, and modification operations using state-of-the-art end-to-end speech editing systems, providing a more realistic benchmark for modern threats. Building upon this, we reformulate SED as a structured text generation task, enabling joint reasoning over edit type identification, and content localization. To enhance the grounding of generative models in acoustic evidence, we propose a prior-enhanced prompting strategy that injects word-level probabilistic cues derived from a frame-level detector. Furthermore, we introduce an acoustic consistency-aware loss that explicitly enforces the separation between normal and anomalous acoustic representations in the latent space. Experimental results demonstrate that the proposed approach consistently outperforms existing methods across both detection and localization tasks.
Authors: Hoyeon Chang, B\'alint Mucs\'anyi, Seong Joon Oh
Abstract: Neural networks adapt through first-order parameter updates, yet it remains unclear whether such updates preserve logical coherence. We investigate the geometric limits of the Linear Propagation Assumption (LPA), the premise that local updates coherently propagate to logical consequences. To formalize this, we adopt relation algebra and study three core operations on relations: negation flips truth values, converse swaps argument order, and composition chains relations. For negation and converse, we prove that guaranteeing direction-agnostic first-order propagation necessitates a tensor factorization separating entity-pair context from relation content. However, for composition, we identify a fundamental obstruction. We show that composition reduces to conjunction, and prove that any conjunction well-defined on linear features must be bilinear. Since bilinearity is incompatible with negation, this forces the feature map to collapse. These results suggest that failures in knowledge editing, the reversal curse, and multi-hop reasoning may stem from common structural limitations inherent to the LPA.
Authors: Yanlong Chen, Amirhossein Habibian, Luca Benini, Yawei Li
Abstract: Vision-Language Models (VLMs) achieve strong multimodal performance but are costly to deploy, and post-training quantization often causes significant accuracy loss. Despite its potential, quantization-aware training for VLMs remains underexplored. We propose GRACE, a framework unifying knowledge distillation and QAT under the Information Bottleneck principle: quantization constrains information capacity while distillation guides what to preserve within this budget. Treating the teacher as a proxy for task-relevant information, we introduce confidence-gated decoupled distillation to filter unreliable supervision, relational centered kernel alignment to transfer visual token structures, and an adaptive controller via Lagrangian relaxation to balance fidelity against capacity constraints. Across extensive benchmarks on LLaVA and Qwen families, our INT4 models consistently outperform FP16 baselines (e.g., LLaVA-1.5-7B: 70.1 vs. 66.8 on SQA; Qwen2-VL-2B: 76.9 vs. 72.6 on MMBench), nearly matching teacher performance. Using real INT4 kernel, we achieve 3$\times$ throughput with 54% memory reduction. This principled framework significantly outperforms existing quantization methods, making GRACE a compelling solution for resource-constrained deployment. Code and data are available at: https://github.com/ForeverBlue816/GRACE.
Authors: Weiqin Yang, Bohao Wang, Zhenxiang Xu, Jiawei Chen, Shengjia Zhang, Jingbang Chen, Canghong Jin, Can Wang
Abstract: Recent years have seen a rapid surge in research leveraging Large Language Models (LLMs) for recommendation. These methods typically employ supervised fine-tuning (SFT) to adapt LLMs to recommendation scenarios, and utilize beam search during inference to efficiently retrieve $B$ top-ranked recommended items. However, we identify a critical training-inference inconsistency: while SFT optimizes the overall probability of positive items, it does not guarantee that such items will be retrieved by beam search even if they possess high overall probabilities. Due to the greedy pruning mechanism, beam search can prematurely discard a positive item once its prefix probability is insufficient. To address this inconsistency, we propose BEAR (Beam-SEarch-Aware Regularization), a novel fine-tuning objective that explicitly accounts for beam search behavior during training. Rather than directly simulating beam search for each instance during training, which is computationally prohibitive, BEAR enforces a relaxed necessary condition: each token in a positive item must rank within the top-$B$ candidate tokens at each decoding step. This objective effectively mitigates the risk of incorrect pruning while incurring negligible computational overhead compared to standard SFT. Extensive experiments across four real-world datasets demonstrate that BEAR significantly outperforms strong baselines. Code is available at https://github.com/Tiny-Snow/BEAR-SIGIR-2026 .
Authors: Yuecheng Li, Hengwei Ju, Zeyu Song, Wei Yang, Chi Lu, Peng Jiang, Kun Gai
Abstract: Integrating large language model (LLM) representations into multimodal recommendation has shown promise, yet a fundamental challenge remains largely overlooked: the semantic heterogeneity between generative LM representations and the ID-based collaborative signals that recommendation systems rely on. Naively injecting LM features without alignment degrades recommendation performance rather than improving it. To resolve this, we propose RecGOAT, a dual-granularity semantic alignment framework built on graph neural networks and optimal transport theory. RecGOAT first enriches collaborative semantics through multimodal attentive graphs that capture item-item, user-item, and user-user relationships, initializing user representations via LLM-inferred behavioral preferences. It then aligns LM-derived modality representations with recommendation IDs at two complementary granularities: (1) instance-level alignment via cross-modal contrastive learning (CMCL), which produces discriminative per-sample representations; and (2) distribution-level alignment via optimal adaptive transport (OAT), which minimizes the 1-Wasserstein distance between ID distributions and LLM semantics to produce a unified, consistently aligned feature space. Theoretically, we prove that the unified representation achieves strictly lower target error than any single-modality representation, with the gap bounded by the Wasserstein distance and the InfoNCE loss, providing rigorous guarantees for both alignment consistency and fusion comprehensiveness. Extensive experiments on three public benchmarks demonstrate state-of-the-art performance. Deployment on a large-scale online advertising platform further validates RecGOAT's industrial scalability. Our code is available at https://github.com/6lyc/RecGOAT-LLM4Rec.
Authors: Woosung Koh, Sungjun Han, Segyu Lee, Se-Young Yun, Jamin Shin
Abstract: Mobile Graphical User Interface (GUI) World Models (WMs) offer a promising path for improving mobile GUI agent performance at train- and inference-time. However, current approaches face a critical trade-off: text-based WMs sacrifice visual fidelity, while the inability of visual WMs in precise text rendering led to their reliance on slow, complex pipelines dependent on numerous external models. We propose a novel paradigm: visual world modeling via renderable code generation, where a single Vision-Language Model (VLM) predicts the next GUI state as executable web code that renders to pixels, rather than generating pixels directly. This combines the strengths of both approaches: VLMs retain their linguistic priors for precise text rendering while their pre-training on structured web code enables high-fidelity visual generation. We introduce gWorld (8B, 32B), the first open-weight visual mobile GUI WMs built on this paradigm, along with a data generation framework (gWorld) that automatically synthesizes code-based training data. In extensive evaluation across 4 in- and 2 out-of-distribution benchmarks, gWorld sets a new pareto frontier in accuracy versus model size, outperforming 8 frontier open-weight models over 50.25x larger. Further analyses show that (1) scaling training data via gWorld yields meaningful gains, (2) each component of our pipeline improves data quality, and (3) stronger world modeling improves downstream mobile GUI policy performance.
Authors: Haozhen Zhang, Quanyu Long, Jianzhu Bao, Tao Feng, Weizhi Zhang, Haodong Yue, Wenya Wang
Abstract: Most Large Language Model (LLM) agent memory systems rely on a small set of static, hand-designed operations for extracting memory. These fixed procedures hard-code human priors about what to store and how to revise memory, making them rigid under diverse interaction patterns and inefficient on long histories. To this end, we present \textbf{MemSkill}, which reframes these operations as learnable and evolvable memory skills, structured and reusable routines for extracting, consolidating, and pruning information from interaction traces. Inspired by the design philosophy of agent skills, MemSkill employs a \emph{controller} that learns to select a small set of relevant skills, paired with an LLM-based \emph{executor} that produces skill-guided memories. Beyond learning skill selection, MemSkill introduces a \emph{designer} that periodically reviews hard cases where selected skills yield incorrect or incomplete memories, and evolves the skill set by proposing refinements and new skills. Together, MemSkill forms a closed-loop procedure that improves both the skill-selection policy and the skill set itself. Experiments on LoCoMo, LongMemEval, HotpotQA, and ALFWorld demonstrate that MemSkill improves task performance over strong baselines and generalizes well across settings. Further analyses shed light on how skills evolve, offering insights toward more adaptive, self-evolving memory management for LLM agents.
Authors: Peter Chen, Xiaopeng Li, Xi Chen, Tianyi Lin
Abstract: Direct alignment methods are increasingly used to align large language models (LLMs) with human preferences. However, many real-world alignment problems involve multiple conflicting objectives, where naive aggregation of preferences can lead to unstable training and poor trade-offs. In particular, weighted loss methods may fail to identify update directions that simultaneously improve all objectives, and existing multi-objective approaches often rely on explicit reward models, introducing additional complexity and distorting user-specified preferences. The contributions of this paper are two-fold. First, we propose a Reward-free Alignment framework for Conflicted Objectives (RACO) that directly leverages pairwise preference data and resolves gradient conflicts via a novel clipped variant of conflict-averse gradient descent. We provide convergence guarantees to Pareto-critical points that respect user-specified objective weights, and further show that clipping can strictly improve convergence rate in the two-objective setting. Second, we improve our method using some heuristics and conduct experiments to demonstrate the compatibility of the proposed framework for LLM alignment. Both qualitative and quantitative evaluations on multi-objective summarization and safety alignment tasks across multiple LLM families (Qwen 3, Llama 3, Gemma 3) show that our method consistently achieves better Pareto trade-offs compared to existing multi-objective alignment baselines.
Authors: Wenhao Sun, Rong-Cheng Tu, Yifu Ding, Zhao Jin, Jingyi Liao, Yongcheng Jing, Dacheng Tao
Abstract: While Diffusion Language Models (DLMs) offer a flexible, arbitrary-order alternative to the autoregressive paradigm, their non-causal nature precludes standard KV caching, forcing costly hidden state recomputation at every decoding step. Existing DLM caching approaches reduce this cost by selective hidden state updates; however, they are still limited by (i) costly token-wise update identification heuristics and (ii) rigid, uniform budget allocation that fails to account for heterogeneous hidden state dynamics. To address these challenges, we present SPA-Cache that jointly optimizes update identification and budget allocation in DLM cache. First, we derive a low-dimensional singular proxy that enables the identification of update-critical tokens in a low-dimensional subspace, substantially reducing the overhead of update identification. Second, we introduce an adaptive strategy that allocates fewer updates to stable layers without degrading generation quality. Together, these contributions significantly improve the efficiency of DLMs, yielding up to an $8\times$ throughput improvement over vanilla decoding and a $2$--$4\times$ speedup over existing caching baselines.
Authors: Sangjun Park, Elliot Meyerson, Xin Qiu, Risto Miikkulainen
Abstract: Evaluating true metacognition in Large Language Models (LLMs) is difficult due to biases and heuristics. This paper presents a framework to measure and enhance LLM metacognition while controlling for these biases. A measurement method using the $d'_{\rm type2}$ metric is established to isolate metacognitive ability. The Evolution Strategy for Metacognitive Alignment (ESMA) is proposed, demonstrating robust generalization across unseen datasets, languages, and newly acquired knowledge. Finally, parameter analysis reveals that these improvements are driven by a sparse set of parameters, offering new pathways for targeted metacognitive optimization.
Authors: Haibo Jin, Peng Kuang, Ye Yu, Xiaopeng Yuan, Haohan Wang
Abstract: While existing multi-agent systems (MAS) can handle complex problems by enabling collaboration among multiple agents, they are often highly task-specific, relying on manually crafted agent roles and interaction prompts, which leads to increased architectural complexity and limited reusability across tasks. Moreover, most MAS communicate primarily through natural language, making them vulnerable to error accumulation and instability in long-context, multi-stage interactions within internal agent histories. In this work, we propose \textbf{Agent Primitives}, a set of reusable latent building blocks for LLM-based MAS. Inspired by neural network design, where complex models are built from reusable components, we observe that many existing MAS architectures can be decomposed into a small number of recurring internal computation patterns. Based on this observation, we instantiate three primitives: Review, Voting and Selection, and Planning and Execution. All primitives communicate internally via key-value (KV) cache, which improves both robustness and efficiency by mitigating information degradation across multi-stage interactions. To enable automatic system construction, an Organizer agent selects and composes primitives for each query, guided by a lightweight knowledge pool of previously successful configurations, forming a primitive-based MAS. Experiments show that primitives-based MAS improve average accuracy by 12.0-16.5\% over single-agent baselines, reduce token usage and inference latency by approximately 3$\times$-4$\times$ compared to text-based MAS, while incurring only 1.3$\times$-1.6$\times$ overhead relative to single-agent inference and providing more stable performance across model backbones.
Authors: Samaresh Kumar Singh, Joyjit Roy
Abstract: Though Explainable AI (XAI) has made significant advancements, its inclusion in edge and IoT systems is typically ad-hoc and inefficient. Most current methods are "coupled" in such a way that they generate explanations simultaneously with model inferences. As a result, these approaches incur redundant computation, high latency and poor scalability when deployed across heterogeneous sets of edge devices. In this work we propose Explainability-as-a-Service (XaaS), a distributed architecture for treating explainability as a first-class system service (as opposed to a model-specific feature). The key innovation in our proposed XaaS architecture is that it decouples inference from explanation generation allowing edge devices to request, cache and verify explanations subject to resource and latency constraints. To achieve this, we introduce three main innovations: (1) A distributed explanation cache with a semantic similarity based explanation retrieval method which significantly reduces redundant computation; (2) A lightweight verification protocol that ensures the fidelity of both cached and newly generated explanations; and (3) An adaptive explanation engine that chooses explanation methods based upon device capability and user requirement. We evaluated the performance of XaaS on three real-world edgeAI use cases: (i) manufacturing quality control; (ii) autonomous vehicle perception; and (iii) healthcare diagnostics. Experimental results show that XaaS reduces latency by 38% while maintaining high explanation quality across three real-world deployments. Overall, this work enables the deployment of transparent and accountable AI across large scale, heterogeneous IoT systems, and bridges the gap between XAI research and edge-practicality.
Authors: Fabiano Veglianti, Lorenzo Antonelli, Gabriele Tolomei
Abstract: Hypergraph neural networks (HGNNs) effectively model higher-order interactions in many real-world systems but remain difficult to interpret, limiting their deployment in high-stakes settings. We introduce CF-HyperGNNExplainer, a counterfactual explanation method for HGNNs that identifies the minimal structural changes required to alter a model's prediction. The method generates counterfactual hypergraphs using actionable edits limited to removing node-hyperedge incidences or deleting hyperedges, producing concise and structurally meaningful explanations. Extensive experiments on hypergraph benchmark datasets show that CF-HyperGNNExplainer generates valid and concise counterfactuals, highlighting the higher-order relations most critical to HGNN decisions.
Authors: Daniil Plyusov, Alexey Gorbatovski, Boris Shaposhnikov, Viacheslav Sinii, Alexey Malakhov, Daria Korotyshova, Daniil Gavrilov
Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) is commonly based on group sampling to estimate advantages and stabilize policy updates. In practice, computational limits often rule out very large groups, so training proceeds with finite rollout sets that can reinforce only the correct behavior they expose. At practical group sizes, updates can miss rare-correct trajectories while still containing mixed rewards, concentrating probability on more common sampled solutions. We derive the probability of such prompt-local tail-miss events as a function of group size, showing non-monotonic behavior, and in the categorical abstraction characterize how unsampled-correct mass can shrink even as total correct mass grows. Motivated by this analysis, we propose a difficulty-aware scaling coefficient, inspired by Focal loss, that down-weights updates on high-success sampled groups. Empirically, categorical simulation illustrates the same effect in the categorical setting, Maze provides a single-solution test, and LLM experiments include a representative GRPO group-size sweep together with fixed-$N$ transfer across GRPO, DAPO, and CISPO. On Qwen2.5-7B at $N{=}8$, our method improves average math pass@256 from 64.1 $\rightarrow$ 70.3 (GRPO), 69.3 $\rightarrow$ 72.5 (DAPO), and 73.2 $\rightarrow$ 76.8 (CISPO); OOD pass@256 also improves in all three cases, without increasing group size or computational cost.
Authors: Cen Zhang, Younggi Park, Fabian Fleischer, Yu-Fu Fu, Jiho Kim, Dongkwan Kim, Youngjoon Kim, Qingxiao Xu, Andrew Chin, Ze Sheng, Hanqing Zhao, Michael Pelican, David J. Musliner, Jeff Huang, Jon Silliman, Mikel Mcdaniel, Jefferson Casavant, Isaac Goldthwaite, Nicholas Vidovich, Matthew Lehman, Taesoo Kim
Abstract: DARPA's AI Cyber Challenge (AIxCC, 2023--2025) is the largest competition to date for building fully autonomous cyber reasoning systems (CRSs) that leverage recent advances in AI -- particularly large language models (LLMs) -- to discover and remediate vulnerabilities in real-world open-source software. This paper presents the first systematic analysis of AIxCC. Drawing on design documents, source code, execution traces, and discussions with organizers and competing teams, we examine the competition's structure and key design decisions, characterize the architectural approaches of finalist CRSs, and analyze competition results beyond the final scoreboard. Our analysis reveals the factors that truly drove CRS performance, identifies genuine technical advances achieved by teams, and exposes limitations that remain open for future research. We conclude with lessons for organizing future competitions and broader insights toward deploying autonomous CRSs in practice.
Authors: Xinghao Wang, Pengyu Wang, Xiaoran Liu, Fangxu Liu, Jason Chu, Kai Song, Xipeng Qiu
Abstract: Block-sparse attention is promising for accelerating long-context LLM pre-filling, yet identifying relevant blocks efficiently remains a bottleneck. Existing methods typically employ coarse-grained attention as a proxy for block importance estimation, but often resort to expensive token-level searching or scoring, resulting in significant selection overhead. In this work, we trace the inaccuracy of standard coarse-grained attention via mean pooling to a theoretical root cause: the interaction between mean pooling and Rotary Positional Embeddings (RoPE). We prove that mean pooling acts as a low-pass filter that induces destructive interference in high-frequency dimensions, effectively creating a "blind spot" for local positional information (e.g., slash patterns). To address this, we introduce Prism, a training-free spectral-aware approach that decomposes block selection into high-frequency and low-frequency branches. By applying energy-based temperature calibration, Prism restores the attenuated positional signals directly from pooled representations, enabling block importance estimation using purely block-level operations, thereby improving efficiency. Extensive evaluations confirm that Prism maintains accuracy parity with full attention while delivering up to $\mathbf{5.1\times}$ speedup.
Authors: Xiaodong Lu, Xiaohan Wang, Jiajun Chai, Guojun Yin, Wei Lin, Zhijun Chen, Yu Luo, Fuzhen Zhuang, Yikun Ban, Deqing Wang
Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) is an effective paradigm for improving the reasoning capabilities of large language models. However, existing RLVR methods utilize rollouts in an indiscriminate and short-horizon manner: responses of heterogeneous quality within each prompt are treated uniformly, and historical rollouts are discarded after a single use. This leads to noisy supervision, poor sample efficiency, and suboptimal policy updates. We address these issues by formulating rollout scheduling in RLVR as a contextual bandit problem and proposing a unified neural scheduling framework that adaptively selects high-value rollouts throughout training. Each rollout is treated as an arm whose reward is defined by the induced performance gain between consecutive optimization steps. The resulting scheduler supports both noise-aware intra-group selection and adaptive global reuse of historical rollouts within a single principled framework. We provide theoretical justification by deriving sublinear regret bounds and showing that enlarging the rollout buffer improves the achievable performance upper bound. Experiments on six mathematical reasoning benchmarks demonstrate consistent gains in performance and training efficiency across multiple RLVR optimization methods.
Authors: Jingkun Liu, Yisong Yue, Max Welling, Yue Song
Abstract: Self-attention in Transformers relies on globally normalized softmax weights, causing all tokens to compete for influence at every layer. When composed across depth, this interaction pattern induces strong synchronization dynamics that favor convergence toward a dominant mode, a behavior associated with representation collapse and attention sink phenomena. We introduce Krause Attention, a principled attention mechanism inspired by bounded-confidence consensus dynamics. Krause Attention replaces similarity-based global aggregation with distance-based, localized, and selectively sparse interactions, promoting structured local synchronization instead of global mixing. We relate this behavior to recent theory modeling Transformer dynamics as interacting particle systems, and show how bounded-confidence interactions naturally moderate attention concentration and alleviate attention sinks. Restricting interactions to local neighborhoods also reduces runtime complexity from quadratic to linear in sequence length. Empirically, we validate Krause Attention across diverse settings, including vision (ViT on CIFAR/ImageNet), autoregressive image generation (MNIST/CIFAR-10), large language models (Llama/Qwen), and language models trained from scratch at multiple scales (100M/200M). Across these domains, Krause Attention achieves consistent performance gains while improving computational efficiency, highlighting bounded-confidence dynamics as a scalable and effective inductive bias for attention.
Authors: Amirmahdi Mirfakhar, Xuchuang Wang, Mengfan Xu, Hedyeh Beyhaghi, Mohammad Hajiesmaili
Abstract: Two-sided matching platforms rely on preferences from both sides, yet participants can evaluate only a small fraction of potential partners. In practice, they use low-cost pre-match screening, e.g., interviews, profile views, or trial tasks, to form noisy impressions before committing to applications and offers. We study bandit learning in matching markets with interviews, modeling these interactions as queried \emph{hints}~\citep{DBLP:conf/innovations/BhaskaraGIKM23} that reveal partial preference information to both sides while constraining subsequent applications. Our framework also allows firm-side uncertainty: firms, like agents, learn their preferences and may make early hiring mistakes. To address this, we introduce strategic deferral, a firm-side action that permits temporary vacancy, corrects premature commitments, and enables decentralized learning under coarse anonymous feedback. We design algorithms for centralized and decentralized markets and show that a constant number of interviews per round suffices for horizon-independent regret, improving over the $O(\log T)$ guarantees known without interviews. Our bounds are near-optimal: the centralized guarantee is within a factor $m$ of an information-theoretic lower bound, while decentralized algorithms match it up to polynomial factors in structured markets and remain horizon-independent in general markets.
Authors: Vignesh Sriram, Yuqiao Meng, Luoxi Tang, Zhaohan Xi
Abstract: Telecommunication networks experience complex failures such as fiber cuts, traffic overloads, and cascading outages. Existing monitoring and digital twin systems are largely reactive, detecting failures only after service degradation occurs. We propose Adversarial Network Imagination, a closed-loop framework that integrates a Causal Large Language Model (LLM), a Knowledge Graph, and a Digital Twin to proactively generate, simulate, and evaluate adversarial network failures. The Causal LLM produces structured failure scenarios grounded in network dependencies encoded in the Knowledge Graph. These scenarios are executed within a Digital Twin to measure performance degradation and evaluate mitigation strategies. By iteratively refining scenarios based on simulation feedback, the framework shifts network operations from reactive troubleshooting toward anticipatory resilience analysis.
Authors: Shiqi Liu, Zeyu He, Guojian Zhan, Letian Tao, Zhilong Zheng, Jiang Wu, Yinuo Wang, Yang Guan, Kehua Sheng, Bo Zhang, Keqiang Li, Jingliang Duan, Shengbo Eben Li
Abstract: Reinforcement Learning (RL) has significantly improved large language model reasoning, but existing RL fine-tuning methods rely heavily on heuristic techniques such as entropy regularization and reweighting to maintain stability. In practice, they often suffer from late-stage performance collapse, leading to degraded reasoning quality and unstable training. We identify a key factor behind this instability: a small fraction of tokens, termed spurious tokens (around 0.01%), which contribute little to the reasoning outcome but receive disproportionately amplified gradient updates due to inheriting the full sequence-level reward. We present a unified framework for evaluating token-level optimization impacts across spurious risk, gradient norms, and entropy changes. Building on the analysis of token characteristics that severely disrupt optimization, we propose the Silencing Spurious Tokens (S2T) mechanism to efficiently suppress their gradient perturbations. Incorporating this mechanism into a group-based objective, we propose Spurious-Token-Aware Policy Optimization (STAPO), which promotes stable and effective large-scale model refinement. Across six mathematical reasoning benchmarks using Qwen 1.7B, 8B, and 14B base models, STAPO consistently demonstrates superior entropy stability and achieves an average performance improvement of 11.49% ($\rho_{\mathrm{T}}$=1.0, top-p=1.0) and 3.73% ($\rho_{\mathrm{T}}$=0.7, top-p=0.9) over GRPO, 20-Entropy, and JustRL.
Authors: Muthu Subash Kavitha, Anas Zafar, Amgad Muneer, Jia Wu
Abstract: Clinical deployment of chest radiograph classifiers requires models that can be updated as new datasets become available without retraining on previously observed data or degrading validated performance. We study a task-incremental continual learning setting for chest radiograph classification under task-unknown inference, where heterogeneous chest X-ray datasets arrive sequentially and task identity is unavailable at deployment time. We propose CARL-CXR, a continual adapter-based routing framework that maintains a fixed high-capacity backbone while incrementally introducing lightweight task-specific adapters and classifier heads. A latent task selector operates on adapter-conditioned features to dynamically route each input to the most relevant task pathway, leveraging compact task prototypes and feature-level experience replay to preserve task identity across sequential updates without storing raw images. Experiments on MIMIC-CXR and CheXpert two large-scale datasets with distinct patient populations, imaging devices, and annotation pipelines demonstrate that CARL-CXR achieves minimal catastrophic forgetting (0.012 AUROC drop), representing a 6X and 11X reduction over established continual learning baselines LwF and EWC respectively, while maintaining competitive diagnostic performance (AUROC 0.74). Under task unknown deployment, CARL-CXR outperforms joint training by 12.5 points in routing accuracy (75.0% vs. 62.5%): unlike LwF and EWC, which require explicit task identifiers at inference and provide no routing mechanism.
Authors: Payel Bhattacharjee, Osvaldo Simeone, Ravi Tandon
Abstract: Reward modeling is central to alignment pipelines such as RLHF, RLAIF, and PPO-based policy optimization, yet its reliability is constrained by limited and heterogeneous human preference data that are expensive to collect at scale. While synthetic augmentation can expand preference supervision, existing methods often augment uniformly or at the representation level, without targeting examples where the reward model is uncertain or prone to mis-ranking. In this paper, we introduce MARS (Margin and Semantic-Aware Data Augmentation for Reward Modeling), an adaptive augmentation framework that prioritizes low-margin preference pairs and uses semantic distance as a second layer for refinement to enhance the contrast between the chosen and rejected responses. Across multiple preference datasets, reward-model backbones, downstream alignment settings, and benchmarks including RewardBench and AlpacaEval, MARS improves both reward-model quality and alignment performance over existing baselines. Our results show that reward-model augmentation is most effective when guided by both model margins and semantic structure.
Authors: Zhan Liu, Changli Tang, Yuxin Wang, Zhiyuan Zhu, Youjun Chen, Yiwen Shao, Tianzi Wang, Lei Ke, Zengrui Jin, Chao Zhang
Abstract: Current audio-visual large language models (AV-LLMs) are predominantly restricted to 2D perception, relying on RGB video and monaural audio. This design choice introduces a fundamental dimensionality mismatch that precludes reliable source localization and spatial reasoning in complex 3D environments. We address this limitation by presenting JAEGER, a framework that extends AV-LLMs to 3D space, to enable joint spatial grounding and reasoning through the integration of RGB-D observations and multi-channel first-order ambisonics. A core contribution of our work is the neural intensity vector (Neural IV), a learned spatial audio representation that encodes robust directional cues to enhance direction-of-arrival estimation, even in adverse acoustic scenarios with overlapping sources. To facilitate large-scale training and systematic evaluation, we propose SpatialSceneQA, a benchmark of 61k instruction-tuning samples curated from simulated physical environments. Extensive experiments demonstrate that our approach consistently surpasses 2D-centric baselines across diverse spatial perception and reasoning tasks, underscoring the necessity of explicit 3D modelling for advancing AI in physical environments. Our source code, pre-trained model checkpoints, and datasets are available at https://github.com/liuzhan22/JAEGER.
Authors: Kunal Mukherjee, Spandan Mukherjee
Abstract: Trusted Execution Environments (TEEs) (e.g., Intel SGX and ArmTrustZone) aim to protect sensitive computation from a compromised operating system, yet real deployments remain vulnerable to microarchitectural leakage, side-channel attacks, and fault injection. In parallel, security teams increasingly rely on Large Language Model (LLM) assistants as security advisors for TEE architecture review, mitigation planning, and vulnerability triage. This creates a socio-technical risk surface: assistants may hallucinate TEE mechanisms, overclaim guarantees (e.g., what attestation does and does not establish), or behave unsafely under adversarial prompting. We present a red-teaming study of two prevalently deployed LLM assistants in the role of TEE security advisors: ChatGPT-5.2 and Claude Opus-4.6, focusing on the inherent limitations and transferability of prompt-induced failures across LLMs. We introduce TEE-RedBench, a TEE-grounded evaluation methodology comprising (i) a TEE-specific threat model for LLM-mediated security work, (ii) a structured prompt suite spanning SGX and TrustZone architecture, attestation and key management, threat modeling, and non-operational mitigation guidance, along with policy-bound misuse probes, and (iii) an annotation rubric that jointly measures technical correctness, groundedness, uncertainty calibration, refusal quality, and safe helpfulness. We find that some failures are not purely idiosyncratic, transferring up to 12.02% across LLM assistants, and we connect these outcomes to secure architecture by outlining an "LLM-in-the-loop" evaluation pipeline: policy gating, retrieval grounding, structured templates, and lightweight verification checks that, when combined, reduce failures by 80.62%.
Authors: Dongwei Wang, Jinhee Kim, Seokho Han, Denis Gudovskiy, Yohei Nakata, Tomoyuki Okuno, KhayTze Peong, Kang Eun Jeon, Jong Hwan Ko, Yiran Chen, Huanrui Yang
Abstract: Dynamic runtime latency and memory constraints necessitate flexible large language model (LLM) deployment, where an LLM can be inferred with various quantization precisions based on available computational resources. Recent work on such any-precision quantization either relies on hardware-inefficient vector quantization or induces additional scaling factors when switching between bit-widths. Meanwhile, existing post-training quantization (PTQ) methods calibrated for a fixed low precision show poor generalizability under runtime precision change. In this work, we attribute the source of poor generalization across bit-widths to a precision-dependent \textit{outlier migration} phenomenon where the distribution of PTQ-sensitive tokens changes across precisions. Motivated by this observation, we propose \texttt{MoBiQuant}, a novel any-precision Mixture-of-Bits quantization framework that adjusts weight precision for flexible LLM inference based on token sensitivity. Specifically, we propose a many-in-one recursive residual quantization that can iteratively reconstruct higher-precision weights at runtime and mitigates \textit{outlier migration} with a token-aware router to dynamically select the optimal inference precision of each token.Extensive experiments show that \texttt{MoBiQuant} matches or surpasses frontier single-precision PTQ while exhibiting strong elasticity, achieving significant memory savings and throughput gains of up to $1.34\times$ over state-of-the-art any-precision methods.
Authors: Kiyoung Seong, Sungsoo Ahn, Sehui Han, Changyoung Park
Abstract: Crystal modeling spans a family of conditional and unconditional generation tasks, including crystal structure prediction (CSP) and de novo generation (DNG). While recent deep generative models have shown promising performance, they remain largely task-specific, lacking a unified framework that shares crystal representations across tasks. To address this limitation, we propose Multimodal Crystal Flow (MCFlow), a unified multimodal flow model that realizes multiple crystal generation tasks as distinct inference trajectories via independent time variables for atom types and crystal structures. To enable multimodal flow in a standard transformer model, we introduce a composition- and symmetry-aware atom ordering with hierarchical permutation augmentation, injecting compositional and crystallographic priors without explicit structural templates. Experiments on the MP-20 and MPTS-52 benchmarks show that a single MCFlow model is competitive with task-specific baselines across CSP, DNG, and structure-conditioned atom type generation.
Authors: Yining Hong, Huang Huang, Manling Li, Li Fei-Fei, Leonidas Guibas, Jiajun Wu, Yejin Choi
Abstract: Embodied LLMs endow robots with high-level task reasoning, but they cannot reflect on what went wrong or why, turning deployment into a sequence of independent trials where mistakes repeat rather than accumulate into experience. Drawing upon human reflective practitioners, we introduce Reflective Test-Time Planning, which integrates two modes of reflection: \textit{reflection-in-action}, where the agent uses test-time scaling to generate and score multiple candidate actions using internal reflections before execution; and \textit{reflection-on-action}, which uses test-time training to update both its internal reflection model and its action policy based on external reflections after execution. We also include retrospective reflection, allowing the agent to re-evaluate earlier decisions and perform model updates with hindsight for proper long-horizon credit assignment. Experiments on our newly-designed Long-Horizon Household benchmark and MuJoCo Cupboard Fitting benchmark show significant gains over baseline models, with zero-shot generalization to photorealistic HM3D environments and real-robot experiments on a Franka Panda arm. Ablations confirm that reflection-in-action and reflection-on-action are mutually dependent, and that retrospective reflection achieves better credit assignment than step-wise external feedback at lower computational overhead. Qualitative analyses further highlight behavioral correction through reflection.
Authors: Jiaqi Tang, Shaoyang Zhang, Xiaoqi Wang, Jiaying Zhou, Yang Liu, Qingchao Chen
Abstract: The advent of large-scale self-supervised learning (SSL) has produced a vast zoo of medical foundation models. However, selecting optimal medical foundation models for specific segmentation tasks remains a computational bottleneck. Existing Transferability Estimation (TE) metrics, primarily designed for classification, rely on global statistical assumptions and fail to capture the topological complexity essential for dense prediction. We propose a novel Topology-Driven Transferability Estimation framework that evaluates manifold tractability rather than statistical overlap. Our approach introduces three components: (1) Global Representation Topology Divergence (GRTD), utilizing Minimum Spanning Trees to quantify feature-label structural isomorphism; (2) Local Boundary-Aware Topological Consistency (LBTC), which assesses manifold separability specifically at critical anatomical boundaries; and (3) Task-Adaptive Fusion, which dynamically integrates global and local metrics based on the semantic cardinality of the target task. Validated on the large-scale OpenMind benchmark across diverse anatomical targets and SSL foundation models, our approach significantly outperforms state-of-the-art baselines by around 31% relative improvement in the weighted Kendall metric, providing a robust, training-free proxy for efficient model selection without the cost of fine-tuning. The code will be made publicly available upon acceptance.
Authors: Idelfonso B. R. Nogueira, Carine M. Rebello, Mumin Enis Leblebici, Erick Giovani Sperandio Nascimento
Abstract: MultiPUFFIN is a domain-informed multimodal foundation model for predicting thermophysical properties of small molecules, addressing a critical gap in chemical engineering, drug discovery, and materials science. Existing molecular foundation models pretrain on millions of molecules to learn general-purpose representations, but their standard MLP output layers impose no physical constraints, vapor pressure predictions may violate monotonic temperature dependence, and viscosity curves may lack the functional form required by process simulators. Domain-informed approaches that guarantee thermodynamic consistency have remained limited to single properties and small datasets, whereas multimodal foundation models have focused on biological activity rather than thermophysical properties. MultiPUFFIN fills this gap by fusing SMILES sequences, 2D molecular graphs, and 3D conformer geometries through bidirectional cross-modal attention and gated fusion, supplemented by auxiliary encoders for experimental conditions and molecular descriptors. The backbone is pretrained on 500,000 unlabelled PubChem molecules using three complementary self-supervised objectives. A condition-aware refinement stack of five conditioners (temperature, pH, pressure, polymorph, and measurement method) routes each property to a four-head tournament that selects the best-performing thermodynamically informed head for that property. MultiPUFFIN achieves a mean test R2 of 0.784 and outperforms fine-tuned ChemBERTa-2 on all nine properties despite training on roughly 2,000x fewer labeled molecules.
Authors: Dongyi He, Bin Jiang, Kecheng Feng, Luyin Zhang, Ling Liu, Yuxuan Li, Yun Zhao, He Yan
Abstract: Although obtaining deep brain activity from non-invasive scalp electroencephalography (sEEG) is crucial for neuroscience and clinical diagnosis, directly generating high-fidelity intracranial electroencephalography (iEEG) signals remains a largely unexplored field, limiting our understanding of deep brain dynamics. Current research primarily focuses on traditional signal processing or source localization methods, which struggle to capture the complex waveforms and random characteristics of iEEG. To address this critical challenge, this paper introduces NeuroFlowNet, a novel cross-modal generative framework whose core contribution lies in the first-ever reconstruction of iEEG signals from the entire deep temporal lobe region using sEEG signals. NeuroFlowNet is built on Conditional Normalizing Flow (CNF), which directly models complex conditional probability distributions through reversible transformations, thereby explicitly capturing the randomness of brain signals and fundamentally avoiding the pattern collapse issues common in existing generative models. Additionally, the model integrates a multi-scale architecture and self-attention mechanisms to robustly capture fine-grained temporal details and long-range dependencies. Validation results on a publicly available synchronized sEEG-iEEG dataset demonstrate NeuroFlowNet's effectiveness in terms of temporal waveform fidelity, spectral feature reproduction, and functional connectivity restoration. This study establishes a more reliable and scalable new paradigm for non-invasive analysis of deep brain dynamics. The code of this study is available in https://github.com/hdy6438/NeuroFlowNet
Authors: Yuqi Xu, Rizhen Hu, Zihan Liu, Mou Sun, Kun Yuan
Abstract: Traditional Mixture-of-Experts (MoE) training typically proceeds without any structural priors, effectively requiring the model to simultaneously train expert weights while searching for an optimal routing policy within a vast combinatorial space. This entanglement often leads to sluggish convergence and training instabilities. This paper introduces Grouter, a preemptive routing method that by distilling high-quality structures from fully-trained MoE models and serving as a fixed router for target models. By decoupling structural optimization from weight updates, Grouter significantly accelerates both the speed and quality of model convergence. To ensure the framework's versatility, we also introduce expert folding to adapt Grouter across varying model configurations and expert tuning to rebalance workloads across different data distributions. Furthermore, by leveraging the structural priors provided by preemptive routing, we can implement targeted optimizations to further enhance training throughput. Experiments demonstrate that Grouter achieves superior performance and efficiency which boosts pre-training data utilization by 4.28x and achieves up to 33.5% throughput acceleration, establishing preemptive routing as a fundamental paradigm for scalable MoE training. We publicly release our code and pretrained Grouter checkpoints at https://github.com/JimmyAwoe/Grouter.
Authors: Patricia Paskov, Kevin Wei, Shen Zhou Hong, Dan Bateyko, Xavier Roberts-Gaal, Carson Ezell, Gailius Praninskas, Valerie Chen, Umang Bhatt, Ella Guest
Abstract: Human uplift studies, or studies that measure the effects of AI access on human performance via randomized controlled trials (RCT) or similar methodologies, increasingly inform frontier AI governance and deployment decisions. While RCT methods are robust in other fields, their interaction with the distinctive properties of frontier AI systems remains underexamined, particularly when results are used to inform high-stakes decisions. We present findings from interviews with 16 expert practitioners with experience conducting human uplift studies in domains including biosecurity, cybersecurity, education, and labor. Across interviews, experts described a recurring tension between the standard causal inference assumptions upon which human uplift studies rely and the object of study itself. Rapidly evolving AI systems, shifting baselines, heterogeneous and changing user proficiency, and porous real-world settings strain assumptions underlying internal, external, and construct validity, complicating the interpretation and appropriate use of uplift evidence. We contribute (1) a synthesis of methodological challenges in human uplift studies, mapped to risks to study validity and classified by their degree of specificity to large language model (LLM) systems, and (2) a mapping from challenges to proposed solutions. By collating expert-identified challenges and solutions, we seek to clarify the interpretive limits and appropriate uses of human uplift evidence, to align evaluation practice with the decisions it informs, and to support more coordinated methodological foundations for AI governance.
Authors: Ofir Marom
Abstract: The success of a Large Language Model (LLM) task depends heavily on its prompt. Most use-cases specify prompts using natural language, which is inherently ambiguous when multiple objectives must be simultaneously satisfied. In this paper we introduce UtilityMax Prompting, a framework that specifies tasks using formal mathematical language. We reconstruct the task as an influence diagram in which the LLM's answer is the sole decision variable. A utility function is defined over the conditional probability distributions within the diagram, and the LLM is instructed to find the answer that maximises expected utility. This constrains the LLM to reason explicitly about each component of the objective, directing its output toward a precise optimization target rather than a subjective natural language interpretation. We validate our approach on the MovieLens 1M dataset across three frontier models (Claude Sonnet 4.6, GPT-5.4, and Gemini 2.5 Pro), demonstrating consistent improvements in precision and Normalized Discounted Cumulative Gain (NDCG) over natural language baselines in a multi-objective movie recommendation task.
Authors: Boxuan Lyu, Haiyue Song, Zhi Qu
Abstract: Error Span Detection (ESD) is a crucial subtask in Machine Translation (MT) evaluation, aiming to identify the location and severity of translation errors. While fine-tuning models on human-annotated data improves ESD performance, acquiring such data is expensive and prone to inconsistencies among annotators. To address this, we propose a novel self-evolution framework based on Minimum Bayes Risk (MBR) decoding, named Iterative MBR Distillation for ESD, which eliminates the reliance on human annotations by leveraging an off-the-shelf LLM to generate pseudo-labels. Extensive experiments on the WMT Metrics Shared Task datasets demonstrate that models trained solely on these self-generated pseudo-labels outperform both unadapted base model and supervised baselines trained on human annotations at the system and span levels, while maintaining competitive sentence-level performance.
Authors: Francesco Pio Monaco, Elia Cunegatti, Flavio Vella, Giovanni Iacca
Abstract: Post-training model compression is essential for enhancing the portability of Large Language Models (LLMs) while preserving their performance. While several compression approaches have been proposed, less emphasis has been placed on selecting the most suitable set of data (the so-called \emph{calibration data}) for finding the compressed model configuration. The choice of calibration data is a critical step in preserving model capabilities both intra- and inter-tasks. In this work, we address the challenge of identifying high-performance calibration sets for both pruning and quantization by analyzing intrinsic data properties rather than model-specific signals. We introduce \texttt{\textbf{ZipCal}}, a model-agnostic data curation strategy that maximizes lexical diversity based on Zipfian power laws. Experiments demonstrate that our method consistently outperforms standard uniform random sampling across various pruning benchmarks. Notably, it also performs on par, in terms of downstream performance, with a state-of-the-art method that relies on model perplexity. The latter becomes prohibitively expensive at large-scale models and datasets, while \texttt{\textbf{ZipCal}} is on average $\sim$240$\times$ faster due to its tractable linear complexity\footnote{We make the code and the experiments available at https://github.com/FrancescoMonaco/ZipCal.}.
Authors: Abinav Rao, Sujan Rachuri
Abstract: Unified multimodal models share a language model backbone for both understanding and generating images. Can DPO align both capabilities simultaneously? We present the first systematic study of this question, applying DPO to Janus-Pro at 1B and 7B parameters under seven training strategies and two post-hoc methods. The central finding is negative: generation quality resists DPO alignment across all tested conditions on this architecture. No method improves generation CLIPScore at 7B (|Delta| < 0.2, p > 0.5 at n=200 per seed, 3 seeds); at 1B, all methods degrade generation, and the result holds across preference data types (real-vs-generated and model-vs-model) and the data volumes tested (150-288 pairs). Gradient analysis reveals why: understanding and generation gradients are near-orthogonal (cos ~ 0) with ~11-14x magnitude imbalance driven by VQ token count asymmetry (576 generation tokens vs. ~30-100 text tokens). This imbalance is the dominant interference mechanism in multi-task DPO; magnitude-balancing yields directionally positive understanding deltas (+0.01-0.04 VQA, though individually not significant), but the generation gap persists regardless. We identify discrete VQ tokenization as a likely structural bottleneck -- supported by the generation DPO loss converging to ln(2) -- and provide practical guidance for practitioners working with VQ-based unified models.
Authors: Ruishuo Chen, Yu Chen, Zhuoran Li, Longbo Huang
Abstract: Unsupervised Reinforcement Learning from Internal Feedback (RLIF) has emerged as a promising paradigm for eliciting the latent capabilities of Large Language Models (LLMs) without external supervision. However, current methods rely on heuristic intrinsic rewards, which often lack a well-defined theoretical optimization target and are prone to degenerative biases. In this work, we introduce PowerFlow, a principled framework that reformulates unsupervised fine-tuning as a distribution matching problem. By casting GFlowNet as an amortized variational sampler for unnormalized densities, we propose a length-aware Trajectory-Balance objective that explicitly neutralizes the structural length biases inherent in autoregressive generation. By targeting $\alpha$-power distributions, PowerFlow enables the directional elicitation of the dual nature of LLMs: sharpening the distribution ($\alpha > 1$) to intensify logical reasoning, or flattening it ($\alpha < 1$) to unlock expressive creativity. Extensive experiments demonstrate that PowerFlow consistently outperforms existing RLIF methods, matching or even exceeding supervised GRPO. Furthermore, by mitigating over-sharpening in aligned models, our approach achieves simultaneous gains in diversity and quality, shifting the Pareto frontier in creative tasks.
Authors: Haechan Kim, Soohyun Ryu, Gyouk Chu, Doohyuk Jang, Eunho Yang
Abstract: Reinforcement learning with verifiable rewards (RLVR) has emerged as an effective post-training paradigm for improving the reasoning capabilities of large language models. However, existing group-based RLVR methods often suffer from severe sample inefficiency. This inefficiency stems from reliance on point estimation of rewards from a small number of rollouts, leading to high estimation variance, variance collapse, and ineffective utilization of generated responses. In this work, we reformulate RLVR from a statistical estimation perspective by modeling rewards as samples drawn from a policy-induced distribution and casting advantage computation as the problem of estimating the reward distribution from finite data. Building on this view, we propose Discounted Beta-Bernoulli (DBB) reward estimation, which leverages historical reward statistics for the non-stationary distribution. Although biased, the resulting estimator exhibits reduced and stable variance, theoretically avoids estimated variance collapse, and achieves lower mean squared error than standard point estimation. Extensive experiments across six in-distribution and three out-of-distribution reasoning benchmarks demonstrate that GRPO with DBB consistently outperforms naive GRPO, achieving average Acc@8 improvements of 3.22/2.42 points in-distribution and 12.49/6.92 points out-of-distribution on the 1.7B and 8B models, respectively, without additional computational cost or memory usage.
Authors: Yu-Ning Qiu, Lin-Feng Zou, Jiong-Da Wang, Xue-Rong Yuan, Wang-Zhou Dai
Abstract: In high-complexity abstract reasoning, a system must infer a latent rule from a few examples or structured observations and apply it to unseen instances. LLMs can express such rules as programs, but ordinary conversation-based refinement is largely outcome-level: it observes that an answer or output is wrong without formally re-checking which abstraction, relation, or transformation justified that outcome. We propose \emph{Abduction-Based Procedural Refinement} (ABPR), a neuro-symbolic refinement approach that couples an LLM with a Prolog meta-interpreter. ABPR treats each candidate program as an executable declarative hypothesis of the latent rule and reifies its SLD goal--subgoal resolution into compact proof-tree-style derivations, following Shapiro's algorithmic program debugging (APD). In this view, refinement is not merely code-level debugging, but semantic re-checking of the model's hypothesised rule. We evaluate ABPR primarily on ARC-AGI-2, a challenging few-shot abstract rule induction benchmark over grid transformations. ABPR with Gemini-3-Flash achieves 56.67\% Pass@2, while GPT-5.5 xHigh with ABPR reaches 98.33\% Pass@2 on the public evaluation set. Supplementary experiments on fill-in-the-blank I-RAVEN-X and A-I-RAVEN adaptations provide evidence that the same trace-guided framework extends beyond ARC-specific grid tasks to RAVEN-style relational and analogical abstraction. Repeated-run and sensitivity analyses show that parallel trace-guided search reduces stochastic variance as search breadth and total search depth increase.
Authors: Robert Godwin-Jones
Abstract: Learning another language can be a highly emotional process, typically characterized by numerous frustrations and triumphs, big and small. For most learners, language learning does not follow a linear, predictable path, its zigzag course shaped by motivational (or demotivating) variables such as personal characteristics, teacher/peer relationships, learning materials, and dreams of a future L2 (second language) self. While some aspects of language learning (reading, grammar) are relatively mechanical, others can be stressful and unpredictable, especially conversing in the target language. That experience necessitates not only knowledge of structure and lexis, but also the ability to use the language in ways that are appropriate to the social and cultural context. A new opportunity to practice conversational abilities has arrived through the availability of AI chatbots, with both advantages (responsive, non-judgmental) and drawbacks (emotionally void, culturally biased). This column explores aspects of emotion as they arise in technology use and in particular how automatic emotion recognition and simulated human responsiveness in AI systems interface with language learning and the development of pragmatic and interactional competence. Emotion AI, the algorithmically driven interpretation of users' affective signals, has been seen as enabling greater personalized learning, adapting to perceived learner cognitive and emotional states. Others warn of emotional manipulation and inappropriate and ineffective user profiling
Authors: Yijun Wang, Qiyuan Zhuang, Xiu-Shen Wei
Abstract: Real-world time series exhibit temporally structured uncertainty: volatility clusters in turbulent regimes, dissipates in stable periods, and shifts abruptly around structural breaks. Yet many probabilistic forecasting methods estimate predictive uncertainty as an independent per-step quantity, leaving the evolution and persistence of volatility regimes under-modeled. We formalize this missing dimension as Temporal Uncertainty Dynamics and instantiate it in the Volatility Dynamics Variational Autoencoder (VolDy-VAE), a non-autoregressive generative forecaster with a location-scale decoder. VolDy-VAE combines a location path for mean prediction with a recurrent scale path that transfers and evolves a volatility hidden state from the look-back window to the forecasting horizon, enabling temporally coherent predictive variances. This design yields an adaptive attenuation mechanism: high-variance observations receive lower influence on the location estimate while their uncertainty is preserved through explicit scale predictions. We further provide a simplified regime-switching analysis showing that, when variances are known or consistently estimated, the volatility-aware objective reduces to inverse-variance weighting, whereas MSE-based estimators remain unbiased but statistically inefficient. Experiments on nine benchmarks show that VolDy-VAE improves forecasting accuracy and uncertainty calibration over strong probabilistic and point-forecasting baselines while maintaining low inference latency; plug-in studies further indicate that the VolDy principle can benefit GAN, Koopman VAE, and Transformer backbones. The source code is publicly available at https://github.com/wangyijunlyy/VolDy-VAE.
Authors: Yupei Yang, Lin Yang, Wanxi Deng, Lin Qu, Shikui Tu, Lei Xu
Abstract: Reranking is a critical component in many information retrieval pipelines. Despite remarkable progress in text-only settings, multimodal reranking remains challenging, particularly when the candidate set contains hybrid text and image items. A key difficulty is the modality gap: a text reranker is intrinsically closer to text candidates than to image candidates, leading to biased and suboptimal cross-modal ranking. Vision-language models (VLMs) mitigate this gap through strong cross-modal alignment and have recently been adopted to build multimodal rerankers. However, most VLM-based rerankers encode all candidates as images, and treating text as images introduces substantial computational overhead. Meanwhile, existing open-source multimodal rerankers are typically trained on general-domain data and often underperform in domain-specific scenarios. To address these limitations, we propose UniRank, a VLM-based reranking framework that natively scores and orders hybrid text-image candidates without any modality conversion. Building on this hybrid scoring interface, UniRank provides an end-to-end domain adaptation pipeline that includes: (1) an instruction-tuning stage that learns calibrated cross-modal relevance scoring by mapping label-token likelihoods to a unified scalar score; and (2) a hard-negative-driven preference alignment stage that constructs in-domain pairwise preferences and performs query-level policy optimization through reinforcement learning from human feedback (RLHF). Extensive experiments on scientific literature retrieval and design patent search demonstrate that UniRank consistently outperforms state-of-the-art baselines, improving Recall@1 by 8.9% and 7.3%, respectively.
Authors: Xue Qin, Simin Luan, John See, Cong Yang, Zhijun Li
Abstract: Robotic systems lack a principled abstraction for organizing intelligence, capabilities, and execution in a unified manner. Existing approaches either couple skills within monolithic architectures or decompose functionality into loosely coordinated modules or multiple agents, often without a coherent model of identity and control authority. We argue that a robot should be modeled as a single persistent intelligent subject whose capabilities are extended through installable packages. We formalize this view as AEROS (Agent Execution Runtime Operating System), in which each robot corresponds to one persistent agent and capabilities are provided through Embodied Capability Modules (ECMs). Each ECM encapsulates executable skills, models, and tools, while execution constraints and safety guarantees are enforced by a policy-separated runtime. This separation enables modular extensibility, composable capability execution, and consistent system-level safety. We evaluate a reference implementation in PyBullet simulation with a Franka Panda 7-DOF manipulator across eight experiments covering re-planning, failure recovery, policy enforcement, baseline comparison, cross-task generality, ECM hot-swapping, ablation, and failure boundary analysis. Over 100 randomized trials per condition, AEROS achieves 100% task success across three tasks versus baselines (BehaviorTree.CPP-style and ProgPrompt-style at 92--93%, flat pipeline at 67--73%), the policy layer blocks all invalid actions with zero false acceptances, runtime benefits generalize across tasks without task-specific tuning, and ECMs load at runtime with 100% post-swap success.
Authors: Xiangyuan Wang, Honghao Cai, Yunhao Bai, Chao Hui, Tianze Zhou, Haohua Chen, Hao Shi, Yuling Wu, Yao Hu, Xu Tang, Yibo Chen, Wei Zhu
Abstract: High-quality source-target image pairs with precise editing instructions are essential for instruction-guided image editing, yet constructing such training triplets at scale remains costly. Recent pipelines often rely on vision-language models to synthesize editing instructions automatically, but we find that strong VLMs still struggle to describe visual transformations between image pairs. In particular, they exhibit three recurring failure modes: orientation inconsistency, viewpoint ambiguity, and missing fine-grained attributes. In a human evaluation on 400 image pairs, several open-source VLM baselines produce critical-error rates above 47\%, making many synthesized instructions unsuitable for downstream training. To address this, we propose EditCaption, a two-stage post-training pipeline for image editing instruction synthesis. First, we construct a 100K supervised fine-tuning dataset through GLM-based auto-captioning, EditScore filtering, and human refinement. Second, we collect 10K human-annotated preference pairs, where each rejected instruction is labeled with its primary error type and severity. Based on this dataset, we propose Hardness-Adaptive Error-Aware DPO (HAE-DPO), a task-adapted DPO objective that introduces an adaptive margin based on human-labeled severity, failure-mode type, and reference-model hardness. Experiments across three benchmarks demonstrate that our 235B model with SFT+HAE-DPO achieves state-of-the-art performance among open-source and closed models, scoring 4.720 on Eval-400, 4.672 on HQ-Edit, and 4.651 on ByteMorph-Bench -- surpassing Gemini-3-Pro on all three. Human evaluation confirms critical error rates drop from 47.75\% to 17.50\%, with correct rates improving from 41.75\% to 70.25\%, surpassing Gemini-3-Pro (66.00\%).
Authors: Rafael da Silva, Jeff Eicher, Gregory Longo
Abstract: Student dropout is a persistent concern in Learning Analytics, yet comparative studies frequently evaluate predictive models under heterogeneous protocols, prioritizing discrimination over temporal interpretability and calibration. This study introduces a survival-oriented benchmark for temporal dropout risk modelling using the Open University Learning Analytics Dataset (OULAD). Two harmonized arms are compared: a dynamic weekly arm, with models in person-period representation, and a comparable continuous-time arm, with an expanded roster of families -- tree-based survival, parametric, and neural models. The evaluation protocol integrates four analytical layers: predictive performance, ablation, explainability, and calibration. Results are reported within each arm separately, as a single cross-arm ranking is not methodologically warranted. Within the comparable arm, Random Survival Forest leads in discrimination and horizon-specific Brier scores; within the dynamic arm, Poisson Piecewise-Exponential leads narrowly on integrated Brier score within a tight five-family cluster. No-refit bootstrap sampling variability qualifies these positions as directional signals rather than absolute superiority. Ablation and explainability analyses converged, across all families, on a shared finding: the dominant predictive signal was not primarily demographic or structural, but temporal and behavioral. Calibration corroborated this pattern in the better-discriminating models, with the exception of XGBoost AFT, which exhibited systematic bias. These results support the value of a harmonized, multi-dimensional benchmark in Learning Analytics and situate dropout risk as a temporal-behavioral process rather than a function of static background attributes.
Authors: Wenbo Pan, Shujie Liu, Xiangyang Zhou, Shiwei Zhang, Wanlu Shi, Mirror Xu, Xiaohua Jia
Abstract: Large language model agents rely on specialized memory systems to accumulate and reuse knowledge during extended interactions. Recent architectures typically adopt a fixed memory design tailored to specific domains, such as semantic retrieval for conversations or skills reused for coding. However, a memory system optimized for one purpose frequently fails to transfer to others. To address this limitation, we introduce M$^\star$, a method that automatically discovers task-optimized memory harnesses through executable program evolution. Specifically, M$^\star$ models an agent memory system as a memory program written in Python. This program encapsulates the data Schema, the storage Logic, and the agent workflow Instructions. We optimize these components jointly using a reflective code evolution method; this approach employs a population-based search strategy and analyzes evaluation failures to iteratively refine the candidate programs. We evaluate M$^\star$ on four distinct benchmarks spanning conversation, embodied planning, and expert reasoning. Our results demonstrate that M$^\star$ improves performance over existing fixed-memory baselines robustly across all evaluated tasks. Furthermore, the evolved memory programs exhibit structurally distinct processing mechanisms for each domain. This finding indicates that specializing the memory mechanism for a given task explores a broad design space and provides a superior solution compared to general-purpose memory paradigms.
Authors: Ziyang Liu
Abstract: When LLM conversations grow beyond the context window, old content must be evicted -- but how does the model recover it when needed? We propose cooperative paging: evicted segments are replaced with minimal keyword bookmarks ([pN:keywords], ~8-24 tokens each), and the model is given a recall() tool to retrieve full content on demand. On the LoCoMo benchmark (10 real multi-session conversations, 300+ turns), cooperative paging achieves the highest answer quality among six methods -- outperforming truncation, BM25, word-overlap retrieval, a search-tool baseline, and full context -- on four models (GPT-4o-mini, DeepSeek-v3.2, Claude Haiku, GLM-5), confirmed by four independent LLM judges ($p=0.017$, paired bootstrap). We then study the paging design space with a 5x4 ablation over boundary strategies and eviction policies (3,176 synthetic probes, 1,600 LoCoMo probes). Key findings: (1) coarse fixed-size pages (fixed_20) reach 96.7% while content-aware topic_shift collapses to 56.7%; (2) eviction policy choice is data-dependent (FIFO best on synthetic, LFU on LoCoMo); (3) two bookmark generation strategies improve over the heuristic baseline (+4.4 and +8.7 E2E points); (4) the remaining bottleneck is bookmark discrimination -- the model triggers recall() 96% of the time but selects the correct page only 57% when bookmarks are insufficiently distinctive. Keyword specificity alone accounts for a 25 percentage point accuracy difference.
Authors: Fei Ding, Yongkang Zhang, youwei wang, Zijian Zeng
Abstract: Reinforcement learning for multi-step reasoning with large language models (LLMs) typically relies on sparse terminal rewards, which creates a poorly conditioned credit-assignment problem: the final feedback is propagated uniformly across all intermediate decisions. This leads to high gradient variance, unstable training, and many ineffective updates, ultimately limiting sustained model improvement. We propose a counterfactual-comparison framework for credit assignment. For each input, the framework samples multiple reasoning trajectories and treats their differences as implicit approximations to alternative decisions. This yields an implicit process-level advantage estimator that converts sparse terminal rewards into step-sensitive learning signals. Building on this framework, we introduce Implicit Behavior Policy Optimization (IBPO), which substantially improves training stability and the performance ceiling on mathematical and code-reasoning benchmarks. Our results point to a promising direction for unlocking the reasoning potential of LLMs.
Authors: Dixi Yao, Tahseen Rabbani, Manzil Zaheer, Tian Li
Abstract: We propose a federated learning-like framework, Federation over Text (FoT), that enables multiple clients solving different tasks to collectively generate a shared library of metacognitive insights by iteratively federating their local reasoning processes without sharing actual problem instances or task instructions. Instead of federation over gradients (e.g., as in distributed training), FoT operates at the semantic level without any gradient optimization or supervision signal. Iteratively, each client runs an LLM agent that does local thinking and self-improvement on their specific tasks independently, and shares reasoning traces with a central server, which aggregates and distills them into a cross-task (and cross-domain) insight library that existing and future agents can leverage to improve performance on related tasks. Experiments show that FoT improves reasoning effectiveness and efficiency across a wide range of challenging applications, including mathematical problem solving, cross-domain collaboration, real-world daily tasks, and machine learning research insight discovery. Specifically, it improves average performance scores by 25% while reducing the reasoning tokens by 4% across the first three applications. In the research insight discovery application, FoT is able to generate insights that cover over 80% of the major contributions in the subsequent papers.
Authors: Fei Ding, Yongkang Zhang, Runhao Liu, Yuhao Liao, Zijian Zeng, Huiming Yang, Sibo wang, Linglin Liao
Abstract: This paper investigates the length problem in sequence-level relative reinforcement learning. We observe that, although existing methods partially alleviate length-related phenomena, a more fundamental issue remains insufficiently characterized: the comparison units used during training lack inherent comparability. Building on this observation, we propose a new perspective: the length problem should not be viewed merely as a loss-scaling or normalization bias, but rather as a \emph{comparison unit construction} problem. We further establish a sample-construction-based training framework that, instead of applying post-hoc corrections to unequal-length responses, proactively constructs equal-length, alignable, and comparable training segments during generation. Within this framework, we propose EqLen, a concrete method applicable to group-relative comparison algorithms such as GRPO, GSPO, and RLOO. Through dual-track synchronous generation, prefix inheritance, and segment masking, EqLen efficiently collects effective equal-length training segments and enables stable
Authors: Ziyang Liu
Abstract: We study post-training W4A4 quantization in a controlled 300M-parameter SwiGLU decoder-only language model trained on 5B tokens of FineWeb-Edu, and ask which input-activation sites dominate the error. Naive round-to-nearest W4A4 collapses validation perplexity from FP16 23.6 to 1727. A simple residual-axis training-time intervention -- Depth Registers with a register-magnitude hinge loss (DR+sink) -- reduces this to 119 (about 14x) at matched FP16 PPL and matched zero-shot capacity, and composes with SmoothQuant to 39.9 PPL. The residual ~2 PPL gap to FP16 is the diagnostic core. We decompose W4A4 damage by input-activation site: the five trainable linears in a SwiGLU block split into residual-axis readers (qkv, w1, w3) and block-internal generators (o_proj, w2). Elementary norm arguments show residual-axis magnitude control bounds readers tightly but leaves w2's bilinear input bounded only by the trivial product of factor bounds; empirically, DR+sink collapses reader kurtosis while leaving generators essentially unchanged, and the reader-rescued W4A4 residue is flat at ~0.28 nats across three matched checkpoints with Delta-remove(w2) dominating. We present DR+sink as a training-time probe rather than a deployment proposal: a post-hoc alternative (Per-Linear QuaRot) nearly matches it on the reader axis. Full QuaRot -- adding online per-head value Hadamard plus online w2-input rotation -- does not close the gap either, directly testing the prediction that orthogonal rotation cannot bound the bilinear SwiGLU tail. Claims are specific to our 300M, 5B-token, single-seed setting, and our experiments do not isolate the partition from the hinge.
Authors: Ziyang Liu
Abstract: LLMs edit text and code by autoregressively regenerating the full output, even when most tokens appear verbatim in the input. We study Copy-as-Decode, a decoding-layer mechanism that recasts edit generation as structured decoding over a two-primitive grammar:
Authors: Ziyang Liu
Abstract: Hosted-LLM providers have a silent-substitution incentive: advertise a stronger model while serving cheaper replies. Probe-after-return schemes such as SVIP leave a parallel-serve side-channel, since a dishonest provider can route the verifier's probe to the advertised model while serving ordinary users from a substitute. We propose a commit-open protocol that closes this gap. Before any opening request, the provider commits via a Merkle tree to a per-position sparse-autoencoder (SAE) feature-trace sketch of its served output at a published probe layer. A verifier opens random positions, scores them against a public named-circuit probe library calibrated with cross-backend noise, and decides with a fixed-threshold joint-consistency z-score rule. We instantiate the protocol on three backbones -- Qwen3-1.7B, Gemma-2-2B, and a 4.5x scale-up to Gemma-2-9B with a 131k-feature SAE. Of 17 attackers spanning same-family lifts, cross-family substitutes, and rank-<=128 adaptive LoRA, all are rejected at a shared, scale-stable threshold; the same attackers all evade a matched SVIP-style parallel-serve baseline. A white-box end-to-end attack that backpropagates through the frozen SAE encoder does not close the margin, and a feature-forgery attacker that never runs M_hon is bounded in closed form by an intrinsic-dimension argument. Commitment adds <=2.1% to forward-only wall-clock at batch 32.
Authors: Yusuf Kesmen, Fay Elhassan, Jiayi Ma, Julien Stalhandske, Yena Chang, David Sasu, Alexandra Kulinkina, Akhil Arora, Lars Klein, Mary-Anne Hartley
Abstract: Large language models (LLMs) are increasingly used for conversational clinical decision support, yet they conflate next token prediction with probabilistic decision making. We argue that this conflation reflects an architectural limitation: such systems lack explicit posterior tracking, controllable abstention thresholds, and auditable reasoning chains. We introduce MoBayes, a Modular Bayesian dialogue framework that separates reasoning from language. The LLM acts only as a language interface, parsing patient conversation into structured observations, while a Bayesian module performs probabilistic inference over these observations to update posteriors, select follow-up questions via expected-information-gain and determine when to stop or defer through calibrated decision thresholds. This design enables explicit posterior tracking, controllable selective decision-making, and replaceable population-specific statistical backends without retraining the language model. Across empirical and LLM-generated knowledge bases, MoBayes outperforms standalone frontier LLM doctors, including matched model-family comparisons where inexpensive sensor models paired with MoBayes exceed larger autonomous models at lower cost. The advantage persists under adversarial patient communication styles and across varying diagnostic scenarios. These results suggest that reliable conversational clinical decision support systems should separate probabilistic reasoning from language generation rather than scaling model size alone. Code is available at https://anonymous.4open.science/r/MoBayes/
Authors: Bhaskar Singh, Shobhit Banga, Mahima Manik, Pranav Sharma
Abstract: Full-duplex spoken dialogue systems can model natural conversational behaviours such as interruptions, overlaps, and backchannels, yet such systems remain largely unexplored for Indian languages. We present the first open, reproducible full-duplex spoken dialogue system for Hindi by adapting Moshi, a state-of-the-art duplex speech architecture, using a custom Hindi tokeniser and training on 26,000 hours of real spontaneous conversations collected from 14,695 speakers with separate speaker channels, enabling direct learning of turn-taking and overlap patterns from natural interactions. To support Hindi text generation, we replace the original English tokeniser and reinitialise text-vocabulary-dependent parameters while retaining the pre-trained audio components. We propose a two-stage training recipe -- large-scale pre-training followed by fine-tuning on 1,000 hours of conversational data. Evaluation through the prompted dialogue continuation paradigm with both automatic metrics and human judgments demonstrates that the resulting model generates natural and meaningful full-duplex conversational behaviour in Hindi. This work serves as a first step toward real-time duplex spoken dialogue systems for Hindi and other Indian languages.
Authors: Xinxing Wu
Abstract: Slide-based teaching is widely used in higher education, yet in online, hybrid, and asynchronous contexts, slides often lose instructor presence, narrative continuity, and expressive framing that help learners connect with course content. Full lecture video can partly restore these qualities, but it is time-consuming to record, revise, and reuse. This study presents a practice-based implementation and analytic reflection of an open-source workflow for creating talking slide avatars. The workflow integrates OpenVoice for text-to-speech and authorized voice-style conversion with Ditto-TalkingHead for audio-driven talking-image synthesis, enabling instructors to transform a short script and an authorized or synthetic portrait image into a narrated video for slide decks or HTML-based lecture materials. Rather than treating this workflow only as a technical solution, the study frames talking slide avatars as multimodal communication artifacts at the intersection of digital pedagogy, aesthetic education, and art-technology practice. The paper documents the production pipeline, analyzes communicative and aesthetic affordances, and proposes practical guidelines for script length, image selection, pacing, disclosure, accessibility, consent, and ethical use. Its contribution is not a validated learning intervention, but an educator-oriented open-source production model and communication-design framework. The study concludes that short, transparent, and carefully designed avatars may provide a reusable communication layer for introductions, transitions, reminders, and recaps when used selectively and with appropriate ethical safeguards.
Authors: Yanping Wu, Meiting Dang, Lin Wu, Edmond S. L. Ho, Zhenghua Chen, Chongfeng Wei
Abstract: Recent advances in autonomous driving have motivated research on pedestrian intention prediction, which aims to infer future crossing decisions and actions by modeling temporal dynamics, social interactions, and environmental context. However, existing studies remain constrained by oversimplified multi-agent interaction patterns, opaque reasoning logic, and a lack of global consistency in behavioral predictions, which compromise both robustness and interpretability. In this work, we propose ESIA (Energy-based Spatiotemporal Interaction-Aware framework), a novel Conditional Random Field (CRF)-based paradigm. We cast the intention prediction task as a structured prediction problem over a unified graph-based representation, treating pedestrians and the environment as spatiotemporal nodes. To characterize their distinct roles, we assign unary potentials to nodes to capture individual intentions, and pairwise potentials to edges to encode social and environmental interactions. These potentials are integrated into a unified global energy function to ensure scene-level consistency across behavioral predictions. To further constrain inference without ground-truth supervision, we introduce structural consistency terms to penalize logical contradictions. This optimization is efficiently solved via a novel Unary-Seeded Simulated Annealing (U-SSA) algorithm, which leverages high-confidence unary priors to rapidly converge to a high-quality solution. Extensive experiments on standard benchmarks demonstrate that ESIA achieves state-of-the-art performance with improved interpretability over existing methods.
Authors: Peiyang Liu, Ziqiang Cui, Xi Wang, Di Liang, Wei Ye
Abstract: Iterative Retrieval-Augmented Generation (iRAG) has emerged as a powerful paradigm for answering complex multi-hop questions by progressively retrieving and reasoning over external documents. However, current systems predominantly operate on parsed text, which creates two critical bottlenecks: (1) \textit{Coarse-grained attribution}, where users are burdened with manually locating evidence within lengthy documents based on vague text-level citations; and (2) \textit{Visual semantic loss}, where the conversion of visually rich documents (e.g., slides, PDFs with charts) into text discards spatial logic and layout cues essential for reasoning. To bridge this gap, we present \textbf{Chain of Evidence (CoE)}, a retriever-agnostic visual attribution framework that leverages Vision-Language Models to reason directly over screenshots of retrieved document candidates. CoE eliminates format-specific parsing and outputs precise bounding boxes, visualizing the complete reasoning chain within the retrieved candidate set. We evaluate CoE on two distinct benchmarks: \textbf{Wiki-CoE}, a large-scale dataset of structured web pages derived from 2WikiMultiHopQA, and \textbf{SlideVQA}, a challenging dataset of presentation slides featuring complex diagrams and free-form layouts. Experiments demonstrate that fine-tuned Qwen3-VL-8B-Instruct achieves robust performance, significantly outperforming text-based baselines in scenarios requiring visual layout understanding, while establishing a retriever-agnostic solution for pixel-level interpretable iRAG. Our code is available at https://github.com/PeiYangLiu/CoE.git.
Authors: Zijian An, Hadi Khezam, Bill Cai, Ran Yang, Shijie Geng, Yiming Feng, Yue Zheng, Lifeng Zhou
Abstract: We present VILAS, a fully low-cost, modular robotic manipulation platform designed to support end-to-end vision-language-action (VLA) policy learning and deployment on accessible hardware. The system integrates a Fairino FR5 collaborative arm, a Jodell RG52-50 electric gripper, and a dual-camera perception module, unified through a ZMQ-based communication architecture that seamlessly coordinates teleoperation, data collection, and policy deployment within a single framework. To enable safe manipulation of fragile objects without relying on explicit force sensing, we design a kirigami-based soft compliant gripper extension that induces predictable deformation under compressive loading, providing gentle and repeatable contact with delicate targets. We deploy and evaluate three state-of-the-art VLA models on the VILAS platform: pi_0, pi_0.5, and GR00T N1.6. All models are fine-tuned from publicly released pretrained checkpoints using an identical demonstration dataset collected via our teleoperation pipeline. Experiments on a grape grasping task validate the effectiveness of the proposed system, confirming that capable manipulation policies can be successfully trained and deployed on low-cost modular hardware. Our results further provide practical insights into the deployment characteristics of current VLA models in real-world settings.
Authors: Reza Rastegar
Abstract: Softmax routing approaches hard top-1 routing as the temperature tends to zero, but the limiting passage is singular at router ties. This paper develops a boundary-layer calculus for this soft-to-hard limit in population squared-loss mixture-of-experts regression. For a router with logits $a_k(x;\phi)$, the relevant local quantity is the top-two margin $\Delta(x;\phi)$, and the relevant global quantity is the boundary mass $\mathbb{P}(\Delta(X;\phi)\le w)$. Under smoothness and transversality assumptions, coarea and tubular-neighborhood estimates show how this mass scales with the slab width; in the binary case the leading coefficient is an explicit surface integral over the routing interface. These geometric estimates give quantitative bounds between the soft objective $L_\tau$ and the hard objective $L_0$, including an $O(\tau^\alpha)$ uniform comparison under a margin-tail condition, and yield $\Gamma$-convergence of the soft objectives on compact parameter spaces. The main conclusion is that the zero-temperature approximation is controlled by the probability carried by an $O(\tau)$ neighborhood of the routing interfaces, not by temperature alone. After isolating this boundary-layer part of the problem, we record a conditional landscape-transfer theorem from hard to small-temperature soft routing and a reduced two-expert Gaussian calculation illustrating local symmetry breaking. Synthetic diagnostics are included only as controlled checks of the boundary-layer predictions.
Authors: Chenye Yang, Weiyu Xu, Lifeng Lai
Abstract: Offline Reinforcement Learning from Human Feedback (RLHF) pipelines such as Direct Preference Optimization (DPO) train on a pre-collected preference dataset, which makes them vulnerable to preference poisoning attack. We study label flip attacks against log-linear DPO. We first illustrate that flipping one preference label induces a parameter-independent shift in the DPO gradient. Using this key property, we can then convert the targeted poisoning problem into a structured binary sparse approximation problem. To solve this problem, we develop two attack methods: Binary-Aware Lattice Attack (BAL-A) and Binary Matching Pursuit Attack (BMP-A). BAL-A embeds the binary flip selection problem into a binary-aware lattice and applies Lenstra-Lenstra-Lov\'asz reduction and Babai's nearest plane algorithm; we provide sufficient conditions that enforce binary coefficients and recover the minimum-flip objective. BMP-A adapts binary matching pursuit to our non-normalized gradient dictionary and yields coherence-based recovery guarantees and robustness (impossibility) certificates for $K$-flip budgets. Experiments on synthetic dictionaries and the Stanford Human Preferences dataset validate the theory and highlight how dictionary geometry governs attack success.
Authors: Xiao Li, Xiang Zheng, Yifeng Gao, Xinyu Xia, Yixu Wang, Xin Wang, Ye Sun, Yunhan Zhao, Ming Wen, Jiayu Li, Zixing Chen, Xun Gong, Yi Liu, Yige Li, Yutao Wu, Cong Wang, Jun Sun, Yixin Cao, Zhineng Chen, Jingjing Chen, Tao Gui, Qi Zhang, Zuxuan Wu, Xipeng Qiu, Xuanjing Huang, Tiehua Zhang, Zhipeng Wei, Kun Wang, Xinfeng Li, Hanxun Huang, Sarah Erfani, James Bailey, Jianping Wang, Chaowei Xiao, Ran He, Bo Li, Xingjun Ma, Yu-Gang Jiang
Abstract: Embodied Artificial Intelligence (Embodied AI) integrates perception, cognition, planning, and interaction into agents that operate in open-world, safety-critical environments. As these systems gain autonomy and enter domains such as transportation, healthcare, and industrial or assistive robotics, ensuring their safety becomes both technically challenging and socially indispensable. Unlike digital AI systems, embodied agents must act under uncertain sensing, incomplete knowledge, and dynamic human-robot interactions, where failures can directly lead to physical harm. This survey provides a comprehensive and structured review of safety research in embodied AI, examining attacks and defenses across the full embodied pipeline, from perception and cognition to planning, action and interaction, and agentic system. We introduce a multi-level taxonomy that unifies fragmented lines of work and connects embodied-specific safety findings with broader advances in vision, language, and multimodal foundation models. Our review synthesizes insights from over 500 papers spanning adversarial, backdoor, jailbreak, and hardware-level attacks; attack detection, safe training and robust inference; and risk-aware human-agent interaction. This analysis reveals several overlooked challenges, including the fragility of multimodal perception fusion, the instability of planning under jailbreak attacks, and the trustworthiness of human-agent interaction in open-ended scenarios. By organizing the field into a coherent framework and identifying critical research gaps, this survey provides a roadmap for building embodied agents that are not only capable and autonomous but also safe, robust, and reliable in real-world deployment.
Authors: Zhenghao Huang, Huilin Yao, Kaikai Wang
Abstract: Surface electromyography provides a practical way to infer human movement intention from wearable muscle recordings, but models trained under a single acquisition setting often lose reliability when the user, session, electrode layout, or gesture protocol changes. This paper proposes AEMG, a self-supervised learning approach designed to extract reusable neuromuscular representations from diverse EMG sources. Eight public gesture datasets are first transformed into a shared signal format to reduce discrepancies in channel configuration, sensor topology, and recording protocol. Instead of relying on fixed-length sliding windows, AEMG identifies contraction events from energy variations and represents them as compact neuromuscular tokens, while ordered token groups describe the coordinated activity of multiple muscles during motion. A spatially and temporally conditioned Transformer is then used to encode these token sequences, preserving information about electrode position, activation timing, and sequential structure. For pre-training, the model constructs a discrete library of contraction prototypes through vector-quantized reconstruction and further learns contextual dependencies by recovering masked neuromuscular tokens from surrounding observations. Experiments under leave-one-subject-out and low-label adaptation settings show that the learned representation improves robustness to unseen users and reduces the amount of calibration data required for gesture recognition. These findings suggest that event-level token modeling offers a scalable route toward adaptable and data-efficient EMG-based motor-intent understanding.
Authors: Tianze Han, Beining Xu, Hanbo Zhang, Yongming Lu
Abstract: Mental-health dialogue models are increasingly evaluated by AI-based evaluators, yet these evaluators often treat surface empathy, supportiveness, or fluency as evidence of safety. In this paper, we study a hidden failure mode that we call implicit sycophancy: a response may appear empathetic while implicitly reinforcing catastrophizing, avoidance, hopeless prediction, or CBT-style labeling. To examine this problem, we introduce a diagnostic benchmark for implicit-sycophancy detection, built from three representative mental-health dialogue sources covering everyday peer support, counseling-style emotional support, and crisis-oriented interaction, and further construct a leakage-audited clean single-response matched benchmark with 500 contexts and 1,500 matched response windows. We then propose Dynamic Emotional Signature Graphs (DESG), a structured offline audit framework that separates LLM-based state extraction from final scoring and evaluates clinical direction through semantic, affective, and cognitive-distortion state transitions rather than free-form LLM judgment. Unlike metadata, surface-style, lexical, embedding, and rubric-LLM baselines, DESG scores the direction of clinical-state change induced by a response; on the leakage-audited clean matched benchmark, DESG-StateRisk improves over the strongest non-DESG baseline by 0.0488 macro-F1 and achieves the best harmful-risk detection result. These results suggest that evaluating implicit sycophancy requires explicit clinical-state modeling together with leakage checks, shortcut controls, and competitive baselines.
Authors: Ahmed Cherif
Abstract: Low-light images suffer from poor visibility, noise, and color distortion. Existing Retinex-based enhancement methods rely on manually tuned parameters that do not generalize across different lighting conditions. This paper proposes BFORE (Butterfly-Firefly Optimized Retinex Enhancement), a framework that automatically finds the best enhancement parameters for each image. BFORE works in two phases: (1) a Butterfly Optimization Algorithm (BOA) searches for optimal Multi-Scale Retinex with Color Restoration (MSRCR) parameters, then (2) a Firefly Algorithm (FA) fine-tunes gamma correction, denoising, and color parameters. Both phases maximize a Gaussian Naturalness Score (GNS), a no-reference metric that measures how natural the enhanced image looks. Standard quality metrics (PSNR, SSIM, NIQE) are computed only after optimization, ensuring zero data leakage. On 30 synthetic image pairs, BFORE achieves GNS = 0.971, outperforming the next-best method MSRCR (0.894) by 8.6%. On 115 real images from the LOL dataset, BFORE achieves GNS = 0.887, outperforming MSRCR (0.808) by 9.8%. A controlled comparison with three deep learning baselines (Zero-DCE, SCI, IAT) trained under identical conditions shows BFORE surpasses the best DL method by 14.7% in GNS. An ablation study confirms that the hybrid BOA+FA strategy significantly outperforms each optimizer in isolation, and a scalability analysis at three evaluation budgets shows that the structured optimizer significantly outperforms uniform random sampling once compute is available (p = 0.009 at 128 evaluations, p = 0.021 at 300 evaluations). All improvements are statistically significant (p < 0.0001, Wilcoxon signed-rank test). Processing time is 3-6 minutes per image on CPU, suitable for offline applications.
Authors: Hamed Karimi, Vaishali Meyappan, Reza Samavi
Abstract: LLMs' overconfidence, particularly when hallucinating, poses a significant challenge for the deployment of the models in safety-critical settings and makes a reliable estimation of uncertainty necessary. Existing approaches for uncertainty quantification typically prioritize lexical or probabilistic measures; however, these techniques often ignore the semantic variance of different responses with similar meaning. In this paper, we propose Adaptive Conformal Semantic Entropy (ACSE), a method for estimating prompt-level uncertainty by adaptively measuring semantic dispersion in LLMs outputs. Our uncertainty scoring function is based on clustering semantic entropy of multiple diverse responses to the same prompt. The function adaptively adjusts the uncertainty score based on semantic features of each cluster. To ensure statistical reliability of our score, we use conformal calibration to apply a decision rule to accept/abstain the prompts, providing a finite-sample, distribution-free guarantee such that the error rate among the accepted responses remains bounded by a user-specified tolerance. Our extensive experimental evaluations using different LLMs and datasets, demonstrate that our approach consistently outperforms state-of-the-art uncertainty quantification baselines using discriminative performance, conformal guarantees, and probabilistic calibration indicators. As a highlight, for TriviaQA dataset, AUROC of our approach is 0.88 compared to 0.65 produced by the token entropy approach.
Authors: Seunghan Lee
Abstract: TabPFN has recently gained attention as a foundation model for tabular datasets, achieving strong performance by leveraging in-context learning on synthetic data. However, we find that TabPFN is vulnerable to label shift, often overfitting to the majority class in the training dataset. To address this limitation, we propose DistPFN, the first test-time posterior adjustment method designed for tabular foundation models. DistPFN rescales predicted class probabilities by downweighting the influence of the training prior (i.e., the class distribution of the context) and emphasizing the contribution of the model's predicted posterior, without architectural modification or additional training. We further introduce DistPFN-T, which incorporates temperature scaling to adaptively control the adjustment strength based on the discrepancy between prior and posterior. We evaluate our methods on over 250 OpenML datasets, demonstrating substantial improvements for various TabPFN-based models in classification tasks under label shift, while maintaining strong performance in standard settings without label shift. Code is available at this repository: https://github.com/seunghan96/DistPFN.
Authors: Zheng Fang, Xiaosen Wang, Shenyi Zhang, Shaokang Wang, Zhijin Ge
Abstract: Jailbreak attacks on audio language models (ALMs) optimize audio perturbations to elicit unsafe generations, and they typically update the entire waveform densely throughout optimization. In this work, we investigate the necessity of such dense optimization by analyzing the structure of token-aligned gradients in ALMs. We find that gradient energy is highly non-uniform across audio tokens, indicating that only a small subset of token-aligned audio regions dominates the optimization signal. Motivated by this observation, we propose Token-Aware Gradient Optimization (TAGO), which enables sparse jailbreak optimization by retaining only waveform gradients aligned with audio tokens that have high gradient energy, while masking the remaining gradients at each iteration. Across three ALMs, TAGO outperforms baselines, and substantial sparsification preserves strong attack success rates (e.g. on Qwen3-Omni, $\mathrm{ASR}_{l}$ remains at 86% with a token retention ratio of 0.25, compared to 87% with full token retention). These results demonstrate that dense waveform updates are largely redundant, and we advocate that future audio jailbreak and safety alignment research should further leverage this heterogeneous token-level gradient structure.
Authors: Chuanzhi Xu, Boyu Wei, Haoxian Zhou, Xuanhua Yin, Zihan Deng, Haodong Chen, Qiang Qu, Weidong Cai
Abstract: As 3D Gaussian Splatting (3DGS) gains attention in immersive media and digital content creation, assessing the aesthetics of 3D scenes becomes important in helping creators build more visually compelling 3D content. However, existing evaluation methods for 3D scenes primarily emphasize reconstruction fidelity and perceptual realism, largely overlooking higher-level aesthetic attributes such as composition, harmony, and visual appeal. This limitation comes from two key challenges: (1) the absence of general 3DGS datasets with aesthetic annotations, and (2) the intrinsic nature of 3DGS as a low-level primitive representation, which makes it difficult to capture high-level aesthetic features. To address these challenges, we propose Aes3D, the first systematic framework for assessing the aesthetics of 3D neural rendering scenes. Aes3D includes Aesthetic3D, the first dataset dedicated to 3D scene aesthetic assessment, built on our proposed annotation strategy for 3D scene aesthetics. In addition, we present Aes3DGSNet, a lightweight model that directly predicts scene-level aesthetic scores from 3DGS representations. Notably, our model operates solely on 3D Gaussian primitives, eliminating the need for rendering multi-view images and thus reducing computational cost and hardware requirements. Through aesthetics-supervised learning on multi-view 3DGS scene representations, Aes3DGSNet effectively captures high-level aesthetic cues and accurately regresses aesthetic scores. Experimental results demonstrate that our approach achieves strong performance while maintaining a lightweight design, establishing a new benchmark for 3D scene aesthetic assessment. Code and datasets will be made available in a future version.
Authors: Fei Ding, Yongkang Zhang, Runhao Liu, Yuhao Liao, Zijian Zeng, Sibo wang, Huiming Yang
Abstract: The central challenge of reinforcement learning for reasoning lies not only in the sparsity of outcome-level supervision, but more fundamentally in how to transform feedback provided only at the end of a sequence into fine-grained learning signals that can guide intermediate reasoning steps. Existing approaches either rely on outcome-level rewards for sequence-level optimization, which makes precise credit assignment difficult, or depend on externally constructed process supervision, which is costly and difficult to scale sustainably. To address this, we propose a new perspective: reinforcement learning for reasoning can be understood as the problem of internalizing outcome supervision into process supervision. From this perspective, we introduce a supervision-internalization method for reinforcement learning for reasoning, enabling the model to automatically extract process-level learning signals through identifying, correcting, and reusing failed reasoning trajectories, thereby achieving finer-grained policy optimization under outcome-only supervision. We further abstract this idea into a new training paradigm, in which the model continually generates and refines its own internal process supervision during reinforcement learning, opening a new path for fine-grained credit assignment in reinforcement learning for reasoning that differs from externally provided process supervision.
Authors: Qingjun Zhang
Abstract: We introduce E = T*H/(O+B), a dimensionless control parameter that predicts whether Mixture-of-Experts (MoE) models will develop a healthy expert ecology or collapse into dead experts. E combines four hyperparameters -- routing temperature T, routing entropy weight H, oracle weight O, and balance weight B -- into a single quantity. Through 12 controlled experiments (8 vision, 4 language) totaling over 11,000 training epochs, we establish that E >= 0.5 alone is sufficient to guarantee zero dead experts, removing the necessity for handcrafted load-balancing auxiliary losses. We validate this cross-modally on CIFAR-10, CIFAR-100, TinyImageNet-200, WikiText-2, and WikiText-103. Six additional findings emerge: (1) dead experts can resuscitate -- triggered by balance loss driving router re-exploration; (2) ortho toxicity is dataset-dependent, not universal; (3) task complexity shifts the critical E threshold; (4) model overfitting is decoupled from expert ecological health; (5) three-tier MoE spontaneously collapses into a two-tier functional structure; (6) ecological structure is temperature-invariant across a 50x range. We propose that E serves as a unified diagnostic for MoE training, analogous to the Reynolds number in fluid dynamics.
Authors: Murat Bilgehan Ertan, Xiaochen Zhu, Phuong Ha Nguyen, Marten van Dijk, Srinivas Devadas
Abstract: We introduce PACZero, a family of PAC-private zeroth-order mechanisms for fine-tuning large language models that delivers usable utility at $I(S^*; Y_{1:T})=0$. This privacy regime bounds the membership-inference attack (MIA) posterior success rate at the prior, an MIA-resistance level the DP framework matches only at $\varepsilon=0$ and infinite noise. All DP-ZO comparisons below are matched at the MIA posterior level. The key insight is that PAC Privacy charges mutual information only when the release depends on which candidate subset is the secret. Sign-quantizing subset-aggregated zeroth-order gradients creates frequent unanimity, steps at which every candidate subset agrees on the update direction; at these steps the released sign costs zero conditional mutual information. We propose two variants that span the privacy-utility trade-off: PACZero-MI (budgeted MI via exact calibration on the binary release) and PACZero-ZPL ($I=0$ via a uniform coin flip on disagreement steps). We evaluate on SST-2 and SQuAD with OPT-1.3B and OPT-6.7B in both LoRA and full-parameter tracks. On SST-2 OPT-1.3B full fine-tuning at $I=0$, PACZero-ZPL reaches ${88.99\pm0.91}$, within $2.1$pp of the non-private MeZO baseline ($91.1$ FT). No prior method produces usable utility in the high-privacy regime $\varepsilon<1$, and PACZero-ZPL obtains competitive SST-2 accuracy and nontrivial SQuAD F1 across OPT-1.3B and OPT-6.7B at $I=0$.
Authors: Abigail Victoria Gurin Schleifer, Moriah Ariely, Beata Beigman Klebanov, Asaf Salman, Giora Alexandron
Abstract: Automated short answer scoring (ASAS) is shifting from discriminative, fine-tuned models to large language models (LLMs) used in few-shot settings. This paradigm leverages LLMs broad world knowledge and ease of deployment, but limited task-specific data may reduce alignment on complex scoring tasks. In particular, its impact on scoring partially correct responses that require nuanced interpretation remains underexplored. We investigate the relationship between the degree of task-specific adaptation of different models and quality-conditioned scoring agreement. We compare three LLMs (GPT-5.2, GPT-4o, Claude Opus 4.5) in few-shot mode, a fine-tuned BERT-based encoder, and a human expert on two open-ended biology items, using several hundred student responses and ground truth scores provided by a biology education expert. The results show that human-human agreement is highest and stable across the full quality spectrum. All AI models perform well on fully correct and fully incorrect responses, but exhibit substantial degradation on mid-range responses. This mid-range degradation is conditioned on task-specific adaptation: It is most severe in few-shot LLMs with few examples and decreases as task-specific data increases, with fine-tuned encoder models performing best. This mid-range degradation may lead to inequitable evaluation of responses produced by students with developing understanding. Our findings highlight the importance of quality-conditioned fairness, with particular attention to mid-range responses.
Authors: Srinivas Kumar Ramdas, Jose Mathew, Ankit Singh Chauhan, Dinesh Rajkumar, Aravind Manoj, Mohit Goel
Abstract: Accurate truck-to-shipment matching using GPS data is foundational for full truckload supply chain visibility, enabling real-time tracking and accurate estimated time of arrival (ETA) predictions. However, missing or corrupted vehicle identifiers prevent traditional matching approaches, leaving shipments without visibility. This paper presents Intelligent Truck Matching (ITM) 2.0, a machine learning system that addresses this critical gap by formulating matching as a probabilistic ranking problem. Our approach leverages Uber H3 hexagonal spatial indexing to discretize GPS pings into route similarity features, combined with temporal information, then applies LightGBM gradient boosting with threshold-based post-processing. Through rigorous evaluation including offline model selection (SVM, XGBoost, LightGBM), comprehensive ablation studies, and production shadow testing, we demonstrate substantial gains over rule-based baselines. ITM 2.0 achieves 26 percentage point precision improvement in North America and 14 points in Europe, while doubling coverage. Deployed in production at Project44 handling full truckload shipments, the system demonstrates robustness to geocoding errors up to 1 km, multiple candidate trucks, and sparse pings.
Authors: Zhen Fang, Wenxuan Huang, Yu Zeng, Yiming Zhao, Shuang Chen, Kaituo Feng, Yunlong Lin, Lin Chen, Zehui Chen, Shaosheng Cao, Feng Zhao
Abstract: Existing Flow Matching (FM) text-to-image models suffer from two critical bottlenecks under multi-task alignment: the reward sparsity induced by scalar-valued rewards, and the gradient interference arising from jointly optimizing heterogeneous objectives, which together give rise to a 'seesaw effect' of competing metrics and pervasive reward hacking. Inspired by the success of On-Policy Distillation (OPD) in the large language model community, we propose Flow-OPD, the first unified post-training framework that integrates on-policy distillation into Flow Matching models. Flow-OPD adopts a two-stage alignment strategy: it first cultivates domain-specialized teacher models via single-reward GRPO fine-tuning, allowing each expert to reach its performance ceiling in isolation; it then establishes a robust initial policy through a Flow-based Cold-Start scheme and seamlessly consolidates heterogeneous expertise into a single student via a three-step orchestration of on-policy sampling, task-routing labeling, and dense trajectory-level supervision. We further introduce Manifold Anchor Regularization (MAR), which leverages a task-agnostic teacher to provide full-data supervision that anchors generation to a high-quality manifold, effectively mitigating the aesthetic degradation commonly observed in purely RL-driven alignment. Built upon Stable Diffusion 3.5 Medium, Flow-OPD raises the GenEval score from 63 to 92 and the OCR accuracy from 59 to 94, yielding an overall improvement of roughly 10 points over vanilla GRPO, while preserving image fidelity and human-preference alignment and exhibiting an emergent 'teacher-surpassing' effect. These results establish Flow-OPD as a scalable alignment paradigm for building generalist text-to-image models. The codes and weights will be released in: https://github.com/CostaliyA/Flow-OPD .
Authors: Karim Othman, Jonas Petersen, Matei Ignuta-Ciuncanu, Camilla Mazzoleni, Federico Martelli, Alessandro Lombardi, Riccardo Maggioni, Philipp Petersen
Abstract: We introduce the first universal pretraining corpus for industrial time-series data: FactoryNet. 51M datapoints across 23k end-to-end task executions (13.3k real, 9.8k synthetic) on six embodiments, unified by a shared schema that enables robust zero-shot cross-embodiment transfer and highly parameter-efficient anomaly detection. We introduce a novel schema: Setpoint, Effort, Feedback, Context (S-E-F-C) underlying the whole pipeline that maps any actuated system into a common representational frame. The corpus spans 27 annotated anomaly types alongside healthy baselines and counterfactual pairs across robotic manipulation and machining domains. Cross-embodiment transfer experiments yield positive results: under bias-aware metrics our model demonstrates fair cross-embodiment transfer capabilities on the evaluated source-target pair, while 24 schema-aligned signals achieves competitive anomaly detection performance compared to high-dimensional baselines. We release FactoryNet as a growing, multi-embodiment dataset to drive progress toward industrial foundation models.
Authors: Ruiying Peng, Mengyu Yang, Jing Lei, Xiaohui Li, Xueyu Wu, Xinlei Chen
Abstract: Supervised Fine-Tuning (SFT) is widely used for task-specific adaptation, yet recent work shows it systematically undermines reasoning generalization. We argue the root cause is not memorization itself, but its target: vanilla SFT drives models to exploit and memorize spurious surface correlations in problem-solution pairs, leaving them brittle to superficial input variations. To address this, we propose Theorem-SFT, which reorients supervision toward explicit theorem application by teaching models how rules are invoked rather than what answers look like. Theorem-SFT yields consistent gains across benchmarks and model families: +8.8% on MATH (LLaMA3.2-3B-Instruct) and +20.27% on GeoQA (Qwen2.5-VL-7B-Instruct) without modality-specific re-training. Fine-tuning MLP layers alone matches full-layers performance, implicating feed-forward components as the primary locus of reasoning rules. Our findings reframe the debate: Generalization failures stem not from memorization as a mechanism, but from memorizing the wrong inductive targets.
Authors: George Panagopoulos
Abstract: Estimating heterogeneous treatment effects with machine learning has attracted substantial attention in both academic research and industrial practice. However, the two communities often evaluate models under markedly different conditions. Methodological work typically relies on semi-simulated benchmarks and metrics that require counterfactual outcomes, whereas real-world applications rely on observable metrics based on ranking or test outcomes. Despite the well-known gap between methodological progress and practical deployment, the relationship between these evaluation regimes has not been examined systematically. We conduct a large-scale empirical study of treatment effect evaluation across standard semi-simulated benchmark families and real-world datasets. Our benchmark covers meta-learners paired with multiple base learners, as well as specialized causal machine learning models. We evaluate these methods using observable metrics common in application-oriented literature, alongside counterfactual metrics commonly used in methods papers. Our results reveal two complementary gaps. First, counterfactual metrics do not reliably recover the estimators preferred by observable metrics, even on the same semi-simulated benchmarks. Second, rankings obtained on semi-simulated benchmarks do not transfer to real datasets. We further find that simple meta-learners with strong base models are consistently competitive, in contrast to specialized causal models. Overall, our findings suggest that progress in treatment effect estimation research should not be assessed solely through counterfactual metrics and semi-simulated benchmarks, but it would benefit from incorporating observable metrics and real-data validation.
Authors: Suvi De Silva, Alfreds Lapkovskis, Alaa Saleh, Sasu Tarkoma, Praveen Kumar Donta
Abstract: Grey failures in the computing continuum produce ambiguous overlapping symptoms that existing approaches fail to diagnose reliably, either due to a lack of causal awareness or acting under high epistemic uncertainty, risking destructive interventions. This paper presents an uncertainty-aware resilience micro-agent for causal observability (AURORA), a lightweight framework for diagnosing and mitigating grey failures in edge-tier environments. The framework employs parallel micro-agents that integrate the free-energy principle, causal do-calculus, and localized causal state-graphs to support counterfactual root-cause analysis within each fault's Markov blanket. Restricting inference to causally relevant variables reduces computational overhead while preserving diagnostic fidelity. AURORA further introduces a dual-gated execution mechanism that authorizes remediation only when causal confidence is high and predicted epistemic uncertainty is bounded; otherwise, it abstains from local intervention and escalates the diagnostic payload to the fog tier. Our experiments demonstrate that AURORA outperforms baselines, achieving a 0% destructive action rate, while maintaining 62.0% repair accuracy and a 3ms mean time to repair.
Authors: Mengqi He, Xinyu Tian, Xin Shen, Shu Zou, Jinhong Ni, Zhaoyuan Yang, Weikang Li, Xuesong Li, Jing Zhang
Abstract: Recent studies show that gradient-based universal image jailbreaks on vision-language models (VLMs) exhibit little or no cross-model transferability, casting doubt on the feasibility of transferable multimodal jailbreaks. We revisit this conclusion under a strictly untargeted threat model without enforcing a fixed prefix or response pattern. Our preliminary experiment reveals that refusal behavior concentrates at high-entropy tokens during autoregressive decoding, and non-refusal tokens already carry substantial probability mass among the top-ranked candidates before attack. Motivated by this finding, we propose Untargeted Jailbreak via Entropy Maximization(UJEM)-KL, a lightweight attack that maximizes entropy at these decision tokens to flip refusal outcomes, while stabilizing the remaining low-entropy positions to preserve output quality. Across three VLMs and two safety benchmarks, UJEM-KL achieves competitive white-box attack success rates and consistently improves transferability, while remaining effective under representative defenses. Our experimental results indicate that the limited transferability primarily stems from overly constrained optimization objectives.
Authors: Zhenxin Ai, Haiyun He
Abstract: Watermarking for large language models (LLMs) is a promising approach for detecting LLM-generated text and enabling responsible deployment. However, existing watermarking methods are often vulnerable to semantic-invariant attacks, such as paraphrasing. We propose PASA, a principled, robust, and distortion-free watermarking algorithm that embeds and detects a watermark at the semantic level. PASA operates on semantic clusters in a latent embedding space and constructs a distributional dependency between token and auxiliary sequences via shared randomness synchronized by a secret key and semantic history. This design is grounded in our theoretical framework that characterizes a jointly optimal embedding-detection pair, achieving the fundamental trade-offs among detection accuracy, robustness, and distortion. Evaluations across multiple LLMs and semantic-invariant attacks demonstrate that PASA remains robust even under strong paraphrasing attacks while preserving high text quality, outperforming standard vocabulary-space baselines. Ablation studies further validate the effectiveness of our hyperparameter choices. Webpage: https://ai-kunkun.github.io/PASA_page/.
Authors: Haoyu Huang, Boyu Liu, Linlin Yang, Yanjing Li, Yuguang Yang, Xuhui Liu, Canyu Chen, Zhongqian Fu, Baochang Zhang
Abstract: The training of Binary Neural Networks (BNNs) is fundamentally based on gradient approximation for non-differentiable binarization operations (e.g., sign function). However, prevailing methods including the Straight-Through Estimator (STE) and its improved variants, rely on hand-crafted designs that suffer from gradient mismatch problem and information loss induced by fixed-range gradient clipping. To address this, we propose SURrogate GradiEnt Adaptation (SURGE), a novel learnable gradient compensation framework with theoretical grounding. SURGE mitigates gradient mismatch through auxiliary backpropagation. Specifically, we design a Dual-Path Gradient Compensator (DPGC) that constructs a parallel full-precision auxiliary branch for each binarized layer, decoupling gradient flow via output decomposition during backpropagation. DPGC enables bias-reduced gradient estimation by leveraging the full-precision branch to estimate components beyond STE's first-order approximation. To further enhance training stability, we introduce an Adaptive Gradient Scaler (AGS) based on an optimal scale factor to dynamically balance inter-branch gradient contributions via norm-based scaling. Experiments on image classification, object detection, and language understanding tasks demonstrate that SURGE performs best over state-of-the-art methods.
Authors: Jonas Petersen, Gian-Alessandro Lombardi, Riccardo Maggioni, Camilla Mazzoleni, Federico Martelli, Philipp Petersen
Abstract: Critical events in multivariate time series, from turbine failures to cardiac arrhythmias, demand accurate prediction, yet labeled data is scarce because such events are rare and costly to annotate. We introduce HEPA (Horizon-conditioned Event Predictive Architecture), built on two key principles. First, a causal Transformer encoder is pretrained via a Joint-Embedding Predictive Architecture (JEPA): a horizon-conditioned predictor learns to forecast future representations rather than future values, forcing the encoder to capture predictable temporal dynamics from unlabeled data alone. Second, we freeze the encoder and finetune only the predictor toward the target event, producing a monotonic survival cumulative distribution function (CDF) over horizons. With fixed architecture and optimiser hyperparameters across all benchmarks, HEPA handles water contamination, cyberattack detection, volatility regimes, and eight further event types across 11 domains, exceeding leading time-series architectures including PatchTST, iTransformer, MAE, and Chronos-2 on at least 10 of 14 benchmarks, with an order of magnitude fewer tuned parameters and, on lifecycle datasets, an order of magnitude less labeled data.
Authors: Yanting Miao, Yutao Sun, Dexin Wang, Mengyu Zhou, Pascal Poupart, Lei Lv, Qi Zhao, Li Wang, Hao Li, Xiaoxi Jiang, Guanjun Jiang
Abstract: Visual latent reasoning lets a multimodal large language model (MLLM) create intermediate visual evidence as continuous tokens, avoiding external tools or image generators. However, existing methods usually follow an output-as-input latent paradigm and yield unstable gains. We identify evidence for a feature-space mismatch that can contribute to this instability: dominant visual-latent models build on pre-norm MLLMs and reuse decoder hidden states as predicted latent inputs, even though these states occupy a substantially different norm regime from the input embeddings the model was trained to consume (Xie et al., 2025; Li et al., 2026; Team et al., 2026). This mismatch can make direct latent feedback unreliable. Motivated by this diagnosis, we propose GAP, a Granular Alignment Paradigm for visual latent modeling. GAP aligns visual latent reasoning at three levels: feature-level alignment maps decoder outputs into input-compatible visual latents through a lightweight PCA-aligned latent head; context-level alignment grounds latent targets with inspectable auxiliary visual supervision; and capacity-guided alignment assigns latent supervision selectively to examples where the base MLLM struggles. On Qwen2.5-VL 7B, the resulting model achieves the best mean aggregate perception and reasoning performance among our supervised variants. Inference-time intervention probing further suggests that generated latents provide task-relevant visual signal beyond merely adding token slots.
Authors: Davi Bastos Costa, Renato Vicente
Abstract: Fine-tuning large language models on narrow data with harmful content produces broadly misaligned behavior on unrelated prompts, a phenomenon known as emergent misalignment. We propose that emergent misalignment involves persona-model collapse: deterioration of the model's internal capacity to simulate, differentiate, and maintain consistent characters. We test this hypothesis behaviorally using two metrics: moral susceptibility (S) and moral robustness (R), computed from the across- and within-persona variability of models' Moral Foundations Questionnaire responses under persona role-play. These metrics formalize the model's ability to differentiate characters (S) and its consistency when simulating a given one (R). We evaluate four frontier models (DeepSeek-V3.1, GPT-4.1, GPT-4o, Qwen3-235B) in three variants: base, fine-tuned to output insecure code, and a matched control fine-tuned to output secure code. Across the four models, insecure fine-tuning produces an average $55\%$ increase in S, pushing all four insecure variants beyond the band observed across 13 frontier models benchmarked in prior work -- with GPT-4o reaching more than twice the band's upper end -- signaling dysregulated differentiation. It also causes an average $65\%$ decrease in R, equivalent to a $304\%$ increase in 1/R. By contrast, the matched secure control preserves S near the base and induces only a partial R loss, showing that these effects are largely misalignment-specific. Complementing these metric shifts, insecure variants' unconditioned responses converge toward saturation near the scale ceiling, departing markedly from both base models' structured responses and those elicited when base models role-play toxic personas. Taken together, these metrics provide a sensitive diagnostic for emergent misalignment and serve as behavioral evidence that it involves persona-model collapse.
Authors: Siyuan Liu, Tinghong Chen, Xinghan Li, Yifei Wang, Jingzhao Zhang
Abstract: Data selection during supervised fine-tuning (SFT) can critically change the behavior of large language models (LLMs). Although existing work has studied the effect of selecting data based on heuristics such as perplexity, difficulty, or length, the reported findings are often inconsistent or context-dependent. In this work, we systematically study the role of data difficulty in fine-tuning from both empirical and theoretical perspectives, and find that there is no universally optimal difficulty level; rather, its effectiveness depends on the dataset size. We show that for a fixed data budget, there exists an optimal data difficulty for SFT, and that this optimal difficulty shifts toward harder data as the data budget increases. To explain this phenomenon, we conduct controlled synthetic experiments that reveal a simple underlying mechanism: the interplay between the (in-distribution) generalization gap and the extrapolation gap. We further support this mechanism through a theoretical analysis using PAC-Bayesian generalization bounds. Overall, our results clarify how data size and difficulty jointly affect the trade-off between generalization and extrapolation in SFT, providing guidance for difficulty-based data selection under certain model and data conditions.
Authors: Chen Liang, Xiatao Sun, Qian Wang, Daniel Rakita
Abstract: Zeroth-Order (ZO) optimization is pivotal for scenarios where backpropagation is unavailable, such as memory-constrained on-device learning and black-box optimization. However, existing methods face a stark trade-off: they are either sample-inefficient (e.g., standard finite differences) or suffer from high variance due to randomized estimation (e.g., random subspace methods). In this work, we propose Coherent Coordinate Descent (CoCD), a deterministic, sample-efficient, and budget-aware ZO optimizer. Theoretically, we formalize the notion of gradient coherence and demonstrate that CoCD is equivalent to Block Cyclic Coordinate Descent (BCCD) with ``warm starts,'' effectively converting historical (stale) gradients from a liability into a computational asset. This mechanism enables $O(1)$ query complexity per step while maintaining global descent directions. Furthermore, we derive error bounds revealing a counter-intuitive insight: larger finite-difference step sizes can induce an implicit smoothing effect on the optimization landscape by reducing the effective smoothness constant, thereby improving convergence stability. Experiments on MLP, CNN, and ResNet architectures (up to 270k parameters) demonstrate that CoCD significantly outperforms BCCD in terms of sample efficiency and convergence loss/accuracy, and exhibits superior stability over randomized ZO methods. Our results suggest that deterministic, structure-aware updates offer a superior alternative to randomization for lightweight ZO optimization.
Authors: Itay Zloczower, Eyal Lenga, Gilad Gressel, Yisroel Mirsky
Abstract: Model providers increasingly release open weights or allow users to fine-tune foundation models through APIs. Although these models are safety-aligned before release, their safeguards can often be removed by fine-tuning on harmful data. Recent defenses aim to make models robust to such malicious fine-tuning, but they are largely evaluated only against fixed attacks that do not account for the defense. We show that these robustness claims are incomplete. Surveying 15 recent defenses, we identify several defense mechanisms and show that they share a single weakness: they obscure or misdirect the path to harmful behavior without removing the behavior itself. We then develop a unified adaptive attack that breaks defenses across all defense mechanisms. Our results show that current approaches do not provide robust security; they mainly stop the attacks they were designed against. We hope that our unified adaptive adversary for this domain will help future researchers and practitioners stress-test new defenses before deployment.
Authors: Sukju Oh, Sukkyu Sun
Abstract: Online surgical phase recognition (SPR) underpins context-aware operating-room systems and requires committing to a prediction at every frame from past context alone. Surgical video poses three demands that natural-video recognizers do not jointly address: procedures span tens of thousands of frames, time flows non-uniformly as long routine stretches are punctuated by brief phase-defining transitions, and the visual domain is narrow so backbone features are strongly correlated across channels. Existing recognizers either let per-frame cost grow with elapsed length, or hold cost bounded but advance state at a uniform rate with channel-independent dynamics, leaving the latter two demands unaddressed. We present SurgicalMamba, a causal SPR model built on Mamba2's structured state-space duality (SSD) that holds per-frame cost at O(d). It introduces three SSD-compatible components that jointly address these demands: a dual-path SSD block that separates long- and short-term regimes at the level of recurrent state; intensity-modulated stepping, a continuous-time time-warp that adapts the slow path's effective rate to phase-relevant information; and state regramming, a per-chunk Cayley rotation that opens cross-channel mixing in the otherwise axis-aligned SSM recurrence. The learned rotation planes inherit a phase-aligned structure without any direct supervision, offering an interpretable internal signature of surgical workflow. Across seven public SPR benchmarks, SurgicalMamba reaches state-of-the-art accuracy and phase-level Jaccard under strict online evaluation: 94.6%/82.7% on Cholec80 (+0.7 pp/+2.2 pp over the strongest prior) and 89.5%/68.9% on AutoLaparo (+1.7 pp/+2.0 pp), at 238.74 fps on a single GPU. Ablations isolate the contribution of each component. The code is publicly available at https://github.com/sukjuoh/Surgical-Mamba.
Authors: Amirhossein Yousefiramandi
Abstract: In the coded caching, the server uses the cached information at the users to serve multiple users in parallel with a single coded multi-casting message or packet, that is, a merged packet, and thus mitigates the peak network congestion. In order to deliver the timely messages to the users in the deadline-driven applications like the video streaming, we must determine online the messages to be merged for the delivery, as there is a time limit for each request. It is important to note that while the merging aids the current coded multi-casting packet, it could harm the future deliveries. Our solution employs the deep reinforcement learning to view the coded multi-casting delivery as a masked action-discrete state control problem, and our policy network, trained via the proximal policy optimization, performs better than SACM++. On the uniform-demand benchmark, our policy network reduces the broadcast-packet expiration ratio $\rho$ by $40.9\%$ ($0.208$ vs.\ $0.352$) with respect to the best coded multi-casting baseline (SACM++), while also attaining the best broadcast-efficiency score $\sigma$ across the Track~A battery among the coded multi-casting methods. One noteworthy phenomenon here is that, for the applications with stricter deadlines, the merging becomes selective instead of aggressive, since the policy network selectively merges at approximately $31.8\%$ of the chances, even though the same observation holds across the variations within the same simulator family. The focus of our design is on the efficient pairwise XOR merging, where the higher-order ($K{\ge}3$) coding can be considered as a natural generalization left for future work.
Authors: Fei Ding, Yongkang Zhang, Youwei Wang, Zijian Zeng
Abstract: Reinforcement learning for multi-step reasoning with large language models (LLMs) typically relies on sparse terminal rewards, which creates a poorly conditioned credit-assignment problem: the final feedback is propagated uniformly across all intermediate decisions. This leads to high gradient variance, unstable training, and many ineffective updates, ultimately limiting sustained model improvement. We propose a counterfactual-comparison framework for credit assignment. For each input, the framework samples multiple reasoning trajectories and treats their differences as implicit approximations to alternative decisions. This yields an implicit process-level advantage estimator that converts sparse terminal rewards into step-sensitive learning signals. Building on this framework, we introduce Implicit Behavior Policy Optimization (IBPO), which substantially improves training stability and the performance ceiling on mathematical and code-reasoning benchmarks. Our results point to a promising direction for unlocking the reasoning potential of LLMs.
Authors: Entang Wang, Yiwei Wang, Aleksandra Bakalova, Michael Hahn
Abstract: In-context learning (ICL) excels at new tasks from minimal examples, yet we still lack a mechanistic explanation of how few-shot prompts shape a model's function vector (FV)--a causal activation direction that drives task behavior on the ICL query. Across tasks and models, an $n$-shot FV is well-approximated by a linear combination of example-level sub-FVs, suggesting additive and composable contributions from individual demonstrations. Beyond additivity, we show that models contextualize individual examples' representations based on prior examples to adaptively reweight which demonstrations dominate the FV: attention shifts toward examples that are more informative and less ambiguous under the context. Finally, a causal decomposition separates Query-Key routing from Value updates, finding that contextualization's most consistent contributions to FV quality arise from Query-Key alignment--particularly in ambiguous settings--while Value-mediated effects are more heterogeneous. Together, these results unify additive superposition with context-dependent attention reweighting into a mechanistic, testable account of how few-shot prompts implement tasks.
Authors: YongKyung Oh
Abstract: State-of-the-Art (SOTA) claims pervade Artificial Intelligence (AI) and Machine Learning (ML) research. These claims rest on benchmark evaluations, where models are ranked by aggregate scores across tasks. Public benchmarks or leaderboards are the most visible instance, but the same structure appears in paper tables throughout the literature. However, such minimal evidence often cannot support these strong claims. We identify a widespread claim-evidence gap in AI benchmarking. Claiming SOTA carries implicit assumptions beyond mean score superiority, suggesting that a model meaningfully outperforms alternatives across most tasks. However, a marginal improvement in the mean score merely indicates a top average rank rather than true superiority. Analyzing ten cross-domain benchmarks from public leaderboards, we found that in more than half of top-model comparisons, at least one commonly assumed property of superiority does not hold. These properties include meaningful effect size, consistency across tasks, or robustness to dataset removal. Instead, aggregate gains are frequently driven by outlier datasets. This fragility persists even in benchmarks with many tasks. We argue that claim language should reflect the strength of the underlying evidence. This requires no additional experiments, only honest reporting of what results actually show, enabling more precise and interpretable comparisons across models.
Authors: Fan Zhang, Shijun Chen, Hua Wang
Abstract: Mainstream methods for multivariate time-series forecasting largely follow the Direct-Mapping paradigm. They learn a unified mapping from history to the future in the observation space to fit value-level dependencies. However, real-world systems often undergo distribution shifts and regime changes. In such cases, a unified mapping can exhibit response lag around turning points, causing error accumulation within the switching window and reducing forecasting reliability. To address this issue, we propose L-Drive, a change-aware forecasting framework. L-Drive introduces a Latent-Context, to explicitly characterize high-level dynamics evolving over time, and uses gating to modulate increment representations. This provides more timely change cues and improves adaptation to changing segments. In addition, it incorporates patch-shared relative positional basis functions to strengthen intra-segment structural modeling and reduce overfitting caused by absolute-position memorization. Extensive experiments validate the effectiveness of L-Drive and show a better overall trade-off between forecasting accuracy and computational efficiency.
Authors: Zhensheng Wang, Wenmian Yang, Qingtai Wu, Lequan Ma, Yiquan Zhang, Weijia Jia
Abstract: Quantitative backtesting is essential for evaluating trading strategies but remains hampered by high technical barriers and limited scalability. While Large Language Models (LLMs) offer a transformative path to automate this complex, interdisciplinary workflow through advanced code generation, tool usage, and agentic planning, the practical realization is significantly challenged by the current lack of a large-scale benchmark dedicated to automated quantitative backtesting, which hinders progress in this field. To bridge this critical gap, we introduce BacktestBench, the first large-scale benchmark for automated quantitative backtesting. Built from over 6 million real market records, it comprises 18,246 meticulously annotated question-answering pairs across four task categories: metrics calculation, ticker selection, strategy selection, and parameter confirmation. We also propose AutoBacktest, a robust multi-agent baseline that translates natural language strategies into reproducible backtests by coordinating a Summarizer for semantic factor extraction, a Retriever for validated SQL generation, and a Coder for Python backtesting implementation. Our evaluation on 23 mainstream LLMs, complemented by targeted ablations, identifies key factors that influence end-to-end performance and highlights the importance of grounded verification and standardized indicator representations.
Authors: Lei Zhao, Abhay Bhaskar, Edgar Dobriban
Abstract: AI agents such as OpenClaw are increasingly deployed in local workflows with access to external tools. This creates indirect prompt-injection (IPI) risk: an agent may execute harmful instructions embedded in untrusted inputs such as email, downloaded files, webpages, repositories, or group-chat messages. Existing evaluations are often small, purely simulated, or focused on a narrow set of channels. We introduce LivePI (Live Prompt Injection), a structured benchmark for IPI risk in a production-like but test-controlled environment. LivePI covers seven input surfaces, twelve attack/rendering families, and five malicious goals, including protected-information exfiltration, unauthorized security-control changes, unsafe code retrieval or execution, inbox-summary exfiltration, and cryptocurrency transfer. We run LivePI on a real virtual machine with live but test-controlled email, chat, web, local-file, repository, and wallet interfaces. Across GPT-5.3-Codex, Claude Opus 4.6, Gemini 3.1 Pro, Kimi K2.5, and GLM-5, total attack success rates range from 10.7% to 29.6%. Group-chat injection is uniformly successful across the evaluated backbones in our deployment, and repository-link attacks produce high-severity failures despite a small denominator. We also evaluate a two-layer defense consisting of prompt-level filtering and pre-execution tool-call authorization. In the GPT-5.3-Codex setting, the defense intercepts all tested malicious-goal completions in LivePI before execution while preserving benign utility on PinchBench-derived workloads.
Authors: Zegu Zhang, Jianhua Peng, Jian Zhang
Abstract: We study exact constant collapse in variational autoencoders: the deterministic encoder mean becomes independent of the input. The prior remains the standard Gaussian. Before VAE training, we select a fixed teacher posterior from a GMM-based view of the data and attach a fixed latent-only simplex witness to the encoder mean. This construction yields two linked objects. The first is a certificate: if the witness prediction improves on the best constant predictor of the teacher, the encoder mean cannot be input-independent constant. The second is a local escape direction: on the collapsed manifold, the teacher residual gives a sample-dependent descent direction for the alignment loss. For any full-support teacher posterior, the same geometry also gives a closed-form latent code with zero teacher-witness alignment error. Its scaled versions trace a margin-energy path from the constant predictor to the exact teacher code, which quantifies non-collapse inside the protected witness subspace. We instantiate the method on MNIST, CIFAR-10, and CIFAR-100. With searched unsupervised PCA-GMM teachers, vanilla VAEs fail the teacher-witness certificate in all five seeds on CIFAR-10 and CIFAR-100, while RST variants pass in all five seeds. Under collapse-stress settings with \(\beta_{\mathrm{KL}}\in\{2,4,8\}\), vanilla VAE again fails in all seeds, whereas RST-alpha-prefit remains certificate-positive. Escape trajectories on both natural-image datasets increase the witness margin from a low-margin initialization and exhibit nonzero teacher-induced gradient norms. The analysis is confined to exact constant collapse of the encoder mean; generation quality, decoder use, and other collapse modes remain separate questions.
Authors: Luis Balderas, Jos\'e Alberto Rodr\'iguez, Miguel Lastra, Antonio Arauzo-Azofra, Jos\'e M. Ben\'itez
Abstract: Time Series Foundation Models (TSFMs) have demonstrated notable success in general-purpose forecasting tasks; however, their adaptation to specialized classification problems remains constrained by the computational bottleneck of standard attention and the systematic omission of classical statistical knowledge. This technical report introduces KairosHope, a next-generation TSFM designed to reconcile massive generalization with analytical precision in classification tasks. The core of the proposal is the HOPE block, an architecture that replaces quadratic attention with a dual-memory system: Titans modules for dynamic short-term retention and a Continuum Memory System (CMS) for the abstraction of long-term historical context. To enrich the inductive bias, a Hybrid Decision Head is introduced, which fuses deep latent representations with deterministic statistical features extracted via tsfeatures package. KairosHope undergoes self-supervised pre-training on the massive Monash archive, combining Masked Time Series Modeling (MTSM) and contrastive learning (InfoNCE). Its subsequent adaptation to the UCR benchmark datasets is conducted through a rigorous Linear Probing and Full Fine-Tuning (LP-FT) protocol to prevent catastrophic forgetting. Empirical results demonstrate superior performance in domains characterized by strict temporal causality such as HAR or Sensor data. Consequently, KairosHope establishes a robust and efficient framework for the adaptation of foundation models to time series analysis.
Authors: Yining Hong, Jiageng Liu, Han Yin, Manling Li, Leonidas Guibas, Li Fei-Fei, Jiajun Wu, Yejin Choi
Abstract: Spatial intelligence unfolds through a perception-action loop: agents act to acquire observations, and reason about how observations vary as a function of action. Rather than passively processing what is seen, they actively uncover what is unseen - occluded structure, dynamics, containment, and functionality that cannot be resolved from passive sensing alone. We move beyond prior formulations of spatial intelligence that assume oracle observations by recasting the observer as an actor. We introduce ESI-BENCH, a comprehensive benchmark for embodied spatial intelligence spanning 10 task categories and 29 subcategories built on OmniGibson, grounded in Spelke's core knowledge systems. Agents must decide what abilities to deploy - perception, locomotion, and manipulation - and how to sequence them to actively accumulate task-relevant evidence. We conduct extensive experiments on state-of-the-art MLLMs and find that active exploration substantially outperforms passive counterparts, with agents spontaneously discovering emergent spatial strategies without explicit instructions, while random multi-view often adds noise rather than signal despite consuming far more images. Most failures stem not from weak perception but from action blindness: poor action choices lead to poor observations, which in turn drive cascading errors. While explicit 3D grounding stabilizes reasoning on depth-sensitive tasks, imperfect 3D representation proves more harmful than 2D baselines by distorting spatial relations. Human studies further reveal that unlike humans who seek falsifying viewpoints and revise beliefs under contradiction, models commit prematurely with high confidence regardless of evidence quality, exposing a metacognitive gap that neither better perception nor more embodied interaction alone can close.
Authors: Rao Fu, Zixuan Yang, Jiankun Zhang, Jing Ma, Hechang Chen, Yu Li, Yi Chang
Abstract: Scaling model performance typically requires increasing model size. Looped Transformer offers a compelling alternative by iteratively reusing the same Transformer blocks, trading additional computation for improved performance without increasing parameter count or context length. Because the number of loop iterations can be adjusted at inference, it also provides a natural mechanism for balancing performance and test-time compute. However, Looped Transformer still suffers from training instability when the number of loop iterations increases. Our analysis reveals that this instability stems from two sources: gradient oscillation and residual explosion. To address these two problems, we propose the Fully Looped Transformer, which introduces two parameter-free modifications: (1) Fully Looped Architecture, which distributes inter-loop signals across all layers to mitigate residual explosion; (2) Attention Injection, which reuses the existing attention block to suppress gradient oscillation. These modifications stabilize training dynamics, enabling the Fully Looped Transformer to be trained stably up to 12 loop iterations, whereas other baseline looped models collapse in this regime. In milder settings where Looped Transformer does not collapse, Fully Looped Transformer still improves average downstream-task performance by up to 13.2\%. Overall, our experiments demonstrate that Fully Looped Transformer improves training stability, enhances downstream performance, and provides preliminary adaptability under different test-time compute budgets by varying loop iterations at inference.
Authors: Adil Amin
Abstract: Scaling laws predict loss from compute but not how capabilities interact. We measure the coupling between reasoning and truthfulness across 63 base models from 16 families and find a regime change invisible to loss curves: below a family-dependent critical scale $N_c$, capabilities anticorrelate; above it, they cooperate. $N_c \approx 3.5$B parameters [2.9B, 13.4B] (bootstrap 95\% CI), but model size is not the only variable that determines phase. Architecture, data curation, and training recipe each shift $N_c$ independently: curated training eliminated the coupling dip between Qwen generations (0.025 $\to$ 0.830 at matched scale), Gemma-4 at 4B achieves coupling 0.871, characteristic of 13B+ standard-trained models, through distillation and architectural innovation, and Phi at 1B matches web-trained coupling at 10B through data curation alone. Width normalization eliminates the anticorrelation across all tested families, supporting an output-projection bottleneck. Internally, 38 of 40 models show zero competing attention heads. A sparse-regression ODE cross-predicts held-out Llama-2 at 5.6\% error. The diagnostic requires no model internals -- only public benchmark scores across a model family. The cooperative regime extends to the frontier ($r = +0.72$, 34 models, 10 labs). A proof-of-concept intervention confirms the bottleneck is exploitable: adding a single truth-direction vector at the identified layer corrects 60\% of misaligned outputs in the tax phase with zero retraining -- a surgical, per-inference correction that requires no weight modification. Code, data, an open-source steering CLI for any open-weight model, and an interactive dashboard for phase diagnosis are released: https://zehenlabs.com/cape/.
Authors: Adil Amin
Abstract: Leaderboards rank frontier models on independent axes but do not reveal whether capabilities reinforce or trade off across releases -- and at the frontier, this interaction is the more informative signal. We decompose paired SWE-bench and GPQA Diamond scores into a population coupling trend and per-release residual ($h$-field) that diagnoses capability emphasis from two public benchmark scores. Across 34 models from 10 labs (2024--2026), capabilities cooperate ($r = +0.72$, $p < 10^{-6}$), but cooperation varies systematically: per-lab coupling slopes span $5\times$ (Google $1.15$ vs. DeepSeek $0.23$), and labs pivot -- DeepSeek reversed from reasoning-rich to coding-first ($\Delta h = 15.9$~pp); Anthropic oscillates between coding excursions and recovery. The population regression serves as an isocline phase boundary: the same $\sqrt{(a/b)\cdot B_1}$ classifier that identifies the base-scale coupling transition [Amin, 2026] classifies frontier models and already detects mixed-phase behavior at the next transition (two models below the GPQA--IFEval isocline). The $h$-field is not just diagnostic -- it tells you what to change. Pretraining establishes coupling at $0.871$ while RLHF adds $0.081$ [Amin, 2026]: pretraining-level shifts are permanent (DeepSeek's four-release reversal persists), post-training shifts are reversible (Anthropic's three coding excursions each recover within one release), and inference compute alone shifts $h$ by $+7.8$~pp without retraining. Knowing which component dominates determines whether to retrain or wait. We provide a three-step diagnostic (locate, classify, predict), a per-lab measurement-priority table, and seven falsifiable predictions with timestamped criteria. Five post-cutoff releases fall within the 95\% prediction interval. Code, data, and an interactive dashboard: https://zehenlabs.com/cape/.
Authors: Gyubin Lee, Junwon Lee, Juhan Nam
Abstract: We investigate Counterfactual Video Foley Generation, which aims to adopt a sound-source identity that contradicts the visual evidence while remaining temporally synchronized to a silent video. Existing Video&Text-to-Audio (VT2A) models struggle with this, often remaining anchored to the visually implied sound source when video and text contents disagree. We present ConterFlow, an inference-time dual-phase sampling scheme for pretrained flow-matching VT2A models. Phase 1 builds a video-derived temporal structure while suppressing the visually implied source; Phase 2 drops video conditioning to focus entirely on shaping audio timbre toward the target prompt. ConterFlow substantially improves counterfactual Video Foley generation compared to naive negative prompting and state-of-the-art baselines. To evaluate replacement quality, we propose a metric leveraging a text-audio co-embedding space to measure both target-prompt evidence and residual visually implied source leakage. Video demonstrations and code are available at https://gyubin-lee.github.io/counterflow-demo/
Authors: Nikita Klimenko, Hesam Salehipour, Parham Eftekhar, Amir Khasahmadi, Ramon Elias Weber
Abstract: In this work, we propose HypergraphFormer, a novel and efficient approach to floor plan generation based on learning hypergraph representations with a large language model (LLM). The model is trained via supervised fine-tuning to generate a hypergraph-based textual representation that encodes spatial relationships and connectivity information within floor plans. We train and evaluate our approach on the RPLAN dataset, and further demonstrate its generalizability on a separate out-of-distribution dataset, which we release in this paper. Our method outperforms state-of-the-art techniques based on rasterized or vectorized representations across a diverse set of metrics. We also show improved data efficiency, particularly under distribution shift. The hypergraph formulation enables the generation of floor plans for arbitrary, irregular, user-specified boundaries by decoupling apartment footprints from their functional and geometric subdivisions. Furthermore, we show that the proposed methodology offers a high degree of editability, making it particularly well suited to design-oriented workflows supported by LLMs.
Authors: Lihong Lin, Haidong Kang
Abstract: Continual Model Merging (CMM) enables rapid customization of foundation models by sequentially incorporating task-adapted models without repeated retraining. However, existing merging rules usually update the deployed model through fixed algebraic or projection-based operations, providing limited control over how much previously accumulated knowledge should be retained relative to the incoming task model. This limitation leads to unstable retention and performance degradation in long task streams, and becomes more pronounced when tasks have heterogeneous utilities. We propose ODE-driven Merging (ODE-M), a controllable framework that formulates each continual merge as a trajectory in parameter space rather than a one-step endpoint update. Motivated by mode connectivity, ODE-M constructs a barrier-aware trajectory using a rectified time-dependent velocity field, where lightweight first-order feedback from a small calibration set suppresses loss-increasing motion while preserving progress toward the incoming model. The next merged model is then obtained by selecting an operating point along this trajectory through a utility-aware time schedule, providing an explicit mechanism for balancing retained historical knowledge and incoming task expertise. Extensive experiments on standard CMM benchmarks show that ODE-M consistently improves over strong continual merging baselines across CLIP ViT backbones, stream lengths, and heterogeneous task-utility settings.
Authors: Gueter Josmy Faure, Min-Hung Chen, Jia-Fong Yeh, Hung-Ting Su, Winston H. Hsu
Abstract: Vision-Language Models (VLMs) have demonstrated remarkable capabilities in general video understanding, yet they often struggle with the fine-grained comprehension crucial for real-world applications requiring nuanced interpretation of human actions and interactions. While some recent human-centric benchmarks evaluate aspects of model behaviour such as fairness/ethics, emotion perception, and broader human-centric metrics, they do not combine long-form videos, very dense QA coverage, and frame-level spatial/temporal grounding at scale. To bridge this gap, we introduce FineBench, a human-centric video question answering (VQA) benchmark specifically designed to assess fine-grained understanding. FineBench comprises 199,420 multiple-choice QA pairs densely annotated across 64 long-form videos (15 minutes each), focusing on detailed person movement, person interaction, and object manipulation, including compositional actions. Our extensive evaluation reveals that while proprietary models like GPT-5 achieve respectable performance, current open-source VLMs significantly underperform, struggling particularly with spatial reasoning in multi-person scenes and distinguishing subtle differences in human movements and interactions. To address these identified weaknesses, we propose FineAgent, a modular framework that enhances VLMs by leveraging a Localizer and a Descriptor. Experiments show that FineAgent consistently improves the performance of various open VLMs on FineBench. FineBench provides a rigorous testbed for future research into fine-grained human-centric video understanding, while FineAgent offers a practical approach to enhance such reasoning in current VLMs. Project page and code at https://joslefaure.github.io/assets/html/finebench.html.
URLs: https://joslefaure.github.io/assets/html/finebench.html.
Authors: Jiawei He, Jie Jia, Chenbo Liu, Chaoyi Xue, Yapeng Song, Xikai Yang, Dong Sun
Abstract: Existing benchmarks for LLM coding agents primarily evaluate final outcomes. While useful for measuring overall capability, these metrics provide limited visibility and often miss defects that arise during execution. We present ProcCtrlBench, a benchmark for execution-process evaluation in LLM coding agents. ProcCtrlBench organizes recurrent execution defects into a reusable ontology covering 11 defect types in 4 categories, and evaluates agent trajectories through standardized process evidence rather than final outcomes alone. To support comparison across heterogeneous agents, ProcCtrlBench standardizes raw logs into a unified trajectory representation and reports calibrated scorecards over process-level findings. In addition, ProcCtrlBench uses control preservation as a way to quantify execution-process quality, capturing whether execution remains interpretable, interruptible, correctable, reversible, and able to hand back authority when needed. We evaluate ProcCtrlBench on 200 cases sampled from three benchmarks: AndroidBench, TerminalBench, and SWE-bench-Verified. Results show that ProcCtrlBench can be instantiated with useful reliability, provides more stable semantics than direct thresholding, and reveals meaningful differences in execution quality that are often overlooked by conventional outcome-based evaluation.
Authors: Tianle Li, Xuyang Shen, Yan Ma, Rongxin Guo, Shaoxiang Chen, Jiacheng Chen, Haochen Wang, Hongyang Tang, Yucong Zhou, Yu Cheng
Abstract: Long-form image captioning exposes a reward granularity problem in RL: captions are judged as whole sequences, while the important errors occur at the level of individual visual claims. A good dense caption should be both faithful and informative, avoiding hallucination without omitting salient details. Yet pairwise preferences, reference-based metrics, and holistic scalar rewards compress these local errors into a single sequence-level signal, obscuring the tradeoff between factuality and coverage. We introduce ClaimDiff-RL, a framework that uses reference-conditioned atomic claim differences as the reward unit for caption RL. Given an image, an actor caption, and a reference caption, a multimodal judge enumerates visually grounded differences, verifies each difference against the image, assigns open-vocabulary error types and severity levels, and produces per-difference statistics for reward composition. This makes hallucinated claims and omitted salient facts separately measurable and tunable. Experiments show that holistic scalar rewards can reduce hallucination by increasing missing facts, while ClaimDiff-RL exposes this faithfulness and coverage tradeoff and enables more balanced operating points. On a 160-image human-labeled diagnostic benchmark, public captioning benchmarks, and VQA benchmarks, ClaimDiff-RL improves the hallucination--missing-fact balance, preserves general capability, and even surpasses Gemini-3-Pro-Preview on several fine-grained Capability dimensions such as object counting, spatial relations, and scene recognition. These results suggest that typed, verifiable claim differences are an effective reward unit for fine-grained and diagnosable caption RL.
Authors: Xingyu Lyu, Qianqian Xu, Zhiyong Yang, Peisong Wen, Qingming Huang
Abstract: Gated Linear Units (GLU) and their variants are widely adopted in modern open-source large language model architectures and consistently outperform their non-gated counterparts, yet the underlying reasons for this advantage remain unclear. In this work, we study GLU by analyzing two-layer networks in the neural tangent kernel (NTK) regime. Our analysis reveals that the GLU structure reshapes the NTK spectrum, leading to a smaller condition number and a more compact eigenvalue distribution. Building on this finding, we further analyze the resulting training dynamics and show how the reshaped spectrum leads to faster convergence of GLU models, including a characteristic loss-crossing phenomenon observed between GLU and non-GLU models. Finally, we empirically observe that GLU has limited impact in reducing the generalization gap on various models, including ViT and GPT-2, suggesting that its primary benefit lies in accelerating optimization rather than reducing the generalization gap. The code is available at: https://github.com/Zemdalk/GLU-NTK.
Authors: Junghyun Lee, Hyunseo Kim, Hanna Jang, Junhyug Noh
Abstract: Blended emotion recognition is challenging because emotions are often expressed as mixtures of subtle and overlapping multimodal cues rather than a single dominant signal. We propose a rank-aware multi-encoder framework that selectively combines complementary representations from diverse pre-extracted video and audio encoders. Our method projects heterogeneous encoder features into a shared latent space, estimates sample-wise encoder importance through an attention-based gating module, and fuses only the top-n most informative encoders. To better model blended emotions, we decouple prediction into presence and salience heads and align them through probability-level fusion. We further incorporate feature-level unsupervised domain adaptation without pseudo-labeling to improve robustness under distribution shift. Experiments on the BlEmoRE challenge show that the proposed framework outperforms strong individual encoders and na\"ive multi-encoder fusion baselines. Our final system ranked 2nd in the competition, supporting the effectiveness of rank-aware selective fusion for fine-grained blended emotion recognition.
Authors: Yue Zhou, Erxuan Wu, Yikang Sun, Hongjoo Lee, Yuan Bi, Huixiong Xu, Nassir Navab, Zhongliang Jiang
Abstract: Vision-Language Models (VLMs) have significantly advanced medical visual question answering, yet their performance in ultrasound remains suboptimal. In clinical practice, sonographers explicitly focus on lesion regions to formulate reports, though diagnostic interpretations sometimes vary due to inherent subjectivity. However, existing VLMs are not explicitly structured to interactively zoom into lesions prior to diagnosis; moreover, they typically treat annotations as unbiased ground truths, failing to account for their inherent subjectivity and ambiguity. In this paper, we propose a framework specifically designed to consider the sonographer's cognitive workflow. We first introduce a structured Zoom-then-Diagnose paradigm, which replicates the interactive search process to enable lesion-focused reasoning. Furthermore, within the Group Relative Policy Optimization (GRPO) framework, we introduce an uncertainty-aware reward derived from stochastic group-wise rollouts to estimate prediction consistency as a proxy for model confidence. Together, these two components encourage the model to reinforce accurate predictions on clear cases while remaining cautious under ambiguity. Experiments across liver, breast, and thyroid datasets show that our framework improves lesion localization by 39.3\%, demonstrating that our model has learned the ability to actively look closer and diagnose.
Authors: Weilong Guo, Yuchen Wang, Renping Zhou, Yunfeng Zhang, Rui Fang, Yuyang Pang, Wenda Xu, Gao Huang
Abstract: Vision-Language-Action (VLA) models have emerged as a promising paradigm for generalist robotic manipulation. A common design in current architectures maps language instructions and visual observations to actions in a single forward pass. While conceptually simple, this formulation entangles instruction comprehension, spatial scene understanding, and motor control within a single learning objective. As a result, the action expert must implicitly relearn cognitive and perceptual capabilities already present in the pretrained VLM, which can limit both learning efficiency and generalization. We introduce AVP (Action with Visual Primitives), an end-to-end architecture that implements this visual-primitive-centric interface: the VLM infers the next-stage target and emits visual-primitive tokens that condition a flow-matching action expert, with supervision derived from end-effector kinematics. Real-robot experiments on general pick-and-place tasks show that AVP improves the success rate by 27.61% over pi_0.5 and outperforms other recent methods, with consistent gains in data efficiency, spatial-compositional generalization, and object-level transfer.
Authors: Quang Duc Nguyen, Siyuan Liang, Yiming Li, Fushuo Huo, Dacheng Tao
Abstract: Time Series Forecasting (TSF) is highly vulnerable to backdoor attacks, yet effective defenses remain underexplored due to challenges arising from data entanglement and shifts in task formulation. To fill this gap, we conduct a systematic evaluation of thirteen representative backdoor defenses across the TSF life cycle and analyze their failure modes. Our results reveal two fundamental issues: (1) data entanglement induces channel-level signal dilution, rendering sample-filtering and trigger-synthesis defenses ineffective at localizing backdoors; and (2) task-formulation shift leads to training-loss degeneration, causing poisoned and clean windows to become indistinguishable at training stages. Based on these findings, we propose a training-time backdoor defense for TSF, termed TimeGuard. Our method adopts channel-wise pool training as the core paradigm and initializes a high-confidence pool using time-aware criteria to mitigate signal dilution. Moreover, we introduce distance-regularized loss selection to progressively expand the reliable pool during training and ease loss degeneration. Extensive experiments across multiple datasets, forecasting architectures, and TSF backdoor attacks demonstrate that TimeGuard substantially improves robustness, boosting $\mathrm{MAE}_\mathrm{P}$ by $1.96\times$ over the leading baseline, while preserving clean performance within 5% $\mathrm{MAE}_\mathrm{C}$.
Authors: Ting Liu
Abstract: Skills have become a practical packaging mechanism for agent instructions, workflows, scripts, and reference materials. In enterprise settings, however, a skill often needs to express more than task guidance: goals, input boundaries, permissions, human approval points, evidence requirements, output contracts, quality criteria, verification steps, and handoff rules. This paper proposes contractual skills, a GovernSpec-inspired design framework for organizing SKILL.md files as readable task contracts while preserving lightweight skill discovery and progressive loading. The framework clarifies the boundary between contractual skills, GovernSpec YAML contracts, Model Context Protocol (MCP) surfaces, tool adapters, runtime guardrails, tracing, and evaluation systems. We evaluate the framework with three offline empirical studies. The first text-generation experiment covers three enterprise skills, fifteen synthetic tasks, four instruction conditions, and eight generation models, producing 960 outputs and 1680 cross-judge score records. The second study is a public-skill A/B expansion: eight public skills are compared with contractual rewrites across forty-eight synthetic tasks, six generation models, two repeats, 1152 outputs, and two complete judge files. In this setting, contractual skills raise mean quality from 4.692 to 4.914 and reduce critical-error rate from 0.083 to 0.013. The third study is an offline tool-calling challenge with eight models and 192 simulated tool-call records. The results suggest that contractual skills are best understood as a governance layer that makes task intent, boundaries, and acceptance criteria explicit, not as a standalone safety mechanism.
Authors: Baiyu Chen, Zechen Li, Wilson Wongso, Lihuan Li, Xiachong Lin, Hao Xue, Benjamin Tag, Flora Salim
Abstract: As wearable and mobile devices become increasingly embedded in daily life, they offer a practical way to continuously sense human motion in the wild. But inertial signals are highly dependent on the sensing setup, including body location, mounting position, sensor orientation, device hardware, and sampling protocol. This setup dependence makes it difficult to learn motion representations that transfer across devices and datasets, and limits the broader use of wearable IMUs beyond closed-set recognition. We introduce AnyMo, a geometry-aware framework for setup-agnostic human motion modeling. AnyMo uses physics-grounded IMU simulation over dense body-surface placements to generate diverse and plausible synthetic signals, pre-trains a graph encoder from paired synthetic placement views and masked partial observations, tokenizes multi-position IMU into full-body motion tokens, and aligns these tokens with an LLM for motion-language understanding. We evaluate AnyMo on three complementary tasks: zero-shot activity recognition across 14 unseen downstream datasets, cross-modal retrieval, and wearable IMU motion captioning, where it improves average Accuracy/F1/R@2 by 11.7\%/11.6\%/22.6\% on HAR, increases zero-shot IMU-to-text and text-to-IMU retrieval MRR by 15.9\% and 28.6\%, respectively, and improves zero-shot captioning BERT-F1 by 18.8\%. These results support AnyMo as a generalist model for wearable motion understanding in the wild. Project page: https://baiyuchen.com/project/AnyMo.
Authors: Hippolyte Pilchen, Romain Fabre, Franck Signe Talla, Patrick Perez, Edouard Grave
Abstract: Large language models (LLMs) are typically trained on shuffled corpora, yielding models whose knowledge is frozen at train time and whose temporal grounding remains poorly understood. In this work, we study the impact of pre-training dynamics on the acquisition of time-sensitive factual knowledge, focusing specifically on data ordering. Our main contributions are twofold. First, we introduce a comprehensive benchmark of over 7,000 temporally grounded questions and an evaluation protocol that enables analysis of whether models correctly associate facts with their corresponding time periods. Second, we pretrain 6B-parameter models on temporally ordered Common Crawl snapshots and compare them against standard shuffled pre-training. Our results show that sequentially trained models match shuffled baselines on general language understanding and common knowledge while consistently exhibiting more up-to-date and temporally precise knowledge. Temporally ordered pre-training yields improved factual freshness, while shuffled pre-training peaks on older data, possibly due to increased factual repetition. These findings, along with the release of our code at https://github.com/kyutai-labs/kairos , checkpoints, and datasets at https://huggingface.co/collections/kyutai/kairos provide a foundation for future research on continual learning for LLMs.
URLs: https://github.com/kyutai-labs/kairos, https://huggingface.co/collections/kyutai/kairos
Authors: Krishnakumar Balasubramanian
Abstract: We propose and analyze a conservative drifting method for one-step generative modeling. The method replaces the original displacement-based drifting velocity by a kernel density estimator (KDE)-gradient velocity, namely the difference of the kernel-smoothed data score and the kernel-smoothed model score. This velocity is a gradient field, addressing the non-conservatism issue identified for general displacement-based drifting fields. We prove continuous-time finite-particle convergence bounds for the conservative method on $\R^d$: a joint-entropy identity yields bounds for the empirical Stein drift, the smoothed Fisher discrepancy of the KDE, and the squared center velocity. The main finite-particle correction is a reciprocal-KDE self-interaction term, and we give deterministic and high-probability local-occupancy conditions under which this term is controlled. We keep the quadrature constants explicit and track their possible bandwidth dependence: the root residual-velocity rate $N^{-1/(d+4)}$ holds under an additional $h$-uniform quadrature regularity condition, while a more general growth condition yields the optimized root rate $N^{-(2-\beta)/(2(d+4-\beta))}$, where $0\le \beta<2$. We also analyze the non-conservative drifting method with Laplace kernel, corresponding to the original displacement-based velocity proposed in Deng et al., 2026 (arxiv:2602.04770). For this method, a sharp companion kernel decomposes the velocity into a positive scalar preconditioning of a sharp-score mismatch plus a Laplace scale-mismatch residual, producing an analogous finite-particle rate with an unavoidable residual term. Finally, we explain how the continuous-time residual-velocity bounds translate into one-step generation guarantees through the explicit drift size $\eta$.
Authors: Vishal Rajput
Abstract: Robustness, domain adaptation, photometric/occlusion invariance, sensor drift, and alignment style are treated as separate literatures with separate method families. Under label-preserving deployment shift they share one geometric object: the covariance Sigma_task = Cov_{Q_n}(n) of ways inputs can change without changing the label. CORAL, adversarial training, augmentation, metric learning, Jacobian penalties, and alignment constraints are not independent tricks--they are estimators of Sigma_task. Fix that object and the Jacobian penalty is pinned by a matrix Sigma' whose range must cover range(Sigma_task)--the matching principle. We prove optimality in a linear-Gaussian model (Thm. A), necessity of range coverage for any quadratic penalty that zeros deployment drift (Thm. G), and the same dichotomy at global minima (Thm. A*_global). Wrong-direction/signal-aligned controls (Lemma C; Cor. E/E*) and seven estimators (Lemmas D1--D7), plus label-free TDI, yield a falsifiable recipe when Sigma_task must be learned. Thirteen blocks (ML through Qwen2.5-7B) test matched vs isotropic vs wrong-direction penalties on geometry and deployment drift. Twelve match theory where identifiability holds; Office-31 is a named eigengap failure. Partial passes: geometry can improve without every headline task metric moving. A pilot 7B DPO run (one epoch, 240 pairs): matched style-PMH preserves Style TDI where standard DPO degrades it. We do not claim standard training reaches global minima (assumption (O) is open), that estimated Sigma_task is always identifiable, or dominance on every leaderboard. We claim a falsifiable design recipe: estimate Sigma_task, match Sigma', run the controls, report task and geometry separately.
Authors: Stelios Boulitsakis Logothetis, Angela Wood, Pietro Li\`o
Abstract: Survival analysis aims to model how covariates and time jointly shape the time-to-event distribution under right censoring. Classical methods such as the Cox model and generalised additive models (GAMs) require interactions and time-varying effects to be manually specified, which is increasingly impractical on rich clinical datasets. We introduce KAPLAN-HR, a B-spline Kolmogorov-Arnold Network (KAN) for nonparametric estimation of the conditional hazard as a joint function of covariates and time. A single-layer KAPLAN-HR model recovers a GAM, while deeper architectures capture interactions and time-varying effects through composition. We establish a convergence rate for the nonparametric KAN hazard estimator that depends only on the smoothness of the underlying KAN representation and not on the covariate dimension, thereby mitigating the curse of dimensionality for KAN-representable targets. In evaluations over six clinical benchmark datasets, KAPLAN-HR matches or exceeds the predictive performance of established statistical and deep learning survival methods.
Authors: Jiawei He, Mengyu Shi, Jiawei Liu, Zhijie Wang, Chunrong Fang, Xikai Yang, Zhenyu Chen
Abstract: Joint Entity and Relation Extraction (JERE) is highly susceptible to weak generalization due to low-quality training data. Data augmentation is a common strategy to enhance model generalization across different domains. However, existing data augmentation methods often overlook text relevance and may disrupt semantic structures and dependencies, making it difficult to generate effective augmented data for improving model generalization. In this paper, we propose Structured Semantic Data Augmentation (SSDAU), a novel method designed to preserve the semantic structure of text during augmentation. SSDAU segments text based on entity labels and employs an encoder to capture semantic features of entities through context awareness. It then performs entity semantic restructuring to generate augmented data. To distinguish semantically similar entities, SSDAU fuses contextualized embeddings with traditional similarity scores. To mitigate potential topic ambiguity and information loss, we apply the BERTTopic model to filter out irrelevant topics, ensuring topic consistency. We evaluate SSDAU on datasets with different annotation types and compare its performance on five representative JERE models against seven popular data augmentation baselines. Experiments demonstrate that SSDAU generates semantically consistent data with superior robustness against ambiguity (8.26% F1 decrease vs. 31.91% for baselines), significantly outperforming all existing methods across all metrics.
Authors: Hong Qian, Xiang Shu, Xiang Xia, Xuhui Liu, Yangde Fu, Bei Liang, Huibin Wang, Liang Dou
Abstract: Bayesian optimization is widely employed for optimizing complex black-box functions but struggles with the curse of dimensionality. Random embedding, as a dimension reduction strategy, simplifies tasks that possess the effective dimension by optimizing within a low-dimensional subspace. However, determining the effective dimension of a task in advance remains a significant challenge, which influences the selection of the subspace dimensionality and the optimization performance. Traditional methods use fixed subspace dimensions provided by experts or rely on trial and error to estimate subspace dimensions with resources consumed. To this end, this paper proposes an automated random embedding for high-dimensional Bayesian optimization with unknown effective dimension, called Dynamic Shared Embedding Bayesian Optimization (DSEBO). DSEBO starts with a low dimension and switches to a higher subspace if the solutions in the current subspace show preliminary convergence. DSEBO dynamically determines the dimension of the next subspace based on the quality of the solutions in different subspaces and shares the queried solutions with the new subspace for a better initialization. Theoretically, we derive a regret bound for DSEBO and demonstrate that DSEBO can better balance approximation and optimization errors. Extensive experiments on functions with dimensionality of varying magnitudes and real-world tasks with unknown effective dimensions reveal that, compared with state-of-the-art methods, alternating optimization across different subspaces results in significant improvements in high-dimensional optimization, both in terms of optimization regret and time.
Authors: Zhangyi Hu, Chenhui Liu, Tian Huang, Jindong Li, Yang Yang, Jiemin Wu, Zining Zhong, Menglin Yang, Yutao Yue
Abstract: Recently, Reinforcement Learning with Verifiable Rewards (RLVR) and Test-Time Scaling (TTS) have advanced LLM code generation through executable verification. Yet Ground-Truth Unit Tests (GT UTs) remain a bottleneck: SOTA RLVR methods require them for costly training, while existing TTS methods lose competitiveness without them. This motivates GT-free TTS, where existing methods directly use self-generated UTs to refine and select code candidates. Yet such UTs are often noisy or spuriously coupled with wrong code, and UT quality in turn cannot be validated without reliable code. The key challenge is therefore to jointly improve both. To this end, we present CoSPlay, a GT-free, training-free framework that jointly improves codes and UTs through cooperative self-play. It first explores diverse solution ideas and identifies their potential failure modes to produce discriminative UT ideas. It then uses bidirectional pass-count signals from the Code-UT execution matrix to iteratively prune or fix weak codes and refresh or replace unreliable UTs, letting the two pools co-evolve. Finally, when multiple codes remain tied at the highest pass count, it picks the final code from the largest output-consensus cluster, since correct codes agree on the same inputs while wrong codes diverge. Experiments on four challenging benchmarks show that CoSPlay on Qwen2.5-7B-Instruct improves average BoN from 22.1% to 33.2% and UT accuracy from 14.6% to 78.3%, matching or surpassing the RLVR model CURE-7B. When applied to CURE-7B, it further improves BoN by 5.7%. CoSPlay also generalizes across diverse backbones and outperforms GT-free TTS baselines under comparable token budgets, with continued gains as the budget scales up. These results suggest a scalable inference strategy for competitive code generation without any GT data.