Authors: Jun He, Deying Yu
Abstract: The rise of autonomous AI agents exposes a fundamental flaw in API-centric architectures: probabilistic systems directly execute state mutations without sufficient context, coordination, or safety guarantees. We introduce OpenKedge, a protocol that redefines mutation as a governed process rather than an immediate consequence of API invocation. OpenKedge requires actors to submit declarative intent proposals, which are evaluated against deterministically derived system state, temporal signals, and policy constraints prior to execution. Approved intents are compiled into execution contracts that strictly bound permitted actions, resource scope, and time, and are enforced via ephemeral, task-oriented identities. This shifts safety from reactive filtering to preventative, execution-bound enforcement. Crucially, OpenKedge introduces an Intent-to-Execution Evidence Chain (IEEC), which cryptographically links intent, context, policy decisions, execution bounds, and outcomes into a unified lineage. This transforms mutation into a verifiable and reconstructable process, enabling deterministic auditability and reasoning about system behavior. We evaluate OpenKedge across multi-agent conflict scenarios and cloud infrastructure mutations. Results show that the protocol deterministically arbitrates competing intents and cages unsafe execution while maintaining high throughput, establishing a principled foundation for safely operating agentic systems at scale.
Authors: Hongyin Zhu, Jinming Liang, Mengjun Hou, Ruifan Tang, Xianbin Zhu, Jingyuan Yang, Yuanman Mao, Feng Wu
Abstract: Existing LLM-based agent systems share a common architectural failure: they answer from the unrestricted knowledge space without first simulating how active business scenarios reshape that space for the event at hand -- producing decisions that are fluent but ungrounded and carrying no audit trail. We present LOM-action, which equips enterprise AI with \emph{event-driven ontology simulation}: business events trigger scenario conditions encoded in the enterprise ontology~(EO), which drive deterministic graph mutations in an isolated sandbox, evolving a working copy of the subgraph into the scenario-valid simulation graph $G_{\text{sim}}$; all decisions are derived exclusively from this evolved graph. The core pipeline is \emph{event $\to$ simulation $\to$ decision}, realized through a dual-mode architecture -- \emph{skill mode} and \emph{reasoning mode}. Every decision produces a fully traceable audit log. LOM-action achieves 93.82% accuracy and 98.74% tool-chain F1 against frontier baselines Doubao-1.8 and DeepSeek-V3.2, which reach only 24--36% F1 despite 80% accuracy -- exposing the \emph{illusive accuracy} phenomenon. The four-fold F1 advantage confirms that ontology-governed, event-driven simulation, not model scale, is the architectural prerequisite for trustworthy enterprise decision intelligence.
Authors: Olivier Jeunen, Eleanor Hanna, Schaun Wheeler
Abstract: In consumer applications, Customer Relationship Management (CRM) has traditionally relied on the manual optimisation of static, rule-based messaging strategies. While adaptive and autonomous learning systems offer the promise of scalable personalisation, it remains unclear to what extent ``human-in-the-loop'' oversight is required to sustain performance uplift over time. This paper presents a longitudinal case study analysing a real-world consumer application that leverages agentic infrastructure to personalise marketing messaging for a large-scale user base over an 11-month period. We compare two distinct periods: an active phase where marketers directly curated content, audiences, and strategies -- followed immediately by a passive phase where agents operated autonomously from a fixed library of components. Our results demonstrate that whilst active human management generates the highest relative lift in engagement metrics, the autonomous agents successfully sustained a positive lift during the passive period. These findings suggest a symbiotic model where human intervention drives strategic initialisation and discovery, yet autonomous agents can ensure the scalable retention and preservation of performance gains.
Authors: Yarin Benyamin, Argaman Mordoch, Shahaf S. Shperberg, Roni Stern
Abstract: Automated planning algorithms require an action model specifying the preconditions and effects of each action, but obtaining such a model is often hard. Learning action models from observations is feasible, but existing algorithms for numeric domains are offline, requiring expert traces as input. We propose the Reinforcement learning, Action Model learning, and Planning (RAMP) strategy for learning numeric planning action models online via interactions with the environment. RAMP simultaneously trains a Deep Reinforcement Learning (DRL) policy, learns a numeric action model from past interactions, and uses that model to plan future actions when possible. These components form a positive feedback loop: the RL policy gathers data to refine the action model, while the planner generates plans to continue training the RL policy. To facilitate this integration of RL and numeric planning, we developed Numeric PDDLGym, an automated framework for converting numeric planning problems to Gym environments. Experimental results on standard IPC numeric domains show that RAMP significantly outperforms PPO, a well-known DRL algorithm, in terms of solvability and plan quality.
Authors: Petr Ku\v{c}era, Petr Martinek
Abstract: Monadic second order logic (MSO2) plays an important role in parameterized complexity due to the Courcelle's theorem. This theorem states that the problem of checking if a given graph has a property specified by a given MSO2 formula can be solved by a parameterized linear time algorithm with respect to the treewidth of the graph and the size of the formula. We extend this result by showing that models of MSO2 formula with free variables can be represented with a decision diagram whose size is parameterized linear in the above mentioned parameter. In particular, we show a parameterized linear upper bound on the size of a sentential decision diagram (SDD) when treewidth is considered and a parameterized linear upper bound on the size of an ordered binary decision diagram (OBDD) when considering the pathwidth in the parameter. In addition, building on a lower bound on the size of OBDD by Razgon (2014), we show that there is an MSO2 formula and a class of graphs with bounded treewidth which do not admit an OBDD with the size parameterized by the treewidth. Our result offers a new perspective on the Courcelle's theorem and connects it to the area of knowledge representation.
Authors: James Oswald, Daniel Oblinsky, Volodymyr Varha, Vasilije Dragovic, Harsha Kokel, Kavitha Srinivas, Michael Katz, Shirin Sohrabi
Abstract: The generation of planning domains from natural language descriptions remains an open problem even with the advent of large language models and reasoning models. Recent work suggests that while LLMs have the ability to assist with domain generation, they are still far from producing high quality domains that can be deployed in practice. To this end, we investigate the ability of an agentic language model feedback framework to generate planning domains from natural language descriptions that have been augmented with a minimal amount of symbolic information. In particular, we evaluate the quality of the generated domains under various forms of symbolic feedback, including landmarks, and output from the VAL plan validator. Using these feedback mechanisms, we experiment using heuristic search over model space to optimize domain quality.
Authors: John D. Martin, Fraser Mince, Esra'a Saleh, Amy Pajak
Abstract: The situated view of cognition holds that intelligent behavior depends not only on internal memory, but on an agent's active use of environmental resources. Here, we begin formalizing this intuition within Reinforcement Learning (RL). We introduce a mathematical framing for how the environment can functionally serve as an agent's memory, and prove that certain observations, which we call artifacts, can reduce the information needed to represent history. We corroborate our theory with experiments showing that when agents observe spatial paths, the amount of memory required to learn a performant policy is reduced. Interestingly, this effect arises unintentionally, and implicitly through the agent's sensory stream. We discuss the implications of our findings, and show they satisfy qualitative properties previously used to ground accounts of external memory. Moving forward, we anticipate further work on this subject could reveal principled ways to exploit the environment as a substitute for explicit internal memory.
Authors: Pengze Li, Jiaquan Zhang, Yunbo Long, Xinping Liu, Zhou wenjie, Encheng Su, Zihang Zeng, Jiaqi Liu, Jiyao Liu, Junchi Yu, Lihao Liu, Philip Torr, Shixiang Tang, Aoran Wang, Xi Chen
Abstract: Recovering analytical solutions of physical fields from visual observations is a fundamental yet underexplored capability for AI-assisted scientific reasoning. We study visual-to-symbolic analytical solution inference (ViSA) for two-dimensional linear steady-state fields: given field visualizations (and first-order derivatives) plus minimal auxiliary metadata, the model must output a single executable SymPy expression with fully instantiated numeric constants. We introduce ViSA-R2 and align it with a self-verifying, solution-centric chain-of-thought pipeline that follows a physicist-like pathway: structural pattern recognition solution-family (ansatz) hypothesis parameter derivation consistency verification. We also release ViSA-Bench, a VLM-ready synthetic benchmark covering 30 linear steady-state scenarios with verifiable analytical/symbolic annotations, and evaluate predictions by numerical accuracy, expression-structure similarity, and character-level accuracy. Using an 8B open-weight Qwen3-VL backbone, ViSA-R2 outperforms strong open-source baselines and the evaluated closed-source frontier VLMs under a standardized protocol.
Authors: Tianyi Wang, Yixia Li, Long Li, Yibiao Chen, Shaohan Huang, Yun Chen, Peng Li, Yang Liu, Guanhua Chen
Abstract: Proximal Policy Optimization (PPO) is central to aligning Large Language Models (LLMs) in reasoning tasks with verifiable rewards. However, standard token-level PPO struggles in this setting due to the instability of temporal credit assignment over long Chain-of-Thought (CoT) horizons and the prohibitive memory cost of the value model. While critic-free alternatives like GRPO mitigate these issues, they incur significant computational overhead by requiring multiple samples for baseline estimation, severely limiting training throughput. In this paper, we introduce Sequence-Level PPO (SPPO), a scalable algorithm that harmonizes the sample efficiency of PPO with the stability of outcome-based updates. SPPO reformulates the reasoning process as a Sequence-Level Contextual Bandit problem, employing a decoupled scalar value function to derive low-variance advantage signals without multi-sampling. Extensive experiments on mathematical benchmarks demonstrate that SPPO significantly surpasses standard PPO and matches the performance of computation-heavy group-based methods, offering a resource-efficient framework for aligning reasoning LLMs.
Authors: Jinghan Zhang, Fengran Mo, Tharindu Cyril Weerasooriya, Ruimin Dai, Xiaoyan Han, Yanjie Fu, Dakuo Wang, Kunpeng Liu
Abstract: Reinforcement learning (RL) is effective in enhancing the accuracy of large language models in complex reasoning tasks. Existing RL policy optimization frameworks rely on final-answer correctness as feedback signals and rarely capture the internal logical structure of the reasoning process. Consequently, the models would generate fluent and semantically relevant responses but logically inconsistent, structurally erratic, or redundant. To this end, we propose StaRPO, a stability-augmented reinforcement learning framework that explicitly incorporates reasoning stability into the optimization objective. Our StaRPO decomposes stability into two computable lightweight metrics: the Autocorrelation Function (ACF) to evaluate local step-to-step coherence, and Path Efficiency (PE) to evaluate global goal-directedness of the reasoning trajectory. These stability rewards are combined with task rewards to provide complementary and process-aware feedback. We validate the effectiveness of using ACF and PE rewards by showing their correlation with logic errors on two backbone models. Experiments on four reasoning benchmarks show that StaRPO consistently outperforms compared baselines and can enhance both final-answer accuracy and logical stability.
Authors: Nurullah Eymen \"Ozdemir, Erhan Oztop
Abstract: Human cognitive development is shaped not only by individual effort but by structured social interaction, where role-based exchanges such as those between a tutor and a learner, enable solutions that neither could achieve alone. Inspired by these developmental principles, we ask the question whether a tutor-student multi-agent system can create a synergistic effect by pushing Large Language Model (LLM) beyond what it can do within existing frameworks. To test the idea, we adopt autonomous coding problem domain where two agents instantiated from the same LLM assigned asymmetric roles: a student agent generates and iteratively refines solutions, while a tutor agent provides structured evaluative feedback without access to ground-truth answers. In our proposed framework (PETITE), we aim to extract better problem-solving performance from one model by structuring its interaction through complementary roles, rather than relying on stronger supervisory models or heterogeneous ensembles. Our model is evaluated on the APPS coding benchmark against state-of-the-art approaches of Self-Consistency, Self-Refine, Multi-Agent Debate, and Multi-Agent Review. The results show that our model achieves similar or higher accuracy while consuming significantly fewer tokens. These results suggest that developmentally grounded role-differentiated interaction structures provide a principled and resource-efficient paradigm for enhancing LLM problem-solving through structured peer-like interactions. Index Terms- Peer Tutoring, Scaffolding, Large Language Models, Multi-Agent Systems, Code Generation
Authors: Yalun Wu, Haotian Liu, Zhoujun Li, Boyang Wang
Abstract: As Large Language Models (LLMs) advance toward embodied AI agents operating in physical environments, a fundamental question emerges: can models trained on text corpora reliably reason about complex physics while adhering to safety constraints? We address this through PilotBench, a benchmark evaluating LLMs on safety-critical flight trajectory and attitude prediction. Built from 708 real-world general aviation trajectories spanning nine operationally distinct flight phases with synchronized 34-channel telemetry, PilotBench systematically probes the intersection of semantic understanding and physics-governed prediction through comparative analysis of LLMs and traditional forecasters. We introduce Pilot-Score, a composite metric balancing 60% regression accuracy with 40% instruction adherence and safety compliance. Comparative evaluation across 41 models uncovers a Precision-Controllability Dichotomy: traditional forecasters achieve superior MAE of 7.01 but lack semantic reasoning capabilities, while LLMs gain controllability with 86--89% instruction-following at the cost of 11--14 MAE precision. Phase-stratified analysis further exposes a Dynamic Complexity Gap-LLM performance degrades sharply in high-workload phases such as Climb and Approach, suggesting brittle implicit physics models. These empirical discoveries motivate hybrid architectures combining LLMs' symbolic reasoning with specialized forecasters' numerical precision. PilotBench provides a rigorous foundation for advancing embodied AI in safety-constrained domains.
Authors: Sihang Jiang, Lipeng Ma, Zhonghua Hong, Keyi Wang, Zhiyu Lu, Shisong Chen, Jinghao Zhang, Tianjun Pan, Weijia Zhou, 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 or optimize strategies across task boundaries. While the Self-Evolving Agent (SEA) paradigm has been previously proposed, this paper contributes a new formal definition of SEA grounded in digital embodiment and continuous cross-task evolution, and introduces SEA-Eval, the first benchmark designed to evaluate SEA characteristics across two dimensions, intra-task execution reliability and long-term evolutionary performance. By organizing tasks into sequential streams and analyzing Success Rate and Token Consumption over time, SEA-Eval quantifies evolutionary gain and structural stability in ways that existing episodic benchmarks cannot. Empirical evaluations reveal a significant evolutionary bottleneck in current state-of-the-art frameworks, where identical success rates mask up to 31.2 times differences in token consumption and divergent evolutionary trajectories under sequential analysis. SEA-Eval provides a rigorous scientific foundation for advancing agents from mere task executors toward genuinely self-evolving digital entities.
Authors: Hiroya Ijima, Koichiro Yawata
Abstract: Enumerating Minimal Unsatisfiable Subsets (MUSes) is a fundamental task in constraint satisfaction problems (CSPs). Its major challenge is the exponential growth of the search space, which becomes particularly severe when satisfiability checks are expensive. Recent machine learning approaches reduce this cost for Boolean satisfiability problems but rely on explicit variable-constraint relationships, limiting their application domains. This paper proposes a domain-agnostic method to accelerate MUS enumeration using Hypergraph Neural Networks (HGNNs). The proposed method incrementally builds a hypergraph with constraints as vertices and MUSes enumerated until the current step as hyperedges, and employs an HGNN-based agent trained via reinforcement learning to minimize the number of satisfiability checks required to obtain an MUS. Experimental results demonstrate the effectiveness of our approach in accelerating MUS enumeration, showing that our method can enumerate more MUSes within the same satisfiability check budget compared to conventional methods.
Authors: Daniele Foffano, Arvid Eriksson, David Broman, Karl H. Johansson, Alexandre Proutiere
Abstract: Model-based reinforcement learning (MBRL) with autoregressive world models suffers from compounding errors, whereas diffusion world models mitigate this by generating trajectory segments jointly. However, existing diffusion guides are either policy-only, discarding value information, or reward-based, which becomes myopic when the diffusion horizon is short. We introduce Advantage-Guided Diffusion for MBRL (AGD-MBRL), which steers the reverse diffusion process using the agent's advantage estimates so that sampling concentrates on trajectories expected to yield higher long-term return beyond the generated window. We develop two guides: (i) Sigmoid Advantage Guidance (SAG) and (ii) Exponential Advantage Guidance (EAG). We prove that a diffusion model guided through SAG or EAG allows us to perform reweighted sampling of trajectories with weights increasing in state-action advantage-implying policy improvement under standard assumptions. Additionally, we show that the trajectories generated from AGD-MBRL follow an improved policy (that is, with higher value) compared to an unguided diffusion model. AGD integrates seamlessly with PolyGRAD-style architectures by guiding the state components while leaving action generation policy-conditioned, and requires no change to the diffusion training objective. On MuJoCo control tasks (HalfCheetah, Hopper, Walker2D and Reacher), AGD-MBRL improves sample efficiency and final return over PolyGRAD, an online Diffuser-style reward guide, and model-free baselines (PPO/TRPO), in some cases by a margin of 2x. These results show that advantage-aware guidance is a simple, effective remedy for short-horizon myopia in diffusion-model MBRL.
Authors: Ruihong Shen, Shiqian Li, Yixin Zhu
Abstract: Humans effortlessly navigate the physical world by predicting how objects behave under gravity and contact forces, yet how such judgments support sequential physical planning under resource constraints remains poorly understood. Research on intuitive physics debates whether prediction relies on the Intuitive Physics Engine (IPE) or fast, cue-based heuristics; separately, decision-making research debates deliberative lookahead versus myopic strategies. These debates have proceeded in isolation, leaving the cognitive architecture of sequential physical planning underspecified. How physical prediction mechanisms and planning strategies jointly adapt under limited cognitive resources remains an open question. Here we show that humans exhibit a dual transition under resource pressure, simultaneously shifting both physical prediction mechanism and planning strategy to match cognitive budget. Using Overhang Tower, a construction task requiring participants to maximize horizontal overhang while maintaining stability, we find that IPE-based simulation dominates early stages while CNN-based visual heuristics prevail as complexity grows; concurrently, time pressure truncates deliberative lookahead, shifting planning toward shallower horizons: a dual transition unpredicted by prior single-mechanism accounts. These findings reveal a hierarchical, resource-rational architecture that flexibly trades computational cost against predictive fidelity. Our results unify two long-standing debates (simulation vs. heuristics and myopic vs. deliberative planning) as a dynamic repertoire reconfigured by cognitive budget.
Authors: Haobo Hu, Qi Mao, Yuanhang Li, Libiao Jin
Abstract: We propose Camera Artist, a multi-agent framework that models a real-world filmmaking workflow to generate narrative videos with explicit cinematic language. While recent multi-agent systems have made substantial progress in automating filmmaking workflows from scripts to videos, they often lack explicit mechanisms to structure narrative progression across adjacent shots and deliberate use of cinematic language, resulting in fragmented storytelling and limited filmic quality. To address this, Camera Artist builds upon established agentic pipelines and introduces a dedicated Cinematography Shot Agent, which integrates recursive storyboard generation to strengthen shot-to-shot narrative continuity and cinematic language injection to produce more expressive, film-oriented shot designs. Extensive quantitative and qualitative results demonstrate that our approach consistently outperforms existing baselines in narrative consistency, dynamic expressiveness, and perceived film quality.
Authors: Young-Suk Lee, Ramon Fernandez Astudillo, Radu Florian
Abstract: Deep research agents increasingly interleave web browsing with multi-step computation, yet existing benchmarks evaluate these capabilities in isolation, creating a blind spot in assessing real-world performance. We introduce DRBENCHER, a synthetic benchmark generator for questions that require both browsing and computation. It enforces four criteria: verifiability (gold answers are computed by executing parameterized code over knowledge-graph values), complexity (multi-hop entity identification, property retrieval, and domain-specific computation), difficulty (a two-stage verification cascade filters out questions solvable by the generating model), and diversity (a greedy max-min embedding filter maximizes coverage). These criteria are realized via a unified answer-first pipeline spanning five domains: biochemistry, financial, geophysical, security, and history. Human evaluation shows 76% validity (84% excluding stale data), with 35% of errors due to outdated knowledge-graph entries, highlighting an inherent limitation of systems that reason over evolving data. Automatic evaluation shows that the strongest frontier model achieves only 20% answer accuracy. Compared to manually constructed benchmarks (BrowseComp+, MATH-500, GPQA), DRBENCHER achieves the highest semantic diversity.
Authors: Ling Shi, Yuqin Dai, Ziyin Wang, Ning Gao, Wei Zhang, Chaozheng Wang, Yujie Wang, Wei He, Jinpeng Wang, Deiyi Xiong
Abstract: The development of Large Language Models (LLMs) has catalyzed automation in customer service, yet benchmarking their performance remains challenging. Existing benchmarks predominantly rely on static paradigms and single-dimensional metrics, failing to account for diverse user behaviors or the strict adherence to structured Standard Operating Procedures (SOPs) required in real-world deployments. To bridge this gap, we propose SAGE (Service Agent Graph-guided Evaluation), a universal multi-agent benchmark for automated, dual-axis assessment. SAGE formalizes unstructured SOPs into Dynamic Dialogue Graphs, enabling precise verification of logical compliance and comprehensive path coverage. We introduce an Adversarial Intent Taxonomy and a modular Extension Mechanism, enabling low-cost deployment across domains and facilitating automated dialogue data synthesis. Evaluation is conducted via a framework where Judge Agents and a Rule Engine analyze interactions between User and Service Agents to generate deterministic ground truth. Extensive experiments on 27 LLMs across 6 industrial scenarios reveal a significant ``Execution Gap'' where models accurately classify intents but fail to derive correct subsequent actions. We also observe ``Empathy Resilience'', a phenomenon where models maintain polite conversational facades despite underlying logical failures under high adversarial intensity. Code and resources are available at https://anonymous.4open.science/r/SAGE-Bench-4CD3/.
Authors: Maochen Sun, Youzhi Zhang, Gaofeng Meng
Abstract: Large language models are making autonomous drug discovery agents increasingly feasible, but reliable success in this setting is not determined by any single action or molecule. It is determined by whether the final returned set jointly satisfies protocol-level requirements such as set size, diversity, binding quality, and developability. This creates a fundamental control problem: the agent plans step by step, while task validity is decided at the level of the whole candidate set. Existing language-based drug discovery systems therefore tend to rely on long raw history and under-specified self-reflection, making failure localization imprecise and planner-facing agent states increasingly noisy. We present CACM (Constraint-Aware Corrective Memory), a language-based drug discovery framework built around precise set-level diagnosis and a concise memory write-back mechanism. CACM introduces protocol auditing and a grounded diagnostician, which jointly analyze multimodal evidence spanning task requirements, pocket context, and candidate-set evidence to localize protocol violations, generate actionable remediation hints, and bias the next action toward the most relevant correction. To keep planning context compact, CACM organizes memory into static, dynamic, and corrective channels and compresses them before write-back, thereby preserving persistent task information while exposing only the most decision-relevant failures. Our experimental results show that CACM improves the target-level success rate by 36.4% over the state-of-the-art baseline. The results show that reliable language-based drug discovery benefits not only from more powerful molecular tools, but also from more precise diagnosis and more economical agent states.
Authors: Lars Benedikt Kaesberg, Tianyu Yang, Niklas Bauer, Terry Ruas, Jan Philip Wahle, Bela Gipp
Abstract: Spatial reasoning is central to navigation and robotics, yet measuring model capabilities on these tasks remains difficult. Existing benchmarks evaluate models in a one-shot setting, requiring full solution generation in a single response, unlike humans, who work in interactive environments step-by-step. We introduce Spatial-Gym, a Gymnasium environment that isolates spatial constraint reasoning by testing pathfinding in 2D-grid puzzles as a sequential decision task with optional backtracking. We evaluate eight models in three settings (one-shot, step-by-step, step-by-step with backtracking) against human, random, and A* baselines on 500 episodes. The best model, GPT-OSS 120B, achieves a solve rate of 16.0%, 82 points below the human baseline (98.0%). Step-by-step format helps weaker models (up to +5.4%) by removing formatting errors, but hurts stronger models (up to 5.6%) by constraining global planning. Backtracking improves episode completion, but increases solve rate only for weaker models; stronger models rarely backtrack and do not benefit from it. Our experiments have three key findings: (1) models fail to scale reasoning effort with difficulty, (2) vision models receiving images of the spatial environment reduce solve rate by 73%, and (3) extended chain-of-thought reasoning retains a 3-5x accuracy advantage over standard inference even in the step-by-step setting. Spatial-Gym enables diagnosis of model limitations and provides a framework for improving spatial reasoning through reinforcement learning.
Authors: Mohamed Elfeki, Tu Trinh, Kelvin Luu, Guangze Luo, Nathan Hunt, Ernesto Montoya, Nandan Marwaha, Yannis He, Charles Wang, Fernando Crabedo, Alessa Castilo, Bing Liu
Abstract: Frontier coding agents solve complex tasks when given complete context but collapse when specifications are incomplete or ambiguous. The bottleneck is not raw capability, but judgment: knowing when to act autonomously and when to ask for help. Current benchmarks are blind to this failure mode. They supply unambiguous detailed instructions and solely reward execution correctness, so an agent that makes a lucky guess for a missing requirement will score identically to one that would have asked to be certain. We present HiL-Bench (Human-in-the-Loop Benchmark) to measure this selective escalation skill. Each task contains human-validated blockers (missing information, ambiguous requests, contradictory information) that surface only through progressive exploration, not upfront inspection. Our core metric, Ask-F1, the harmonic mean of question precision and blocker recall, captures the tension between over-asking and silent guessing; its structure architecturally prevents gaming through question spam. Evaluation across SWE and text-to-SQL domains reveals a large universal judgment gap: no frontier model recovers more than a fraction of its full-information performance when deciding whether to ask. Failure analysis identifies three key help-seeking patterns: overconfident wrong beliefs with no gap detection; high uncertainty detection yet persistent errors; broad, imprecise escalation without self-correction. These consistent patterns confirm poor help-seeking is a model-level flaw, not task-specific. RL training on shaped Ask-F1 reward shows judgment is trainable: a 32B model improves both help-seeking quality and task pass rate, with gains that transfer across domains. The model does not learn domain-specific heuristics for when to ask; it learns to detect unresolvable uncertainty and act on it.
Authors: Chao Jiang, Jingyu Huang, Miqing Li
Abstract: Many-objective optimisation, a subset of multi-objective optimisation, involves optimisation problems with more than three objectives. As the number of objectives increases, the number of solutions needed to adequately represent the entire Pareto front typically grows substantially. This makes it challenging, if not infeasible, to design a search algorithm capable of effectively exploring the entire Pareto front. This difficulty is particularly acute in the Bayesian optimisation paradigm, where sample efficiency is critical and only a limited number of solutions (often a few hundred) are evaluated. Moreover, after the optimisation process, the decision-maker eventually selects just one solution for deployment, regardless of how many high-quality, diverse solutions are available. In light of this, we argue an idea that under a very limited evaluation budget, it may be more useful to focus on finding a single solution of the highest possible quality for the decision-maker, rather than aiming to approximate the entire Pareto front as existing many-/multi-objective Bayesian optimisation methods typically do. Bearing this idea in mind, this paper proposes a \underline{s}ingle \underline{p}oint-based \underline{m}ulti-\underline{o}bjective search framework (SPMO) that aims to improve the quality of solutions along a direction that leads to a good tradeoff between objectives. Within SPMO, we present a simple acquisition function, called expected single-point improvement (ESPI), working under both noiseless and noisy scenarios. We show that ESPI can be optimised effectively with gradient-based methods via the sample average approximation (SAA) approach and theoretically prove its convergence guarantees under the SAA. We also empirically demonstrate that the proposed SPMO is computationally tractable and outperforms state-of-the-arts on a wide range of benchmark and real-world problems.
Authors: Weiyang Guo, Zesheng Shi, Liye Zhao, Jiayuan Ma, Zeen Zhu, Junxian He, Min Zhang, Jing Li
Abstract: While Large Language Models (LLMs) have demonstrated significant potential in Tool-Integrated Reasoning (TIR), existing training paradigms face significant limitations: Zero-RL suffers from inefficient exploration and mode degradation due to a lack of prior guidance, while SFT-then-RL is limited by high data costs and capability plateaus caused by low-entropy collapse. To address these challenges, we propose E3-TIR (Enhanced Experience Exploitation), a warm-up paradigm for the early stages of agent training. Specifically, we formulate training as the dynamic integration of three experience types: Expert Prefixes, Expert Guided, and Self-Exploration. By executing diverse branching exploration around expert "anchors" and employing a mix policy optimization mechanism, we effectively mitigate distribution shifts and resolve optimization conflicts arising from shared prefixes. Our method dynamically adapts the model's knowledge boundaries, effectively balancing exploration diversity with training efficiency.Experimental results demonstrate that E3-TIR achieves a 6 performance improvement over traditional paradigms on tool-use tasks, while requiring less than 10 of the synthetic data. Furthermore, in terms of ROI, a comprehensive metric integrating performance, data cost, and training efficiency we achieve a 1.46x gain compared to baselines. Code is available at https://github.com/yuki-younai/E3-TIR.
Authors: Jiwoong Sohn, Tomasz Sternal, Kenneth Styppa, Torsten Hoefler, Michael Moor
Abstract: Reasoning in knowledge-intensive domains remains challenging as intermediate steps are often not locally verifiable: unlike math or code, evaluating step correctness may require synthesizing clues across large external knowledge sources. As a result, subtle errors can propagate through reasoning traces, potentially never to be detected. Prior work has proposed process reward models (PRMs), including retrieval-augmented variants, but these methods operate post hoc, scoring completed trajectories, which prevents their integration into dynamic inference procedures. Here, we introduce Process Reward Agents (PRA), a test-time method for providing domain-grounded, online, step-wise rewards to a frozen policy. In contrast to prior retrieval-augmented PRMs, PRA enables search-based decoding to rank and prune candidate trajectories at every generation step. Experiments on multiple medical reasoning benchmarks demonstrate that PRA consistently outperforms strong baselines, achieving 80.8% accuracy on MedQA with Qwen3-4B, a new state of the art at the 4B scale. Importantly, PRA generalizes to unseen frozen policy models ranging from 0.5B to 8B parameters, improving their accuracy by up to 25.7% without any policy model updates. More broadly, PRA suggests a paradigm in which frozen reasoners are decoupled from domain-specific reward modules, allowing the deployment of new backbones in complex domains without retraining.
Authors: Gonzalo Ballestero, Hadi Hosseini, Samarth Khanna, Ran I. Shorrer
Abstract: AI agents increasingly operate in multi-agent environments where outcomes depend on coordination. We distinguish primary algorithmic monoculture -- baseline action similarity -- from strategic algorithmic monoculture, whereby agents adjust similarity in response to incentives. We implement a simple experimental design that cleanly separates these forces, and deploy it on human and large language model (LLM) subjects. LLMs exhibit high levels of baseline similarity (primary monoculture) and, like humans, they regulate it in response to coordination incentives (strategic monoculture). While LLMs coordinate extremely well on similar actions, they lag behind humans in sustaining heterogeneity when divergence is rewarded.
Authors: Tiago da Silva, Eliezer de Souza da Silva, Diego Mesquita
Abstract: Generative Flow Networks (GFlowNets) are amortized inference models designed to sample from unnormalized distributions over composable objects, with applications in generative modeling for tasks in fields such as causal discovery, NLP, and drug discovery. Traditionally, the training procedure for GFlowNets seeks to minimize the expected log-squared difference between a proposal (forward policy) and a target (backward policy) distribution, which enforces certain flow-matching conditions. While this training procedure is closely related to variational inference (VI), directly attempting standard Kullback-Leibler (KL) divergence minimization can lead to proven biased and potentially high-variance estimators. Therefore, we first review four divergence measures, namely, Renyi-$\alpha$'s, Tsallis-$\alpha$'s, reverse and forward KL's, and design statistically efficient estimators for their stochastic gradients in the context of training GFlowNets. Then, we verify that properly minimizing these divergences yields a provably correct and empirically effective training scheme, often leading to significantly faster convergence than previously proposed optimization. To achieve this, we design control variates based on the REINFORCE leave-one-out and score-matching estimators to reduce the variance of the learning objectives' gradients. Our work contributes by narrowing the gap between GFlowNets training and generalized variational approximations, paving the way for algorithmic ideas informed by the divergence minimization viewpoint.
Authors: Jiawei Chen, Ruoxi Xu, Boxi Cao, Ruotong Pan, Yunfei Zhang, Yifei Hu, Yong Du, Tingting Gao, Yaojie Lu, Yingfei Sun, Xianpei Han, Le Sun, Xiangyu Wu, Hongyu Lin
Abstract: The emergence of Large Language Models (LLMs) has illuminated the potential for a general-purpose user simulator. However, existing benchmarks remain constrained to isolated scenarios, narrow action spaces, or synthetic data, failing to capture the holistic nature of authentic human behavior. To bridge this gap, we introduce OmniBehavior, the first user simulation benchmark constructed entirely from real-world data, integrating long-horizon, cross-scenario, and heterogeneous behavioral patterns into a unified framework. Based on this benchmark, we first provide empirical evidence that previous datasets with isolated scenarios suffer from tunnel vision, whereas real-world decision-making relies on long-term, cross-scenario causal chains. Extensive evaluations of state-of-the-art LLMs reveal that current models struggle to accurately simulate these complex behaviors, with performance plateauing even as context windows expand. Crucially, a systematic comparison between simulated and authentic behaviors uncovers a fundamental structural bias: LLMs tend to converge toward a positive average person, exhibiting hyper-activity, persona homogenization, and a Utopian bias. This results in the loss of individual differences and long-tail behaviors, highlighting critical directions for future high-fidelity simulation research.
Authors: Milo\v{s} Ko\v{s}prdi\'c, Adela Ljaji\'c, Bojana Ba\v{s}aragin, Darija Medvecki, Lorenzo Cassano, Nikola Milo\v{s}evi\'c
Abstract: We introduce VerifAI, an open-source expert system for biomedical question answering that integrates retrieval-augmented generation (RAG) with a novel post-hoc claim verification mechanism. Unlike standard RAG systems, VerifAI ensures factual consistency by decomposing generated answers into atomic claims and validating them against retrieved evidence using a fine-tuned natural language inference (NLI) engine. The system comprises three modular components: (1) a hybrid Information Retrieval (IR) module optimized for biomedical queries (MAP@10 of 42.7%), (2) a citation-aware Generative Component fine-tuned on a custom dataset to produce referenced answers, and (3) a Verification Component that detects hallucinations with state-of-the-art accuracy, outperforming GPT-4 on the HealthVer benchmark. Evaluations demonstrate that VerifAI significantly reduces hallucinated citations compared to zero-shot baselines and provides a transparent, verifiable lineage for every claim. The full pipeline, including code, models, and datasets, is open-sourced to facilitate reliable AI deployment in high-stakes domains.
Authors: Qiyu Qin, Yichen Li, Haozhao Wang, Cheng Wang, Rui Zhang, Ruixuan Li
Abstract: Fake orders pose increasing threats to sequential recommender systems by misleading recommendation results through artificially manipulated interactions, including click farming, context-irrelevant substitutions, and sequential perturbations. Unlike injecting carefully designed fake users to influence recommendation performance, fake orders embedded within genuine user sequences aim to disrupt user preferences and mislead recommendation results, thereby manipulating exposure rates of specific items to gain competitive advantages. To protect users' authentic interest preferences and eliminate misleading information, this paper aims to perform precise and efficient rectification on compromised sequential recommender systems while avoiding the enormous computational and time costs of retraining existing models. Specifically, we identify that fake orders are not absolutely harmful - in certain cases, partial fake orders can even have a data augmentation effect. Based on this insight, we propose Dual-view Identification and Targeted Rectification (DITaR), which primarily identifies harmful samples to achieve unbiased rectification of the system. The core idea of this method is to obtain differentiated representations from collaborative and semantic views for precise detection, and then filters detected suspicious fake orders to select truly harmful ones for targeted rectification with gradient ascent. This ensures that useful information in fake orders is not removed while preventing bias residue. Moreover, it maintains the original data volume and sequence structure, thus protecting system performance and trustworthiness to achieve optimal unbiased rectification. Extensive experiments on three datasets demonstrate that DITaR achieves superior performance compared to state-of-the-art methods in terms of recommendation quality, computational efficiency, and system robustness.
Authors: Josef Hardi, Martin J. O'Connor, Marcos Martinez-Romero, Jean G. Rosario, Stephen A. Fisher, Mark A. Musen
Abstract: Scientific metadata are often incomplete and noncompliant with community standards, limiting dataset findability, interoperability, and reuse. When reporting guidelines exist, they typically lack machine-actionable representations. Producing FAIR datasets requires encoding metadata standards as machine-actionable templates with rich field specifications and precise value constraints. Recent work has shown that LLMs guided by field names and ontology constraints can improve metadata standardization, but these approaches treat constraints as static text prompts, relying on the model's training knowledge alone. We present an LLM-based metadata standardization system that queries authoritative biomedical terminology services in real time to retrieve canonically correct vocabulary terms on demand. We evaluate this approach on 839 legacy metadata records from the Human BioMolecular Atlas Program (HuBMAP) using an expert-curated gold standard for exact-match assessment. Our evaluation shows that augmenting the LLM with real-time tool access consistently improves prediction accuracy over the LLM alone across both ontology-constrained and non-ontology-constrained fields, demonstrating a practical, scalable approach to automated standardization of biomedical metadata.
Authors: Ruiyao Xu, Kaize Ding
Abstract: Large Language Models (LLMs) have shown strong performance on text-attributed graphs (TAGs) due to their superior semantic understanding ability on textual node features. However, their effectiveness as predictors in the low-resource setting, where labeled nodes are severely limited and scarce, remains constrained since fine-tuning LLMs usually requires sufficient labeled data, especially when the TAG shows complex structural patterns. In essence, this paper targets two key challenges: (i) the difficulty of generating and selecting reliable pseudo labels on TAGs for LLMs, and (ii) the need to mitigate potential label noise when fine-tuning LLMs with pseudo labels. To counter the challenges, we propose a new framework, GNN-as-Judge, which can unleash the power of LLMs for few-shot semi-supervised learning on TAGs by incorporating the structural inductive bias of Graph Neural Networks (GNNs). Specifically, GNN-as-Judge introduces a collaborative pseudo-labeling strategy that first identifies the most influenced unlabeled nodes from labeled nodes, then exploits both the agreement and disagreement patterns between LLMs and GNNs to generate reliable labels. Furthermore, we develop a weakly-supervised LLM fine-tuning algorithm that can distill the knowledge from informative pseudo labels while mitigating the potential label noise. Experiments on multiple TAG datasets demonstrate that GNN-as-Judge significantly outperforms existing methods, particularly in low-resource regimes where labeled data are scarce.
Authors: S{\o}ren Riis
Abstract: The public text record -- the material from which both people and AI systems now learn -- is increasingly shaped by its own outputs. Generated text enters the public record, later agents learn from it, and the cycle repeats. Here we develop an exactly solvable mathematical framework for this recursive process, based on variable-order $n$-gram agents, and separate two forces acting on the public corpus. The first is drift: unfiltered reuse progressively removes rare forms, and in the infinite-corpus limit we characterise the stable distributions exactly. The second is selection: publication, ranking and verification filter what enters the record, and the outcome depends on what is selected. When publication merely reflects the statistical status quo, the corpus converges to a shallow state in which further lookahead brings no benefit. When publication is normative -- rewarding quality, correctness or novelty -- deeper structure persists, and we establish an optimal upper bound on the resulting divergence from shallow equilibria. The framework therefore identifies when recursive publication compresses public text and when selective filtering sustains richer structure, with implications for the design of AI training corpora.
Authors: Arth Singh
Abstract: What exactly do efficient sequence models gain over simple temporal averaging? We use exponential moving average (EMA) traces, the simplest recurrent context (no gating, no content-based retrieval), as a controlled probe to map the boundary between what fixed-coefficient accumulation can and cannot represent. EMA traces encode temporal structure: a Hebbian architecture with multi-timescale traces achieves 96% of a supervised BiGRU on grammatical role assignment with zero labels, surpassing the supervised model on structure-dependent roles. EMA traces destroy token identity: a 130M-parameter language model using only EMA context reaches C4 perplexity 260 (8x GPT-2), and a predictor ablation (replacing the linear predictor with full softmax attention) yields identical loss, localizing the entire gap to the traces. The traces apply lossy, data-independent compression; by the data processing inequality, no downstream predictor can recover the discarded information. Fixed-coefficient accumulation, whether across time or depth, suffers irreversible information dilution that only learned, input-dependent selection can resolve.
Authors: Arth Singh
Abstract: Diffusion-based language models (dLLMs) generate text by iteratively denoising masked token sequences. We show that their safety alignment rests on a single fragile assumption: that the denoising schedule is monotonic and committed tokens are never re-evaluated. Safety-aligned dLLMs commit refusal tokens within the first 8-16 of 64 denoising steps, and the schedule treats these commitments as permanent. A trivial two-step intervention - re-masking these tokens and injecting a 12-token affirmative prefix - achieves 76.1% ASR on HarmBench (n=159, Lg=128) against LLaDA-8B-Instruct and 81.8% ASR (n=159) against Dream-7B-Instruct, without any gradient computation or adversarial search. The simplicity of this exploit is itself the central finding: augmenting with gradient-optimized perturbation via a differentiable Gumbel-softmax chain consistently degrades ASR (e.g., 41.5% vs. 76.1% at Lg=128), confirming that the vulnerability is structural rather than requiring sophisticated exploitation. These findings reveal that dLLM safety is not adversarially robust but architecturally shallow - it holds only because the denoising schedule is never violated. We discuss defenses including safety-aware unmasking schedules, step-conditional prefix detection, and post-commitment re-verification.
Authors: Hanna Lee, Tan Dat Nguyen, Jaehoon Kang, Kyuhong Shim
Abstract: Recent decoder-only autoregressive text-to-speech (AR-TTS) models produce high-fidelity speech, but their memory and compute costs scale quadratically with sequence length due to full self-attention. In this paper, we propose WAND, Windowed Attention and Knowledge Distillation, a framework that adapts pretrained AR-TTS models to operate with constant computational and memory complexity. WAND separates the attention mechanism into two: persistent global attention over conditioning tokens and local sliding-window attention over generated tokens. To stabilize fine-tuning, we employ a curriculum learning strategy that progressively tightens the attention window. We further utilize knowledge distillation from a full-attention teacher to recover high-fidelity synthesis quality with high data efficiency. Evaluated on three modern AR-TTS models, WAND preserves the original quality while achieving up to 66.2% KV cache memory reduction and length-invariant, near-constant per-step latency.
Authors: Xiaohan Ren, Chenxiao Fan, Wenyin Ma, Hongliang He, Chongming Gao, Xiaoyan Zhao, Fuli Feng
Abstract: Large language models (LLMs) have achieved strong performance on medical exam-style tasks, motivating growing interest in their deployment in real-world clinical settings. However, clinical decision-making is inherently safety-critical, context-dependent, and conducted under evolving evidence. In such situations, reliable LLM performance depends not on factual recall alone, but on robust medical reasoning. In this work, we present a comprehensive review of medical reasoning with LLMs. Grounded in cognitive theories of clinical reasoning, we conceptualize medical reasoning as an iterative process of abduction, deduction, and induction, and organize existing methods into seven major technical routes spanning training-based and training-free approaches. We further conduct a unified cross-benchmark evaluation of representative medical reasoning models under a consistent experimental setting, enabling a more systematic and comparable assessment of the empirical impact of existing methods. To better assess clinically grounded reasoning, we introduce MR-Bench, a benchmark derived from real-world hospital data. Evaluations on MR-Bench expose a pronounced gap between exam-level performance and accuracy on authentic clinical decision tasks. Overall, this survey provides a unified view of existing medical reasoning methods, benchmarks, and evaluation practices, and highlights key gaps between current model performance and the requirements of real-world clinical reasoning.
Authors: Leonid Erlygin, Alexey Zaytsev
Abstract: Accurate uncertainty estimation is essential for building robust and trustworthy recognition systems. In this paper, we consider the open-set text classification (OSTC) task - and uncertainty estimation for it. For OSTC a text sample should be classified as one of the existing classes or rejected as unknown. To account for the different uncertainty types encountered in OSTC, we adapt the Holistic Uncertainty Estimation (HolUE) method for the text domain. Our approach addresses two major causes of prediction errors in text recognition systems: text uncertainty that stems from ill formulated queries and gallery uncertainty that is related the ambiguity of data distribution. By capturing these sources, it becomes possible to predict when the system will make a recognition error. We propose a new OSTC benchmark and conduct extensive experiments on a wide range of data, utilizing the authorship attribution, intent and topic classification datasets. HolUE achieves 40-365% improvement in Prediction Rejection Ratio (PRR) over the quality-based SCF baseline across datasets: 365% on Yahoo Answers (0.79 vs 0.17 at FPIR 0.1), 347% on DBPedia (0.85 vs 0.19), 240% on PAN authorship attribution (0.51 vs 0.15 at FPIR 0.5), and 40% on CLINC150 intent classification (0.73 vs~0.52). We make public our code and protocols https://github.com/Leonid-Erlygin/text_uncertainty.git
URLs: https://github.com/Leonid-Erlygin/text_uncertainty.git
Authors: Ilya Trofimenko, David Kocharyan, Aleksandr Zaitsev, Pavel Repnikov, Mark Levin, Nikita Shevtsov
Abstract: Ensuring that Text-to-Speech (TTS) systems deliver human-perceived quality at scale is a central challenge for modern speech technologies. Human subjective evaluation protocols such as Mean Opinion Score (MOS) and Side-by-Side (SBS) comparisons remain the de facto gold standards, yet they are expensive, slow, and sensitive to pervasive assessor biases. This study addresses these barriers by formulating, and implementing a suite of novel neural models designed to approximate expert judgments in both relative (SBS) and absolute (MOS) settings. For relative assessment, we propose NeuralSBS, a HuBERT-backed model achieving 73.7% accuracy (on SOMOS dataset). For absolute assessment, we introduce enhancements to MOSNet using custom sequence-length batching, as well as WhisperBert, a multimodal stacking ensemble that combines Whisper audio features and BERT textual embeddings via weak learners. Our best MOS models achieve a Root Mean Square Error (RMSE) of ~0.40, significantly outperforming the human inter-rater RMSE baseline of 0.62. Furthermore, our ablation studies reveal that naively fusing text via cross-attention can degrade performance, highlighting the effectiveness of ensemble-based stacking over direct latent fusion. We additionally report negative results with SpeechLM-based architectures and zero-shot LLM evaluators (Qwen2-Audio, Gemini 2.5 flash preview), reinforcing the necessity of dedicated metric learning frameworks.
Authors: Mousa Salah, Amgad Muneer
Abstract: Extended reasoning models represent a transformative shift in Large Language Model (LLM) capabilities by enabling explicit test-time computation for complex problem solving. However, the optimal configuration of sampling temperature and prompting strategy for these systems remains largely underexplored. We systematically evaluate chain-of-thought and zero-shot prompting across four temperature settings (0.0, 0.4, 0.7, and 1.0) using Grok-4.1 with extended reasoning on 39 mathematical problems from AMO-Bench, a challenging International Mathematical Olympiad-level benchmark. We find that zero-shot prompting achieves peak performance at moderate temperatures, reaching 59% accuracy at T=0.4 and T=0.7, while chain-of-thought prompting performs best at the temperature extremes. Most notably, the benefit of extended reasoning increases from 6x at T=0.0 to 14.3x at T=1.0. These results suggest that temperature should be optimized jointly with prompting strategy, challenging the common practice of using T=0 for reasoning tasks.
Authors: Reza Sedghi, Robin Schiewer, Anand Subramoney, David Kappel
Abstract: At typical context lengths, the feed-forward MLP block accounts for a large share of a transformer's compute budget, motivating sparse alternatives to dense MLP blocks. We study sparse, tree-structured feed-forward layers as drop-in replacements for MLP blocks in deep transformer architectures, enabling conditional computation via hard hierarchical routing without a separate router network. We demonstrate for the first time that this form of tree-structured conditional sparsity can be applied for autoregressive language modeling and downstream question answering, including zero- and few-shot settings, and its scalability beyond 1B parameters. Despite activating fewer than 5% of the feed-forward block's units per token, our models match dense baselines under controlled training and fine-tuning protocols. We further analyze training dynamics and identify an emergent auto-pruning effect: the interaction of hard routing with asymmetric nonlinearities progressively deactivates unused paths, yielding partial conversion of dynamic routing into static structural sparsity. We show that simple architectural choices can modulate this behavior and recover balanced trees without auxiliary losses. Overall, our work demonstrates that tree-structured feed-forward layers provide a scalable and controllable mechanism for sparsifying large transformer models.
Authors: Nabelanita Utami, Sasano Ryohei
Abstract: The evolution of writing assistance tools from machine translation to large language models (LLMs) has changed how researchers write. This study investigates whether this shift is homogenizing research papers by analyzing native language identification (NLI) trends in ACL Anthology papers across three eras: pre-neural network (NN), pre-LLM, and post-LLM. We construct a labeled dataset using a semi-automated framework and fine-tune a classifier to detect linguistic fingerprints of author backgrounds. Our analysis shows a consistent decline in NLI performance over time. Interestingly, the post-LLM era reveals anomalies: while Chinese and French show unexpected resistance or divergent trends, Japanese and Korean exhibit sharper-than-expected declines.
Authors: Ali Slim, Haydar Hamieh, Jawad Kotaich, Yehya Ghosn, Mahdi Chehimi, Ammar Mohanna, Hasan Abed Al Kader Hammoud, Bernard Ghanem
Abstract: Large Language Models (LLMs) are increasingly used for code generation, yet quantum code generation is still evaluated mostly within single frameworks, making it difficult to separate quantum reasoning from framework familiarity. We introduce QuanBench+, a unified benchmark spanning Qiskit, PennyLane, and Cirq, with 42 aligned tasks covering quantum algorithms, gate decomposition, and state preparation. We evaluate models with executable functional tests, report Pass@1 and Pass@5, and use KL-divergence-based acceptance for probabilistic outputs. We additionally study Pass@1 after feedback-based repair, where a model may revise code after a runtime error or wrong answer. Across frameworks, the strongest one-shot scores reach 59.5% in Qiskit, 54.8% in Cirq, and 42.9% in PennyLane; with feedback-based repair, the best scores rise to 83.3%, 76.2%, and 66.7%, respectively. These results show clear progress, but also that reliable multi-framework quantum code generation remains unsolved and still depends strongly on framework-specific knowledge.
Authors: Pavel Golikov, Evgenii Opryshko, Gennady Pekhimenko, Mark C. Jeffrey
Abstract: While Large Language Models (LLMs) achieve high performance on standard mathematical benchmarks, their underlying reasoning processes remain highly overfit to standard textual formatting. We propose a perturbation pipeline consisting of 14 techniques to evaluate robustness of LLM reasoning. We apply this pipeline to AIME 2024 dataset and evalute 8 state-of-the-art models on the resulting benchmark. While frontier models exhibit resilience, open weights reasoning models suffer catastrophic collapses (up to 55% average accuracy drops across perturbations and up to 100% on some), exposing structural fragility. To further disentangle mechanical parsing failures from downstream reasoning failures, we strictly isolate the models' working memory capacity by forcing models to solve multiple unperturbed mathematical problems sequentially within a single context window. Our results indicate that open weight models ranging from 7B to 120B parameters and Claude Opus 4.6 exhibit accuracy decay on subsequent problems. This degradation demonstrates that intermediate reasoning steps permanently pollute standard dense attention mechanisms. We argue that to achieve reliable reasoning, future reasoning architectures must integrate explicit contextual resets within a model's own Chain-of-Thought, leading to fundamental open questions regarding the optimal granularity of atomic reasoning tasks.
Authors: Matheus Vin\'icius Todescato, Joel Lu\'is Carbonera
Abstract: Learning discriminative representations is a central goal of supervised deep learning. While cross-entropy (CE) remains the dominant objective for classification, it does not explicitly enforce desirable geometric properties in the embedding space, such as intra-class compactness and inter-class separation. Existing metric learning approaches, including supervised contrastive learning (SupCon) and proxy-based methods, address this limitation by operating on pairwise or proxy-based relationships, but often increase computational cost and complexity. In this work, we introduce Soft Silhouette Loss, a novel differentiable objective inspired by the classical silhouette coefficient from clustering analysis. Unlike pairwise objectives, our formulation evaluates each sample against all classes in the batch, providing a batch-level notion of global structure. The proposed loss directly encourages samples to be closer to their own class than to competing classes, while remaining lightweight. Soft Silhouette Loss can be seamlessly combined with cross-entropy, and is also complementary to supervised contrastive learning. We propose a hybrid objective that integrates them, jointly optimizing local pairwise consistency and global cluster structure. Extensive experiments on seven diverse datasets demonstrate that: (i) augmenting CE with Soft Silhouette Loss consistently improves over CE and other metric learning baselines; (ii) the hybrid formulation outperforms SupCon alone; and (iii) the combined method achieves the best performance, improving average top-1 accuracy from 36.71% (CE) and 37.85% (SupCon2) to 39.08%, while incurring substantially lower computational overhead. These results suggest that classical clustering principles can be reinterpreted as differentiable objectives for deep learning, enabling efficient optimization of both local and global structure in representation spaces.
Authors: Rasched Haidari, Sam Martin, Maxime Allard
Abstract: Large Genomic Foundation Models have recently achieved remarkable results and in-vivo translation capabilities. However these models quickly grow to over a few Billion of parameters and are expensive to run when compute is limited. To overcome this challenge, we present a distillation framework for transferring mRNA representations from a state of the art genomic foundation model into a much smaller model specialized for mRNA sequences, reducing the size by 200-fold. Embedding-level distillation worked better than logit based methods, which we found unstable. Benchmarking on mRNA-bench demonstrates that the distilled model achieves state-of-the-art performance among models of comparable size and competes with larger architectures for mRNA-related tasks. Our results highlight embedding-based distillation of mRNA sequences as an effective training strategy for biological foundation models. This enables similar efficient and scalable sequence modelling in genomics, particularly when large models are computationally challenging or infeasible.
Authors: Syed Rameez Naqvi, Lu Peng
Abstract: Molecular generative models must jointly ensure validity, diversity, and property control, yet existing approaches typically trade off among these objectives. We present MOLPAQ, a modular quantum-classical generator that assembles molecules from quantum-generated latent patches. A \b{eta}-VAE pretrained on QM9 learns a chemically aligned latent manifold; a reduced conditioner maps molecular descriptors into this space; and a parameter-efficient quantum patch generator produces entangled node embeddings that a valence-aware aggregator reconstructs into valid molecular graphs. Adversarial fine-tuning with a latent critic and chemistry-shaped reward yields 100\% RDKit validity, 99.75\% novelty, and 0.905 diversity. Beyond aggregate metrics, the pretrained quantum generator, steered by the conditioner, improves mean QED by approx. 2.3\% and increases aromatic motif incidence by approx. 10-12\% relative to a parameter-matched classical generator, highlighting its role as a compact topology-shaping operator.
Authors: Daniel Benniah John
Abstract: Sixth-generation (6G) wireless networks must support heterogeneous services: enhanced Mobile Broadband (eMBB) requiring 1 Tbps data rates, massive Machine-Type Communications (mMTC) supporting 10 million devices per km, and Ultra-Reliable Low-Latency Communications (URLLC) with 0.1-1 ms latency. Current resource allocation suffers from three limitations: (1) semantic blindness wasting 35% bandwidth on redundant data, (2) discrete action quantization, and (3) limited training diversity. This paper proposes GAN-DDPG, a Generative Adversarial Network-enhanced Deep Deterministic Policy Gradient framework integrating conditional GANs for traffic synthesis, continuous action DDPG, and semantic-aware reward optimization. Extensive simulations with statistical validation demonstrate significant improvements: 22% URLLC, 20% eMBB, 25% mMTC spectral efficiency gains (all p < 0.001) compared to baseline DDPG, with 18% latency and 31% packet loss reduction.
Authors: Yeping Jin, Jiaming Hu, Ioannis Ch. Paschalidis
Abstract: Large Language Models (LLMs) tend to respond correctly to prompts that align to the data they were trained and fine-tuned on. Yet, small shifts in wording, format, or language can trigger surprisingly large failures, especially on multi-step reasoning problems. To address this problem, we propose a Distributionally Robust Token Optimization (DRTO) approach, which combines token-level Reinforcement Learning from Human Feedback (RLHF) with Distributionally Robust Optimization (DRO). DRTO bounds worst case token-wise rewards by constructing an f-divergence ambiguity set over a loss minibatch, leading to a theoretical robustness. Empirically, DRTO enhances consistency under distribution shifts in mathematical reasoning benchmarks, achieving 9.17\% improvement on GSM8K and 2.49% improvement on MathQA.
Authors: Phong Lam, Ha-Linh Nguyen, Thu-Trang Nguyen, Son Nguyen, Hieu Dinh Vo
Abstract: High-quality labeled data is critical for training reliable machine learning and deep learning models, yet manual annotation remains costly and error-prone. Programmatic labeling addresses this challenge by using label functions (LFs), i.e., heuristic rules that automatically generate weak labels for training datasets. However, existing automated LF generation methods either rely on large language models (LLMs) to synthesize surface-level heuristics or employ model-based synthesis over hand-crafted primitives. These approaches often result in limited coverage and unreliable label quality. In this paper, we introduce EXPONA, an automated framework for programmatic labeling that formulates LF generation as a principled process balancing diversity and reliability. EXPONA systematically explores multi-level LFs, spanning surface, structural, and semantic perspectives. EXPONA further applies reliability-aware mechanisms to suppress noisy or redundant heuristics while preserving complementary signals. To evaluate EXPONA, we conducted extensive experiments on eleven classification datasets across diverse domains. Experimental results show that EXPONA consistently outperformed state-of-the-art automated LF generation methods. Specifically, EXPONA achieved nearly complete label coverage (up to 98.9%), improved weak label quality by up to 87%, and yielded downstream performance gains of up to 46% in weighted F1. These results indicate that EXPONA's combination of multi-level LF exploration and reliability-aware filtering enabled more consistent label quality and downstream performance across diverse tasks by balancing coverage and precision in the generated LF set.
Authors: Krisanu Sarkar
Abstract: We study cross-modal alignment between independently pretrained vision (DINOv2) and language (all-MiniLM-L6-v2) encoders using the functional map framework from computational geometry, which represents correspondence between representation manifolds as a compact linear operator between graph Laplacian eigenbases. While the framework underperforms Procrustes alignment and relative representations for cross-modal retrieval across all supervision budgets, it reveals a structural property of multimodal representations. We find that the Laplacian eigenvalue spectra of the two encoders are quantitatively similar (normalized spectral distance 0.043), indicating that independently trained models develop manifolds of comparable intrinsic complexity. However, the functional map exhibits near-zero diagonal dominance (mean below 0.05) and large orthogonality error (70.15), showing that the eigenvector bases are effectively unaligned. We term this decoupling the spectral complexity--orientation gap: models converge in how much structure they capture but not in how they organize it. This gap defines a boundary condition for spectral alignment methods and motivates three diagnostic quantities : diagonal dominance, orthogonality deviation, and Laplacian commutativity error for characterizing cross-modal representation compatibility.
Authors: Jun Liu, Ying Chen, Ziqian Lu, Qinyue Tong, Jun Tang
Abstract: Multivariate Time Series Anomaly Detection (MTSAD) is critical for real-world monitoring scenarios such as industrial control and aerospace systems. Mainstream reconstruction-based anomaly detection methods suffer from two key limitations: first, overfitting to spurious correlations induced by an overemphasis on cross-variable modeling; second, the generation of misleading anomaly scores by simply summing up multivariable reconstruction errors, which makes it difficult to distinguish between hard-to-reconstruct samples and genuine anomalies. To address these issues, we propose DBR-AF, a novel framework that integrates a dual-branch reconstruction (DBR) encoder and an autoregressive flow (AF) module. The DBR encoder decouples cross-variable correlation learning and intra-variable statistical property modeling to mitigate spurious correlations, while the AF module employs multiple stacked reversible transformations to model the complex multivariate residual distribution and further leverages density estimation to accurately identify normal samples with large reconstruction errors. Extensive experiments on seven benchmark datasets demonstrate that DBR-AF achieves state-of-the-art performance, with ablation studies validating the indispensability of its core components.
Authors: Chuxu Song, Zhencan Peng, Jiuqi Wei, Chuanhui Yang
Abstract: Long-context LLMs increasingly rely on extended, reusable prefill prompts for agents and domain Q&A, pushing attention and KV-cache to become the dominant decode-time bottlenecks. While sparse attention reduces computation and transfer costs, it often struggles to maintain accuracy at high sparsity levels due to the inherent distribution shift between Queries and Keys. We propose Centroid-Scoring Attention (CSAttention), a training-free sparse attention method optimized for high-throughput serving of reusable contexts. CSAttention adopts a storage-for-computation strategy tailored to the offline-prefill/online-decode setting: it front-loads computation into a one-time offline prefill phase that can be amortized across multiple queries, while aggressively optimizing per-step decoding latency. Specifically, CSAttention constructs query-centric lookup tables during offline prefill, whose size remains fixed during decoding, and enables online decoding to replace full-context scans with efficient table lookups and GPU-friendly score accumulation. Extensive experiments demonstrate that CSAttention achieves near-identical accuracy to full attention. Under high sparsity (95%) and long-context settings (32K-128K), CSAttention consistently outperforms state-of-the-art sparse attention methods in both model accuracy and inference speed, achieving up to 4.6x inference speedup over the most accurate baseline at a context length of 128K.
Authors: Jianxin Yan, Zeheng Qian, Wangze Ni, Zhitao Shen, Zhiping Wang, Haoyang Li, Jia Zhu, Lei Chen, Kui Ren
Abstract: Cache fusion accelerates generation process of LLMs equipped with RAG through KV caching and selective token recomputation, thereby reducing computational costs and improving efficiency. However, existing methods primarily rely on local perspectives for token selection and lack global awareness from the user query. Utilizing this global awareness is challenging due to the high cost of obtaining context-aware query representations and the strict pipeline constraints required for efficient attention analysis. Thus, this demonstration introduces QCFuse, an innovative KV cache fusion system centered on the user query. QCFuse leverages semantic summary anchors to enhance query representations and selectively recomputes query-related tokens to improve accuracy, updating tokens based on the attention distribution of the most critical Transformer layer to preserve the high efficiency of the pipeline structure. Evaluations on real-world datasets demonstrate that QCFuse significantly improves the response efficiency of LLMs by 40\% while maintaining equivalent accuracy compared to current methods. Additionally, in certain scenarios, QCFuse achieves an attention denoising effect that yields higher response accuracy, demonstrating substantial potential in the optimization of LLM inference.
Authors: David Ramos, Lucas Lacasa, Ferm\'in Guti\'errez, Eusebio Valero, Gonzalo Rubio
Abstract: Computational fluid dynamics (CFD) provides high-fidelity simulations of fluid flows but remains computationally expensive for many-query applications. In recent years deep learning (DL) has been used to construct data-driven fluid-dynamic surrogate models. In this work we consider a different learning paradigm and embrace generative modelling as a framework for constructing scalable fluid-dynamics surrogate models. We introduce FluidFlow, a generative model based on conditional flow-matching, a recent alternative to diffusion models that learns deterministic transport maps between noise and data distributions. FluidFlow is specifically designed to operate directly on CFD data defined on both structured and unstructured meshes alike, without the needs to perform any mesh interpolation pre-processing and preserving geometric fidelity. We assess the capabilities of FluidFlow using two different core neural network architectures, a U-Net and diffusion transformer (DiT), and condition their learning on physically meaningful parameters. The methodology is validated on two benchmark problems of increasing complexity: prediction of pressure coefficients along an airfoil boundary across different operating conditions, and prediction of pressure and friction coefficients over a full three-dimensional aircraft geometry discretized on a large unstructured mesh. In both cases, FluidFlow outperform strong multilayer perceptron baselines, achieving significantly lower error metrics and improved generalisation across operating conditions. Notably, the transformer-based architecture enables scalable learning on large unstructured datasets while maintaining high predictive accuracy. These results demonstrate that flow-matching generative models provide an effective and flexible framework for surrogate modelling in fluid dynamics, with potential for realistic engineering and scientific applications.
Authors: Matthew DosSantos DiSorbo, Harang Ju
Abstract: Effective automation hinges on deciding when to act and when to escalate. We model this as a decision under uncertainty: an LLM forms a prediction, estimates its probability of being correct, and compares the expected costs of acting and escalating. Using this framework across five domains of recorded human decisions-demand forecasting, content recommendation, content moderation, loan approval, and autonomous driving-and across multiple model families, we find marked differences in the implicit thresholds models use to trade off these costs. These thresholds vary substantially and are not predicted by architecture or scale, while self-estimates are miscalibrated in model-specific ways. We then test interventions that target this decision process by varying cost ratios, providing accuracy signals, and training models to follow the desired escalation rule. Prompting helps mainly for reasoning models. SFT on chain-of-thought targets yields the most robust policies, which generalize across datasets, cost ratios, prompt framings, and held-out domains. These results suggest that escalation behavior is a model-specific property that should be characterized before deployment, and that robust alignment benefits from training models to reason explicitly about uncertainty and decision costs.
Authors: Brendan R. Hogan, Xiwen Chen, James T. Wilson, Kashif Rasul, Adel Boyarsky, Thomas Kamei, Anderson Schneider, Yuriy Nevmyvaka
Abstract: We present AlphaLab, an autonomous research harness that leverages frontier LLM agentic capabilities to automate the full experimental cycle in quantitative, computation-intensive domains. Given only a dataset and a natural-language objective, AlphaLab proceeds through three phases without human intervention: (1) it adapts to the domain and explores the data, writing analysis code and producing a research report; (2) it constructs and adversarially validates its own evaluation framework; and (3) it runs large-scale GPU experiments via a Strategist/Worker loop, accumulating domain knowledge in a persistent playbook that functions as a form of online prompt optimization. All domain-specific behavior is factored into adapters generated by the model itself, so the same pipeline handles qualitatively different tasks without modification. We evaluate AlphaLab with two frontier LLMs (GPT-5.2 and Claude Opus 4.6) on three domains: CUDA kernel optimization, where it writes GPU kernels that run 4.4x faster than torch.compile on average (up to 91x); LLM pretraining, where the full system achieves 22% lower validation loss than a single-shot baseline using the same model; and traffic forecasting, where it beats standard baselines by 23-25% after researching and implementing published model families from the literature. The two models discover qualitatively different solutions in every domain (neither dominates uniformly), suggesting that multi-model campaigns provide complementary search coverage. We additionally report results on financial time series forecasting in the appendix, and release all code at https://brendanhogan.github.io/alphalab-paper/.
Authors: Ivan Viakhirev, Kirill Borodin, Grach Mkrtchian
Abstract: Hallucinations in large ASR models present a critical safety risk. In this work, we propose the \textit{Spectral Sensitivity Theorem}, which predicts a phase transition in deep networks from a dispersive regime (signal decay) to an attractor regime (rank-1 collapse) governed by layer-wise gain and alignment. We validate this theory by analyzing the eigenspectra of activation graphs in Whisper models (Tiny to Large-v3-Turbo) under adversarial stress. Our results confirm the theoretical prediction: intermediate models exhibit \textit{Structural Disintegration} (Regime I), characterized by a $13.4\%$ collapse in Cross-Attention rank. Conversely, large models enter a \textit{Compression-Seeking Attractor} state (Regime II), where Self-Attention actively compresses rank ($-2.34\%$) and hardens the spectral slope, decoupling the model from acoustic evidence.
Authors: Junjie Wang, Xianyang Gan, Dan Liu, Jingxian He, Stefania Ferraro, Keith M. Kendrick, Weihua Zhao, Shuxia Yao, Christian Montag, Benjamin Becker
Abstract: The widespread adoption of generative artificial intelligence conversational agents (AICAs) among university students constitutes a novel cognitive social environment whose impact on the maturing brain remains elusive. Combining surveys with high resolution structural MRI, we examined patterns of general, functional, and socio emotional AICA use, academic performance, mental health, and brain structural signatures in a comparatively large sample of 222 young individuals. Across computational anatomy, meta analytic network level, and behavioral decoding analyses, we observed use specific associations. Higher general and functional AICA use frequencies were linked to better academic outcomes (GPA), larger dorsolateral prefrontal and calcarine gray matter volume, and enhanced hippocampal network clustering and local efficiency. In contrast, more frequent socio emotional AICA use was associated with poorer mental health (depression, social anxiety) and lower volume of superior temporal and amygdalar regions central to social and affective processing. These findings indicate that the same class of AI tools exerts distinct effects depending on usage patterns and motivations, engaging prefrontal hippocampal systems that support cognition versus socio emotional systems that may track distress linked usage. These heterogeneities are crucial for designing environments that harness the educational benefits of AI while mitigating mental health risks.
Authors: Aleksandr Meshkov
Abstract: Existing evaluation methods for LLM-based AI systems, such as LLM-as-a-Judge, verdict systems, and NLI, do not always align well with human assessment because they cannot adapt their strictness to the application domain. This paper presents Temperature-Controlled Verdict Aggregation (TCVA), a method that combines a five-level verdict-scoring system with generalized power-mean aggregation and an intuitive temperature parameter T [0.1, 1.0] to control evaluation rigor. Low temperatures yield pessimistic scores suited for safety-critical domains; high temperatures produce lenient scores appropriate for conversational AI. Experimental evaluation on three benchmark datasets with human Likert-scale annotations (SummEval and USR) shows that TCVA achieves correlation with human judgments comparable to RAGAS on faithfulness (Spearman = 0.667 vs. 0.676) while consistently outperforming DeepEval. The method requires no additional LLM calls when adjusting the temperature parameter.
Authors: Wenxiao Zhang, Yu Liu, Qiang sun, Yihao Ding, Sirui Li, Yanbing Liu, Jin B. Hong, Wei Liu
Abstract: Extracting structured knowledge from unstructured data still faces practical limitations: entity and event extraction pipelines remain brittle, knowledge graph construction requires costly ontology engineering, and cross-domain generalization is rarely production-ready. In contrast, space and time provide universal contextual anchors that naturally align heterogeneous information and benefit downstream tasks such as retrieval and reasoning. We introduce \textbf{STIndex}, an end-to-end system that structures unstructured content into a multidimensional spatiotemporal data warehouse. Users define domain-specific analysis dimensions with configurable hierarchies, while large language models perform context-aware extraction and grounding. \textbf{STIndex} integrates document-level memory, geocoding correction, and quality validation, and offers an interactive analytics dashboard for visualization, clustering, burst detection, and entity network analysis. In evaluation on a public health benchmark, \textbf{STIndex} improves spatiotemporal entity extraction F1 by 4.37\% (GPT-4o-mini) and 3.60\% (Qwen3-8B). A live demonstration and open-source code are available at https://stindex.ai4wa.com/dashboard.
Authors: Yuki Kataoka, Masahiro Banno, Michihito Kyo, Shuri Nakao, Tomoo Sato, Shunsuke Taito, Tomohiro Takayama, Takahiro Tsuge, Yasushi Tsujimoto, Ryuhei So, Toshi A. Furukawa
Abstract: Background: Server-based screening tools impose subscription costs, while open-source alternatives require coding skills. Objectives: We developed a browser extension that provides no-code, serverless artificial intelligence (AI)-assisted title and abstract screening and examined its functionality. Methods: TiAb Review Plugin is an open-source Chrome browser extension (available at https://chromewebstore.google.com/detail/tiab-review-plugin/alejlnlfflogpnabpbplmnojgoeeabij). It uses Google Sheets as a shared database, requiring no dedicated server and enabling multi-reviewer collaboration. Users supply their own Gemini API key, stored locally and encrypted. The tool offers three screening modes: manual review, large language model (LLM) batch screening, and machine learning (ML) active learning. For ML evaluation, we re-implemented the default ASReview active learning algorithm (TF-IDF with Naive Bayes) in TypeScript to enable in-browser execution, and verified equivalence against the original Python implementation using 10-fold cross-validation on six datasets. For LLM evaluation, we compared 16 parameter configurations across two model families on a benchmark dataset, then validated the optimal configuration (Gemini 3.0 Flash, low thinking budget, TopP=0.95) with a sensitivity-oriented prompt on five public datasets (1,038 to 5,628 records, 0.5 to 2.0 percent prevalence). Results: The TypeScript classifier produced top-100 rankings 100 percent identical to the original ASReview across all six datasets. For LLM screening, recall was 94 to 100 percent with precision of 2 to 15 percent, and Work Saved over Sampling at 95 percent recall (WSS@95) ranged from 48.7 to 87.3 percent. Conclusions: We developed a functional browser extension that integrates LLM screening and ML active learning into a no-code, serverless environment, ready for practical use in systematic review screening.
URLs: https://chromewebstore.google.com/detail/tiab-review-plugin/alejlnlfflogpnabpbplmnojgoeeabij).
Authors: Abhimanyu Pallavi Sudhir, Long Tran-Thanh
Abstract: One of the impediments to the efficiency of information markets is the inherent information asymmetry present in them, exacerbated by the "buyer's inspection paradox" (the buyer cannot mitigate the asymmetry by "inspecting" the information, because in doing so the buyer obtains the information without paying for it). Previous work has suggested that using Large Language Model (LLM) buyers to inspect and purchase information could overcome this information asymmetry, as an LLM buyer can simply "forget" the information it inspects. In this work, we analyze this mechanism formally through a "value-of-information" paradigm, i.e. whether it incentivizes information to be priced and provided in accordance with its "true value". We focus in particular on our new recursive version of the mechanism, which we believe has a range of applications including in AI alignment research, where it is related to Extrapolated Volition and Scalable Oversight.
Authors: H. Xu, B. He, S. Wang
Abstract: Precise interference detection and identification are crucial for enhancing the survivability of communication systems in non-cooperative wireless environments. While deep learning (DL) has advanced this field, existing single-task learning (STL) approaches neglect inherent task correlations. Furthermore, emerging multi-task learning (MTL) methods often lack a theoretical foundation for quantifying and modeling task relationships. To bridge this gap, we establish a theoretically grounded MTL framework for joint interference detection, modulation identification, and interference identification. First, we derive an upper bound for the weighted expected loss in MTL frameworks. This bound explicitly connects MTL performance to task similarity, quantified by the Wasserstein distance and learnable task relation coefficients. Guided by this theory, we present the adversarial multi-task interference detection and identification network (AMTIDIN), which integrates adversarial training to minimize distributional discrepancies across tasks and uses adaptive coefficients to model task correlations dynamically. Crucially, we conducted a quantitative analysis of task similarity to reveal intrinsic task relationships, specifically that modulation identification and interference identification share a substantial feature overlap distinct from interference detection. Extensive comparative experiments demonstrate that AMTIDIN significantly outperforms both its task-specific STL baseline and state-of-the-art MTL baselines in robustness and generalization, particularly under challenging conditions with limited training data, short signal lengths, and low signal-to-noise ratios (SNRs).
Authors: Tanzim Ahad, Ismail Hossain, Md Jahangir Alam, Sai Puppala, Yoonpyo Lee, Syed Bahauddin Alam, Sajedul Talukder
Abstract: We introduce Semantic Intent Fragmentation (SIF), an attack class against LLM orchestration systems where a single, legitimately phrased request causes an orchestrator to decompose a task into subtasks that are individually benign but jointly violate security policy. Current safety mechanisms operate at the subtask level, so each step clears existing classifiers -- the violation only emerges at the composed plan. SIF exploits OWASP LLM06:2025 through four mechanisms: bulk scope escalation, silent data exfiltration, embedded trigger deployment, and quasi-identifier aggregation, requiring no injected content, no system modification, and no attacker interaction after the initial request. We construct a three-stage red-teaming pipeline grounded in OWASP, MITRE ATLAS, and NIST frameworks to generate realistic enterprise scenarios. Across 14 scenarios spanning financial reporting, information security, and HR analytics, a GPT-20B orchestrator produces policy-violating plans in 71% of cases (10/14) while every subtask appears benign. Three independent signals validate this: deterministic taint analysis, chain-of-thought evaluation, and a cross-model compliance judge with 0% false positives. Stronger orchestrators increase SIF success rates. Plan-level information-flow tracking combined with compliance evaluation detects all attacks before execution, showing the compositional safety gap is closable.
Authors: Ponkoj Chandra Shill
Abstract: Digital forensic investigations increasingly rely on heterogeneous evidence such as images, scanned documents, and contextual reports. These artifacts may contain explicit or implicit expressions of harm, hate, threat, violence, or intimidation, yet existing automated approaches often assume clean text input or apply vision models without forensic justification. This paper presents a case-driven multimodal approach for hate and threat detection in forensic analysis. The proposed framework explicitly determines the presence and source of textual evidence, distinguishing between embedded text, associated contextual text, and image-only evidence. Based on the identified evidence configuration, the framework selectively applies text analysis, multimodal fusion, or image-only semantic reasoning using vision language models with vision transformer backbones (ViT). By conditioning inference on evidence availability, the approach mirrors forensic decision-making, improves evidentiary traceability, and avoids unjustified modality assumptions. Experimental evaluation on forensic-style image evidence demonstrates consistent and interpretable behavior across heterogeneous evidence scenarios.
Authors: Xingming Liao, Ning Chen, Muying Shu, Yunpeng Yin, Peijian Zeng, Zhuowei Wang, Nankai Lin, Lianglun Cheng
Abstract: Fine-grained visual understanding and high-level reasoning in real-world open-water environments remain under-explored due to the lack of dedicated benchmarks. We introduce MARINER, a comprehensive benchmark built under the novel Entity-Environment-Event (3E) paradigm. MARINER contains 16,629 multi-source maritime images with 63 fine-grained vessel categories, diverse adverse environments, and 5 typical dynamic maritime incidents, covering fine-grained classification, object detection, and visual question answering tasks. We conduct extensive evaluations on mainstream Multimodal Large language models (MLLMs) and establish baselines, revealing that even advanced models struggle with fine-grained discrimination and causal reasoning in complex marine scenes. As a dedicated maritime benchmark, MARINER fills the gap of realistic and cognitive-level evaluation for maritime multimodal understanding, and promotes future research on robust vision-language models for open-water applications. Appendix and supplementary materials are available at https://lxixim.github.io/MARINER.
Authors: Zhuang Qi, Ying-Peng Tang, Lei Meng, Guoqing Chao, Lei Wu, Han Yu, Xiangxu Meng
Abstract: Exemplar replay has become an effective strategy for mitigating catastrophic forgetting in federated continual learning (FCL) by retaining representative samples from past tasks. Existing studies focus on designing sample-importance estimation mechanisms to identify information-rich samples. However, they typically overlook strategies for effectively utilizing the selected exemplars, which limits their performance under continual dynamic heterogeneity across clients and tasks. To address this issue, this paper proposes a Federated gEometry-Aware correcTion method, termed FEAT, which alleviates imbalance-induced representation collapse that drags rare-class features toward frequent classes across clients. Specifically, it consists of two key modules: 1) the Geometric Structure Alignment module performs structural knowledge distillation by aligning the pairwise angular similarities between feature representations and their corresponding Equiangular Tight Frame prototypes, which are fixed and shared across clients to serve as a class-discriminative reference structure. This encourages geometric consistency across tasks and helps mitigate representation drift; 2) the Energy-based Geometric Correction module removes task-irrelevant directional components from feature embeddings, which reduces prediction bias toward majority classes. This improves sensitivity to minority classes and enhances the model's robustness under class-imbalanced distributions.
Authors: Xingyan Liu, Xiyue Luo, Linyu Li, Ganghong Huang, Jianfeng Liu, Honglin Qiao
Abstract: Deploying LLM-powered agents in enterprise scenarios such as cloud technical support demands high-quality, domain-specific skills. However, existing skill creators lack domain grounding, producing skills poorly aligned with real-world task requirements. Moreover, once deployed, there is no systematic mechanism to trace execution failures back to skill deficiencies and drive targeted refinements, leaving skill quality stagnant despite accumulating operational evidence. We introduce SkillForge, a self-evolving framework that closes an end-to-end creation-evaluation-refinement loop. To produce well-aligned initial skills, a Domain-Contextualized Skill Creator grounds skill synthesis in knowledge bases and historical support tickets. To enable continuous self-optimization, a three-stage pipeline -- Failure Analyzer, Skill Diagnostician, and Skill Optimizer -- automatically diagnoses execution failures in batch, pinpoints the underlying skill deficiencies, and rewrites the skill to eliminate them. This cycle runs iteratively, allowing skills to self-improve with every round of deployment feedback. Evaluated on five real-world cloud support scenarios spanning 1,883 tickets and 3,737 tasks, experiments show that: (1) the Domain-Contextualized Skill Creator produces substantially better initial skills than the generic skill creator, as measured by consistency with expert-authored reference responses from historical tickets; and (2) the self-evolution loop progressively improves skill quality from diverse starting points (including expert-authored, domain-created, and generic skills) across successive rounds, demonstrating that automated evolution can surpass manually curated expert knowledge.
Authors: Ivo Nowak
Abstract: Reinforcement learning is typically treated as a uniform, data-driven optimization process, where updates are guided by rewards and temporal-difference errors without explicitly exploiting global structure. In contrast, dynamic programming methods rely on structured information propagation, enabling efficient and stable learning. In this paper, we provide evidence that such structure can be recovered from the learning dynamics of distributional reinforcement learning. By analyzing the temporal evolution of return distributions, we identify signals that capture when and where learning occurs in the state space. In particular, we introduce a temporal learning indicator t*(s) that reflects when a state undergoes its strongest learning update during training. Empirically, this signal induces an ordering over states that is consistent with a dynamic programming-style propagation of information. Building on this observation, we propose StructRL, a framework that exploits these signals to guide sampling in alignment with the emerging propagation structure. Our preliminary results suggest that distributional learning dynamics provide a mechanism to recover and exploit dynamic programming-like structure without requiring an explicit model. This offers a new perspective on reinforcement learning, where learning can be interpreted as a structured propagation process rather than a purely uniform optimization procedure.
Authors: Yesmine Abdennadher, Philip N. Garner
Abstract: Spiking Neural Networks (SNNs) are naturally suited for speech processing tasks due to their specific dynamics, which allows them to handle temporal data. However, the threshold-based generation of spikes in SNNs intuitively causes an angular or irregular predictive landscape. We explore the effect of using the Bayesian learning approach for the weights on the irregular predictive landscape. For the surrogate-gradient SNNs, we also explore the application of the Improved Variational Online Newton (IVON) approach, which is an efficient variational approach. The performance of the proposed approach is evaluated on the Heidelberg Digits and Speech Commands datasets. The hypothesis is that the Bayesian approach will result in a smoother and more regular predictive landscape, given the angular nature of the deterministic predictive landscape. The experimental evaluation of the proposed approach shows improved performance on the negative log-likelihood and Brier score. Furthermore, the proposed approach has resulted in a smoother and more regular predictive landscape compared to the deterministic approach, based on the one-dimensional slices of the weight space
Authors: Yongchan Chun, Chanhee Park, Jeongho Yoon, Jaehyung Seo, Heuiseok Lim
Abstract: Pretrained models have become standard in both vision and language, yet they typically do not provide reliable measures of confidence. Existing uncertainty estimation methods, such as deep ensembles and MC dropout, are often too computationally expensive to deploy in practice. Evidential Deep Learning (EDL) offers a more efficient alternative, but it requires models to be trained to output evidential quantities from the start, which is rarely true for pretrained networks. To enable EDL-style uncertainty estimation in pretrained models, we propose the Evidential Transformation Network (ETN), a lightweight post-hoc module that converts a pretrained predictor into an evidential model. ETN operates in logit space: it learns a sample-dependent affine transformation of the logits and interprets the transformed outputs as parameters of a Dirichlet distribution for uncertainty estimation. We evaluate ETN on image classification and large language model question-answering benchmarks under both in-distribution and out-of-distribution settings. ETN consistently improves uncertainty estimation over post-hoc baselines while preserving accuracy and adding only minimal computational overhead.
Authors: Yeseul E. Chang, Rahul Kailasa, Simon Shim, Byunghoon Oh, Jaewoo Lee
Abstract: Unauthorized disclosure of confidential documents demands robust, low-leakage classification. In real work environments, there is a lot of inflow and outflow of documents. To continuously update knowledge, we propose a methodology for classifying confidential documents using Retrieval Augmented Classification (RAC). To confirm this effectiveness, we compare RAC and supervised fine tuning (FT) on the WikiLeaks US Diplomacy corpus under realistic sequence-length constraints. On balanced data, RAC matches FT. On unbalanced data, RAC is more stable while delivering comparable performance--about 96% Accuracy on both the original (unbalanced) and augmented (balanced) sets, and up to 94% F1 with proper prompting--whereas FT attains 90% F1 trained on the augmented, balanced set but drops to 88% F1 trained on the original, unbalanced set. When robust augmentation is infeasible, RAC provides a practical, security-preserving path to strong classification by keeping sensitive content out of model weights and under your control, and it remains robust as real-world conditions change in class balance, data, context length, or governance requirements. Because RAC grounds decisions in an external vector store with similarity matching, it is less sensitive to label skew, reduces parameter-level leakage, and can incorporate new data immediately via reindexing--a difficult step for FT, which typically requires retraining. The contributions of this paper are threefold: first, a RAC-based classification pipeline and evaluation recipe; second, a controlled study that isolates class imbalance and context-length effects for FT versus RAC in confidential-document grading; and third, actionable guidance on RAC design patterns for governed deployments.
Authors: Jihwan Yoon, Taemoon Jeong, Jeongeun Park, Chanwoo Kim, Jaewoon Kwon, Yonghyeon Lee, Kyungjae Lee, Sungjoon Choi
Abstract: Designing robot morphologies and kinematics has traditionally relied on human intuition, with little systematic foundation. Motion-design co-optimization offers a promising path toward automation, but two major challenges remain: (i) the vast, unstructured design space and (ii) the difficulty of constructing task-specific loss functions. We propose a new paradigm that minimizes human involvement by (i) learning the design search space from existing mechanical designs, rather than hand-crafting it, and (ii) defining the loss directly from human motion data via motion retargeting and Procrustes analysis. Using screw-theory-based joint axis representation and isometric manifold learning, we construct a compact, geometry-preserving latent space of humanoid upper body designs in which optimization is tractable. We then solve design optimization in this latent space using gradient-free optimization. Our approach establishes a principled framework for data-driven robot design and demonstrates that leveraging existing designs and human motion can effectively guide the automated discovery of novel robot design.
Authors: Rahul D Ray, Utkarsh Srivastava
Abstract: Uncertainty quantification (UQ) is essential for deploying deep learning models in safety critical applications, yet no consensus exists on which UQ method performs best across different data modalities and distribution shifts. This paper presents a comprehensive benchmark of ten widely used UQ baselines including MC Dropout, SWAG, ensemble methods, temperature scaling, energy based OOD, Mahalanobis, hyperbolic classifiers, ENN, Taylor Sensus, and split conformal prediction against a simplified yet highly effective variant of VOLTA that retains only a deep encoder, learnable prototypes, cross entropy loss, and post hoc temperature scaling. We evaluate all methods on CIFAR 10 (in distribution), CIFAR 100, SVHN, uniform noise (out of distribution), CIFAR 10 C (corruptions), and Tiny ImageNet features (tabular). VOLTA achieves competitive or superior accuracy (up to 0.864 on CIFAR 10), significantly lower expected calibration error (0.010 vs. 0.044 to 0.102 for baselines), and strong OOD detection (AUROC 0.802). Statistical testing over three random seeds shows that VOLTA matches or outperforms most baselines, with ablation studies confirming the importance of adaptive temperature and deep encoders. Our results establish VOLTA as a lightweight, deterministic, and well calibrated alternative to more complex UQ approaches.
Authors: Ruixiang Jiang, Changwen Chen
Abstract: Interpretation is essential to deciphering the language of art: audiences communicate with artists by recovering meaning from visual artifacts. However, current Generative Art (GenArt) evaluators remain fixated on surface-level image quality or literal prompt adherence, failing to assess the deeper symbolic or abstract meaning intended by the creator. We address this gap by formalizing a Peircean computational semiotic theory that models Human-GenArt Interaction (HGI) as cascaded semiosis. This framework reveals that artistic meaning is conveyed through three modes - iconic, symbolic, and indexical - yet existing evaluators operate heavily within the iconic mode, remaining structurally blind to the latter two. To overcome this structural blindness, we propose SemJudge. This evaluator explicitly assesses symbolic and indexical meaning in HGI via a Hierarchical Semiosis Graph (HSG) that reconstructs the meaning-making process from prompt to generated artifact. Extensive quantitative experiments show that SemJudge aligns more closely with human judgments than prior evaluators on an interpretation-intensive fine-art benchmark. User studies further demonstrate that SemJudge produces deeper, more insightful artistic interpretations, thereby paving the way for GenArt to move beyond the generation of "pretty" images toward a medium capable of expressing complex human experience. Project page: https://github.com/songrise/SemJudge.
Authors: Makanjuola Ogunleye, Eman Abdelrahman, Ismini Lourentzou
Abstract: Large multimodal models are increasingly used as the reasoning core of embodied agents operating in 3D environments, yet they remain prone to hallucinations that can produce unsafe and ungrounded decisions. Existing inference-time hallucination mitigation methods largely target 2D vision-language settings and do not transfer to embodied 3D reasoning, where failures arise from object presence, spatial layout, and geometric grounding rather than pixel-level inconsistencies. We introduce 3D-VCD, the first inference-time visual contrastive decoding framework for hallucination mitigation in 3D embodied agents. 3D-VCD constructs a distorted 3D scene graph by applying semantic and geometric perturbations to object-centric representations, such as category substitutions and coordinate or extent corruption. By contrasting predictions under the original and distorted 3D contexts, our method suppresses tokens that are insensitive to grounded scene evidence and are therefore likely driven by language priors. We evaluate 3D-VCD on the 3D-POPE and HEAL benchmarks and show that it consistently improves grounded reasoning without any retraining, establishing inference-time contrastive decoding over structured 3D representations as an effective and practical route to more reliable embodied intelligence.
Authors: Tiejin Chen, Huaiyuan Yao, Jia Chen, Evangelos E. Papalexakis, Hua Wei
Abstract: While Large Language Model-based Multi-Agent Systems (MAS) consistently outperform single-agent systems on complex tasks, their intricate interactions introduce critical reliability challenges arising from communication dynamics and role dependencies. Existing Uncertainty Quantification methods, typically designed for single-turn outputs, fail to address the unique complexities of the MAS. Specifically, these methods struggle with three distinct challenges: the cascading uncertainty in multi-step reasoning, the variability of inter-agent communication paths, and the diversity of communication topologies. To bridge this gap, we introduce MATU, a novel framework that quantifies uncertainty through tensor decomposition. MATU moves beyond analyzing final text outputs by representing entire reasoning trajectories as embedding matrices and organizing multiple execution runs into a higher-order tensor. By applying tensor decomposition, we disentangle and quantify distinct sources of uncertainty, offering a comprehensive reliability measure that is generalizable across different agent structures. We provide comprehensive experiments to show that MATU effectively estimates holistic and robust uncertainty across diverse tasks and communication topologies.
Authors: Vrushank Ahire, Vivek Kurumanghat, Mudasir Ganaie, Lipika Kabiraj
Abstract: The disintegration of liquid sheets into ligaments and droplets involves highly transient, multi-scale dynamics that are difficult to quantify from high-speed shadowgraphy images. Identifying droplets, ligaments, and blobs formed during breakup, along with tracking across frames, is essential for spray analysis. However, conventional multi-object tracking frameworks impose strict one-to-one temporal associations and cannot represent one-to-many fragmentation events. In this study, we present a two-stage deep learning framework for object detection and temporal relationship modeling across frames. The framework captures ligament deformation, fragmentation, and parent-child lineage during liquid sheet disintegration. In the first stage, a Faster R-CNN with a ResNet-50 backbone and Feature Pyramid Network detects and classifies ligaments and droplets in high-speed shadowgraphy recordings of an impinging Carbopol gel jet. A morphology-preserving synthetic data generation strategy augments the training set without introducing physically implausible configurations, achieving a held-out F1 score of up to 0.872 across fourteen original-to-synthetic configurations. In the second stage, a Transformer-augmented multilayer perceptron classifies inter-frame associations into continuation, fragmentation (one-to-many), and non-association using physics-informed geometric features. Despite severe class imbalance, the model achieves 86.1% accuracy, 93.2% precision, and perfect recall (1.00) for fragmentation events. Together, the framework enables automated reconstruction of fragmentation trees, preservation of parent-child lineage, and extraction of breakup statistics such as fragment multiplicity and droplet size distributions. By explicitly identifying children droplets formed from ligament fragmentation, the framework provides automated analysis of the primary atomization mode.
Authors: Xinmiao Xiong, Bangya Liu, Hao Wang, Dayou Li, Nuo Chen, Andrew Feng, Mingyu Ding, Suman Banerjee, Yang Zhou, Zhiwen Fan
Abstract: Geometric Foundation Models (GFMs) have recently advanced monocular SLAM by providing robust, calibration-free 3D priors. However, deploying these models on dense video streams introduces significant computational redundancy. Current GFM-based SLAM systems typically rely on post hoc keyframe selection. Because of this, they must perform expensive dense geometric decoding simply to determine whether a frame contains novel geometry, resulting in late rejection and wasted computation. To mitigate this inefficiency, we propose LeanGate, a lightweight feed-forward frame-gating network. LeanGate predicts a geometric utility score to assess a frame's mapping value prior to the heavy GFM feature extraction and matching stages. As a predictive plug-and-play module, our approach bypasses over 90% of redundant frames. Evaluations on standard SLAM benchmarks demonstrate that LeanGate reduces tracking FLOPs by more than 85% and achieves a 5x end-to-end throughput speedup. Furthermore, it maintains the tracking and mapping accuracy of dense baselines.
Authors: Hao Shao, Letian Wang, Yang Zhou, Yuxuan Hu, Zhuofan Zong, Steven L. Waslander, Wei Zhan, Hongsheng Li
Abstract: Recent years have seen remarkable progress in autonomous driving, yet generalization to long-tail and open-world scenarios remains a major bottleneck for large-scale deployment. To address this challenge, some works use LLMs and VLMs for vision-language understanding and reasoning, enabling vehicles to interpret rare and safety-critical situations when generating actions. Others study generative world models to capture the spatio-temporal evolution of driving scenes, allowing agents to imagine possible futures before acting. Inspired by human intelligence, which unifies understanding and imagination, we explore a unified model for autonomous driving. We present LMGenDrive, the first framework that combines LLM-based multimodal understanding with generative world models for end-to-end closed-loop driving. Given multi-view camera inputs and natural-language instructions, LMGenDrive generates both future driving videos and control signals. This design provides complementary benefits: video prediction improves spatio-temporal scene modeling, while the LLM contributes strong semantic priors and instruction grounding from large-scale pretraining. We further propose a progressive three-stage training strategy, from vision pretraining to multi-step long-horizon driving, to improve stability and performance. LMGenDrive supports both low-latency online planning and autoregressive offline video generation. Experiments show that it significantly outperforms prior methods on challenging closed-loop benchmarks, with clear gains in instruction following, spatio-temporal understanding, and robustness to rare scenarios. These results suggest that unifying multimodal understanding and generation is a promising direction for more generalizable and robust embodied decision-making systems.
Authors: Meiziniu Li, Dongze Li, Jianmeng Liu, Shing-Chi Cheung
Abstract: Performance optimization of AI infrastructure is key to the fast adoption of large language models (LLMs). The PyTorch compiler (torch.compile), a core optimization tool for deep learning (DL) models (including LLMs), has received due attention. However, torch.compile is prone to correctness bugs, which cause incorrect outputs of compiled DL models without triggering exceptions, crashes, or warnings. These bugs pose a serious threat to the reliability of downstream LLM applications. Data from the PyTorch community shows that 19.2% of high-priority issues are incorrect outputs of compiled DL models induced by torch.compile bugs, the second-most-common bug category (only behind program crashes at 19.57%). However, no systematic study has been conducted to specifically characterize and thereby detect these bugs. In this paper, we present the first empirical study of the correctness bugs in torch.compile, examine their characteristics, and assess the effectiveness of existing fuzzers in detecting them. Based on our findings, we propose a proof-of-concept testing technique named AlignGuard, tailored specifically for detecting correctness bugs in torch.compile. AlignGuard incorporates bug characteristics distilled from our empirical study, applying LLM-based test mutation to existing test cases for correctness bug detection. At the time of writing, AlignGuard has successfully detected 23 new correctness bugs in recent torch.compile. All these bugs have been confirmed or fixed by the PyTorch development team, and over half (14/23) of them are even marked as high-priority bugs, underscoring the usefulness of our technique.
Authors: Adrian Manchado, Tanner Cellio, Jonathan Keane, Yiyang Wang
Abstract: Sport analysis is crucial for team performance since it provides actionable data that can inform coaching decisions, improve player performance, and enhance team strategies. To analyze more complex features from game footage, a computer vision model can be used to identify and track key entities from the field. We propose the use of an object detection and tracking system to predict player positioning throughout the game. To translate this to positioning in relation to the field dimensions, we use a point prediction model to identify key points on the field and combine these with known field dimensions to extract actual distances. For the player-identification model, object detection models like YOLO and Faster R-CNN are evaluated on the accuracy of our custom video footage using multiple different evaluation metrics. The goal is to identify the best model for object identification to obtain the most accurate results when paired with SAM2 (Segment Anything Model 2) for segmentation and tracking. For the key point detection model, we use a CNN model to find consistent locations in the soccer field. Through homography, the positions of points and objects in the camera perspective will be transformed to a real-ground perspective. The segmented player masks from SAM2 are transformed from camera perspective to real-world field coordinates through homography, regardless of camera angle or movement. The transformed real-world coordinates can be used to calculate valuable tactical insights including player speed, distance covered, positioning heatmaps, and more complex team statistics, providing coaches and players with actionable performance data previously unavailable from standard video analysis.
Authors: Chia-Hsuan Lee, Mingyang Zhou, Renkun Ni, Zelei Cheng, Sihui Dai, Supriyo Chakraborty, Shixiong Zhang, Sambit Sahu, William Campbell
Abstract: Preference optimization methods such as DPO and KTO are widely used for aligning language models, yet little is understood about what properties of preference data drive downstream reasoning gains. We ask: what aspects of a preference pair improve a reasoning model's performance on general reasoning tasks? We investigate two distinct notions of quality delta in preference data: generator-level delta, arising from the differences in capability between models that generate chosen and rejected reasoning traces, and sample-level delta, arising from differences in judged quality differences within an individual preference pair. To study generator-level delta, we vary the generator's scale and model family, and to study sample-level delta, we employ an LLM-as-a-judge to rate the quality of generated traces along multiple reasoning-quality dimensions. We find that increasing generator-level delta steadily improves performance on out-of-domain reasoning tasks and filtering data by sample-level delta can enable more data-efficient training. Our results suggest a twofold recipe for improving reasoning performance through preference optimization: maximize generator-level delta when constructing preference pairs and exploit sample-level delta to select the most informative training examples.
Authors: Paolo Gajo, Domenic Rosati, Hassan Sajjad, Alberto Barr\'on-Cede\~no
Abstract: Relation extraction represents a fundamental component in the process of creating knowledge graphs, among other applications. Large language models (LLMs) have been adopted as a promising tool for relation extraction, both in supervised and in-context learning settings. However, in this work we show that their performance still lags behind much smaller architectures when the linguistic graph underlying a text has great complexity. To demonstrate this, we evaluate four LLMs against a graph-based parser on six relation extraction datasets with sentence graphs of varying sizes and complexities. Our results show that the graph-based parser increasingly outperforms the LLMs, as the number of relations in the input documents increases. This makes the much lighter graph-based parser a superior choice in the presence of complex linguistic graphs.
Authors: Yousra Fettach, Guillaume Bied, Hannu Toivonen, Tijl De Bie
Abstract: Humor is one of the most culturally embedded and socially significant dimensions of human communication, yet it remains largely unexplored as a dimension of Large Language Model (LLM) alignment. In this study, five frontier language models play the same Cards Against Humanity games (CAH) as human players. The models select the funniest response from a slate of ten candidate cards across 9,894 rounds. While all models exceed the random baseline, alignment with human preference remains modest. More striking is that models agree with each other substantially more often than they agree with humans. We show that this preference is partly explained by systematic position biases and content preferences, raising the question whether LLM humor judgment reflects genuine preference or structural artifacts of inference and alignment.
Authors: Zhuoyi Yang, Jiapeng Yu, Reuben Tan, Boyang Li, Huijuan Xu
Abstract: Understanding instructional videos requires recognizing fine-grained actions and modeling their temporal relations, which remains challenging for current Video Foundation Models (VFMs). This difficulty stems from noisy web supervision and a pervasive "static bias", where models rely on objects rather than motion cues. To address this, we propose InstrAction, a pretraining framework for instructional videos' action-centric representations. We first introduce a data-driven strategy, which filters noisy captions and generates action-centric hard negatives to disentangle actions from objects during contrastive learning. At the visual feature level, an Action Perceiver extracts motion-relevant tokens from redundant video encodings. Beyond contrastive learning, we introduce two auxiliary objectives: Dynamic Time Warping alignment (DTW-Align) for modeling sequential temporal structure, and Masked Action Modeling (MAM) for strengthening cross-modal grounding. Finally, we introduce the InstrAct Bench to evaluate action-centric understanding, where our method consistently outperforms state-of-the-art VFMs on semantic reasoning, procedural logic, and fine-grained retrieval tasks.
Authors: Arda Atalik, Hui Xue, Rhodri H. Davies, Thomas A. Treibel, Daniel K. Sodickson, Michael S. Hansen, Peter Kellman
Abstract: Purpose: To develop and evaluate a deep learning (DL) method for free-breathing phase-sensitive inversion recovery (PSIR) late gadolinium enhancement (LGE) cardiac MRI that produces diagnostic-quality images from a single acquisition over two heartbeats, eliminating the need for 8 to 24 motion-corrected (MOCO) signal averages. Materials and Methods: Raw data comprising 800,653 slices from 55,917 patients, acquired on 1.5T and 3T scanners across multiple sites from 2016 to 2024, were used in this retrospective study. Data were split by patient: 640,000 slices (42,822 patients) for training and the remainder for validation and testing, without overlap. The training and testing data were from different institutions. PSIRNet, a physics-guided DL network with 845 million parameters, was trained end-to-end to reconstruct PSIR images with surface coil correction from a single interleaved IR/PD acquisition over two heartbeats. Reconstruction quality was evaluated using SSIM, PSNR, and NRMSE against MOCO PSIR references. Two expert cardiologists performed an independent qualitative assessment, scoring image quality on a 5-point Likert scale across bright blood, dark blood, and wideband LGE variants. Paired superiority and equivalence (margin = 0.25 Likert points) were tested using exact Wilcoxon signed-rank tests at a significance level of 0.05 using R version 4.5.2. Results: Both readers rated single-average PSIRNet reconstructions superior to MOCO PSIR for dark blood LGE (conservative P = .002); for bright blood and wideband, one reader rated it superior and the other confirmed equivalence (all P < .001). Inference required approximately 100 msec per slice versus more than 5 sec for MOCO PSIR. Conclusion: PSIRNet produces diagnostic-quality free-breathing PSIR LGE images from a single acquisition, enabling 8- to 24-fold reduction in acquisition time.
Authors: Mahdi Alizadeh
Abstract: User-space Adaptive Bitrate (ABR) algorithms cannot see the transport layer signals that matter most, such as minimum RTT and instantaneous delivery rate, and they respond to network changes only after damage has already propagated to the playout buffer. We present eBandit, a framework that relocates both network monitoring and ABR algorithm selection into the Linux kernel using eBPF. A lightweight epsilon-greedy Multi-Armed Bandit (MAB) runs inside a sockops program, evaluating three ABR heuristics against a reward derived from live TCP metrics. On an adversarial synthetic trace eBandit achieves $416.3 \pm 4.9$ cumulative QoE, outperforming the best static heuristic by $7.2\%$. On 42 real-world sessions eBandit achieves a mean QoE per chunk of $1.241$, the highest across all policies, demonstrating that kernel-resident bandit learning transfers to heterogeneous mobile conditions.
Authors: Sophie Wu, Andrew Piper
Abstract: Stories are key to transmitting values across cultures, but their interpretation varies across linguistic and cultural contexts. Thus, we introduce multilingual story moral generation as a novel culturally grounded evaluation task. Using a new dataset of human-written story morals collected across 14 language-culture pairs, we compare model outputs with human interpretations via semantic similarity, a human preference survey, and value categorization. We show that frontier models such as GPT-4o and Gemini generate story morals that are semantically similar to human responses and preferred by human evaluators. However, their outputs exhibit markedly less cross-linguistic variation and concentrate on a narrower set of widely shared values. These findings suggest that while contemporary models can approximate central tendencies of human moral interpretation, they struggle to reproduce the diversity that characterizes human narrative understanding. By framing narrative interpretation as an evaluative task, this work introduces a new approach to studying cultural alignment in language models beyond static benchmarks or knowledge-based tests.
Authors: Marc B\"ohlen, Sai Krishna
Abstract: This paper considers AI model churn as an opportunity for frugal investigation of large AI models. It describes how the incessant push for ever more powerful AI systems leaves in its wake a collection of obsolete yet powerful AI models, discarded in a veritable scrapyard of AI production. This scrapyard offers a potent opportunity for resource-constrained experimentation into AI systems. As in the physical scrapyard, nothing ever truly disappears in the AI scrapyard, it is just waiting to be reconfigured into something else. Project Nudge-x is an example of what can emerge from the AI scrapyard. Nudge-x seeks to manipulate legacy AI models to describe how mining sites across the planet are impacting landscapes and lives. By sharing this collection of brutal landscape interventions with people and AI systems alike, Nudge-x creates a venue for the appreciation of a history sadly shared between AI and people.
Authors: Chris Hicks, Elizabeth Bates, Shae McFadden, Isaac Symes Thompson, Myles Foley, Ed Chapman, Nickolas Espinosa Dice, Ankita Samaddar, Joshua Sylvester, Himanshu Neema, Nicholas Butts, Nate Foster, Ahmad Ridley, Zoe M, Paul Jones
Abstract: In November 2025, the authors ran a workshop on the topic of what makes a good reinforcement learning (RL) environment for autonomous cyber defence (ACD). This paper details the knowledge shared by participants both during the workshop and shortly afterwards by contributing herein. The workshop participants come from academia, industry, and government, and have extensive hands-on experience designing and working with RL and cyber environments. While there is now a sizeable body of literature describing work in RL for ACD, there is nevertheless a great deal of tradecraft, domain knowledge, and common hazards which are not detailed comprehensively in a single resource. With a specific focus on building better environments to train and evaluate autonomous RL agents in network defence scenarios, including government and critical infrastructure networks, the contributions of this work are twofold: (1) a framework for decomposing the interface between RL cyber environments and real systems, and (2) guidelines on current best practice for RL-based ACD environment development and agent evaluation, based on the key findings from our workshop.
Authors: Fatih Cagatay Akyon, Alptekin Temizel
Abstract: Content moderation systems classify images as safe or unsafe but lack spatial grounding and interpretability: they cannot explain what sensitive behavior was detected, who is involved, or where it occurs. We introduce the Sensitive Benchmark (SenBen), the first large-scale scene graph benchmark for sensitive content, comprising 13,999 frames from 157 movies annotated with Visual Genome-style scene graphs (25 object classes, 28 attributes including affective states such as pain, fear, aggression, and distress, 14 predicates) and 16 sensitivity tags across 5 categories. We distill a frontier VLM into a compact 241M student model using a multi-task recipe that addresses vocabulary imbalance in autoregressive scene graph generation through suffix-based object identity, Vocabulary-Aware Recall (VAR) Loss, and a decoupled Query2Label tag head with asymmetric loss, yielding a +6.4 percentage point improvement in SenBen Recall over standard cross-entropy training. On grounded scene graph metrics, our student model outperforms all evaluated VLMs except Gemini models and all commercial safety APIs, while achieving the highest object detection and captioning scores across all models, at $7.6\times$ faster inference and $16\times$ less GPU memory.
Authors: Mehran Taghian, Yunke Peng, Xing Huang, Yao Wang, Yaoyuan Wang, Wei Guo, Yuanyong Luo, Tianchi Hu, Junsong Wang, Xin Wang, Hu Liu, Yu Cheng, Ziwei Yu, Hongliang Li, Mehdi Rahimifar, Lei Yan, Xuefei Wang, Zhuang Ma, Lei Liu, Hui Yu, Anandharaju Durai Raju, Hoang Le, Hei Yi Mak, Tanzila Rahman, Shadan Golestan
Abstract: Large foundation models have become central to modern machine learning, with performance scaling predictably with model size and data. However, training and deploying such models incur substantial computational and memory costs, motivating the development of low-precision training techniques. Recent work has demonstrated that 4-bit floating-point (FP4) formats--such as MXFP4 and NVFP4--can be successfully applied to linear GEMM operations in large language models (LLMs), achieving up to 4x improvements in compute throughput and memory efficiency compared to higher-precision baselines. In this work, we investigate the recently proposed HiFloat4 FP4 format for Huawei Ascend NPUs and systematically compare it with MXFP4 in large-scale training settings. All experiments are conducted on Ascend NPU clusters, with linear and expert GEMM operations performed entirely in FP4 precision. We evaluate both dense architectures (e.g., Pangu and LLaMA-style models) and mixture-of-experts (MoE) models, where both standard linear layers and expert-specific GEMMs operate in FP4. Furthermore, we explore stabilization techniques tailored to FP4 training that significantly reduce numerical degradation, maintaining relative error within 1% of full-precision baselines while preserving the efficiency benefits of 4-bit computation. Our results provide a comprehensive empirical study of FP4 training on NPUs and highlight the practical trade-offs between FP4 formats in large-scale dense and MoE models.
Authors: Jinqi Luo, Jinyu Yang, Tal Neiman, Lei Fan, Bing Yin, Son Tran, Mubarak Shah, Ren\'e Vidal
Abstract: Multimodal Large Language Models (MLLMs) have been shown to be vulnerable to malicious queries that can elicit unsafe responses. Recent work uses prompt engineering, response classification, or finetuning to improve MLLM safety. Nevertheless, such approaches are often ineffective against evolving malicious patterns, may require rerunning the query, or demand heavy computational resources. Steering the activations of a frozen model at inference time has recently emerged as a flexible and effective solution. However, existing steering methods for MLLMs typically handle only a narrow set of safety-related concepts or struggle to adjust specific concepts without affecting others. To address these challenges, we introduce Dictionary-Aligned Concept Control (DACO), a framework that utilizes a curated concept dictionary and a Sparse Autoencoder (SAE) to provide granular control over MLLM activations. First, we curate a dictionary of 15,000 multimodal concepts by retrieving over 400,000 caption-image stimuli and summarizing their activations into concept directions. We name the dataset DACO-400K. Second, we show that the curated dictionary can be used to intervene activations via sparse coding. Third, we propose a new steering approach that uses our dictionary to initialize the training of an SAE and automatically annotate the semantics of the SAE atoms for safeguarding MLLMs. Experiments on multiple MLLMs (e.g., QwenVL, LLaVA, InternVL) across safety benchmarks (e.g., MM-SafetyBench, JailBreakV) show that DACO significantly improves MLLM safety while maintaining general-purpose capabilities.
Authors: Cyrus Zhou, Yufei Jin, Yilin Xu, Yu-Chiang Wang, Chieh-Ju Chao, Monica S. Lam
Abstract: Clinical trials are central to evidence-based medicine, yet many struggle to meet enrollment targets, despite the availability of over half a million trials listed on ClinicalTrials.gov, which attracts approximately two million users monthly. Existing retrieval techniques, largely based on keyword and embedding-similarity matching between patient profiles and eligibility criteria, often struggle with low recall, low precision, and limited interpretability due to complex constraints. We propose SatIR, a scalable clinical trial retrieval method based on constraint satisfaction, enabling high-precision and interpretable matching of patients to relevant trials. Our approach uses formal methods -- Satisfiability Modulo Theories (SMT) and relational algebra -- to efficiently represent and match key constraints from clinical trials and patient records. Beyond leveraging established medical ontologies and conceptual models, we use Large Language Models (LLMs) to convert informal reasoning regarding ambiguity, implicit clinical assumptions, and incomplete patient records into explicit, precise, controllable, and interpretable formal constraints. Evaluated on 59 patients and 3,621 trials, SatIR outperforms TrialGPT on all three evaluated retrieval objectives. It retrieves 32%-72% more relevant-and-eligible trials per patient, improves recall over the union of useful trials by 22-38 points, and serves more patients with at least one useful trial. Retrieval is fast, requiring 2.95 seconds per patient over 3,621 trials. These results show that SatIR is scalable, effective, and interpretable.
Authors: Greg Nyilasy, Brock Bastian, Jennifer Overbeck, Abraham Ryan Ade Putra Hito
Abstract: As organizations increasingly deploy AI as a teammate rather than a standalone tool, morally consequential mistakes often arise from joint human-AI workflows in which causality is ambiguous. We ask how people allocate responsibility in these hybrid-agent settings. Across four experiments (N = 1,801) in an AI-assisted lending context (e.g., discriminatory rejection, irresponsible lending, and low-harm filing errors), participants consistently attributed more responsibility to the human decision maker when the human was paired with AI than when paired with another human (by an average of 10 points on a 0-100 scale across studies). This AI-Induced Human Responsibility (AIHR) effect held across high and low harm scenarios and persisted even where self-serving blame-shifting (when the human in question was the self) would be expected. Process evidence indicates that AIHR is explained by inferences of agent autonomy: AI is seen as a constrained implementer, which makes the human the default locus of discretionary responsibility. Alternative mechanisms (mind perception; self-threat) did not account for the effect. These findings extend research on algorithm aversion, hybrid AI-human organizational behavior and responsibility gaps in technology by showing that AI-human teaming can increase (rather than dilute) human responsibility, with implications for accountability design in AI-enabled organizations.
Authors: Mintong Kang, Chen Fang, Bo Li
Abstract: Audio has rapidly become a primary interface for foundation models, powering real-time voice assistants. Ensuring safety in audio systems is inherently more complex than just "unsafe text spoken aloud": real-world risks can hinge on audio-native harmful sound events, speaker attributes (e.g., child voice), impersonation/voice-cloning misuse, and voice-content compositional harms, such as child voice plus sexual content. The nature of audio makes it challenging to develop comprehensive benchmarks or guardrails against this unique risk landscape. To close this gap, we conduct large-scale red teaming on audio systems, systematically uncover vulnerabilities in audio, and develop a comprehensive, policy-grounded audio risk taxonomy and AudioSafetyBench, the first policy-based audio safety benchmark across diverse threat models. AudioSafetyBench supports diverse languages, suspicious voices (e.g., celebrity/impersonation and child voice), risky voice-content combinations, and non-speech sound events. To defend against these threats, we propose AudioGuard, a unified guardrail consisting of 1) SoundGuard for waveform-level audio-native detection and 2) ContentGuard for policy-grounded semantic protection. Extensive experiments on AudioSafetyBench and four complementary benchmarks show that AudioGuard consistently improves guardrail accuracy over strong audio-LLM-based baselines with substantially lower latency.
Authors: Mohammed Maaz Sibhai, Abedalrhman Alkhateeb, Saad B. Ahmed
Abstract: To ensure safe clinical integration, deep learning models must provide more than just high accuracy; they require dependable uncertainty quantification. While current Medical Vision Transformers perform well, they frequently struggle with overconfident predictions and a lack of transparency, issues that are magnified by the noisy and imbalanced nature of clinical data. To address this, we enhanced the modified Medical Transformer (MedFormer) that incorporates prototype-based learning and uncertainty-guided routing, by utilizing a Dirichlet distribution for per-token evidential uncertainty, our framework can quantify and localize ambiguity in real-time. This uncertainty is not just an output but an active participant in the training process, filtering out unreliable feature updates. Furthermore, the use of class-specific prototypes ensures the embedding space remains structured, allowing for decisions based on visual similarity. Testing across four modalities (mammography, ultrasound, MRI, and histopathology) confirms that our approach significantly enhances model calibration, reducing expected calibration error (ECE) by up to 35%, and improves selective prediction, even when accuracy gains are modest.
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: Rafael da Silva, Jeff Eicher, Gregory Longo
Abstract: This study proposes a temporal modeling framework with a counterfactual policy-simulation layer for student dropout in higher education, using LMS engagement data and administrative withdrawal records. Dropout is operationalized as a time-to-event outcome at the enrollment level; weekly risk is modeled in discrete time via penalized, class-balanced logistic regression over person--period rows. Under a late-event temporal holdout, the model attains row-level AUCs of 0.8350 (train) and 0.8405 (test), with aggregate calibration acceptable but sparsely supported in the highest-risk bins. Ablation analyses indicate performance is sensitive to feature set composition, underscoring the role of temporal engagement signals. A scenario-indexed policy layer produces survival contrasts $\Delta S(T)$ under an explicit trigger/schedule contract: positive contrasts are confined to the shock branch ($T_{\rm policy}=18$: 0.0102, 0.0260, 0.0819), while the mechanism-aware branch is negative ($\Delta S_{\rm mech}(18)=-0.0078$, $\Delta S_{\rm mech}(38)=-0.0134$). A subgroup analysis by gender quantifies scenario-induced survival gaps via bootstrap; contrasts are directionally stable but small. Results are not causally identified; they demonstrate the framework's capacity for internal structural scenario comparison under observational data constraints.
Authors: Tokio Kajitsuka, Ukyo Honda, Sho Takase
Abstract: Chain-of-thought (CoT) distillation transfers reasoning behaviors from a strong teacher to a smaller student, but prior work reports a capacity gap: distillation may fail when the teacher-student capability mismatch is large. We revisit the capacity gap from a practical perspective by re-examining commonly used experimental settings. Notably, we find that CoT distillation often degrades performance compared to the student's pre-distillation baseline, an issue obscured when only post-distillation comparisons are reported. We therefore propose a more realistic evaluation protocol and find that the impact of capacity gap effects does not consistently dominate across tasks and settings, especially when candidate teachers differ substantially in performance. Our results offer practical guidance for selecting teacher-student pairs in CoT distillation.
Authors: Chengjie Fan, Cong Pan, Zijian Liu, Ningzhong Liu, Jie Qin
Abstract: Inspired by the general Vision-and-Language Navigation (VLN) task, aerial VLN has attracted widespread attention, owing to its significant practical value in applications such as logistics delivery and urban inspection. However, existing methods face several challenges in complex urban environments, including insufficient generalization to unseen scenes, suboptimal performance in long-range path planning, and inadequate understanding of spatial continuity. To address these challenges, we propose HTNav, a new collaborative navigation framework that integrates Imitation Learning (IL) and Reinforcement Learning (RL) within a hybrid IL-RL framework. This framework adopts a staged training mechanism to ensure the stability of the basic navigation strategy while enhancing its environmental exploration capability. By integrating a tiered decision-making mechanism, it achieves collaborative interaction between macro-level path planning and fine-grained action control. Furthermore, a map representation learning module is introduced to deepen its understanding of spatial continuity in open domains. On the CityNav benchmark, our method achieves state-of-the-art performance across all scene levels and task difficulties. Experimental results demonstrate that this framework significantly improves navigation precision and robustness in complex urban environments.
Authors: Xinyu Zhang, Zurong Mai, Qingmei Li, Zjin Liao, Yibin Wen, Yuhang Chen, Xiaoya Fan, Chan Tsz Ho, Bi Tianyuan, Haoyuan Liang, Ruifeng Su, Zihao Qian, Juepeng Zheng, Jianxi Huang, Yutong Lu, Haohuan Fu
Abstract: While multimodal large language models (MLLMs) have made significant strides in natural image understanding, their ability to perceive and reason over hyperspectral image (HSI) remains underexplored, which is a vital modality in remote sensing. The high dimensionality and intricate spectral-spatial properties of HSI pose unique challenges for models primarily trained on RGB data.To address this gap, we introduce Hyperspectral Multimodal Benchmark (HM-Bench), the first benchmark designed specifically to evaluate MLLMs in HSI understanding. We curate a large-scale dataset of 19,337 question-answer pairs across 13 task categories, ranging from basic perception to spectral reasoning. Given that existing MLLMs are not equipped to process raw hyperspectral cubes natively, we propose a dual-modality evaluation framework that transforms HSI data into two complementary representations: PCA-based composite images and structured textual reports. This approach facilitates a systematic comparison of different representation for model performance. Extensive evaluations on 18 representative MLLMs reveal significant difficulties in handling complex spatial-spectral reasoning tasks. Furthermore, our results demonstrate that visual inputs generally outperform textual inputs, highlighting the importance of grounding in spectral-spatial evidence for effective HSI understanding. Dataset and appendix can be accessed at https://github.com/HuoRiLi-Yu/HM-Bench.
Authors: Hang Gao, Kunyu Li, Huang Hong, Baoquan Cui, Fengge Wu
Abstract: Modeling causal relationships in graph representation learning remains a fundamental challenge. Existing approaches often draw on theories and methods from causal inference to identify causal subgraphs or mitigate confounders. However, due to the inherent complexity of graph-structured data, these approaches frequently aggregate diverse graph elements into single causal variables, an operation that risks violating the core assumptions of causal inference. In this work, we prove that such aggregation compromises causal validity. Building on this conclusion, we propose a theoretical model grounded in the smallest indivisible units of graph data to ensure that the causal validity is guaranteed. With this model, we further analyze the costs of achieving precise causal modeling in graph representation learning and identify the conditions under which the problem can be simplified. To empirically support our theory, we construct a controllable synthetic dataset that reflects realworld causal structures and conduct extensive experiments for validation. Finally, we develop a causal modeling enhancement module that can be seamlessly integrated into existing graph learning pipelines, and we demonstrate its effectiveness through comprehensive comparative experiments.
Authors: Mohsen Yaghoubi Suraki
Abstract: Glioma is a harmful brain tumor that requires early detection to ensure better health results. Early detection of this tumor is key for effective treatment and requires an automated segmentation process. However, it is a challenging task to find tumors due to tumor characteristics like location and size. A reliable method to accurately separate tumor zones from healthy tissues is deep learning models, which have shown promising results over the last few years. In this research, an Adaptive Dual Residual U-Net with Attention Gate and Multiscale Spatial Attention Mechanisms (ADRUwAMS) is introduced. This model is an innovative combination of adaptive dual residual networks, attention mechanisms, and multiscale spatial attention. The dual adaptive residual network architecture captures high-level semantic and intricate low-level details from brain images, ensuring precise segmentation of different tumor parts, types, and hard regions. The attention gates use gating and input signals to compute attention coefficients for the input features, and multiscale spatial attention generates scaled attention maps and combines these features to hold the most significant information about the brain tumor. We trained the model for 200 epochs using the ReLU activation function on BraTS 2020 and BraTS 2019 datasets. These improvements resulted in high accuracy for tumor detection and segmentation on BraTS 2020, achieving dice scores of 0.9229 for the whole tumor, 0.8432 for the tumor core, and 0.8004 for the enhancing tumor.
Authors: Zecheng Hao, Shenghao Xie, Kang Chen, Wenxuan Liu, Zhaofei Yu, Tiejun Huang
Abstract: Spiking Neural Networks (SNNs) offer superior energy efficiency over Artificial Neural Networks (ANNs). However, they encounter significant deficiencies in training and inference metrics when applied to Spiking Vision Transformers (S-ViTs). Existing paradigms including ANN-SNN Conversion and Spatial-Temporal Backpropagation (STBP) suffer from inherent limitations, precluding concurrent optimization of memory, accuracy and energy consumption. To address these issues, we propose Ge$^\text{2}$mS-T, a novel architecture implementing grouped computation across temporal, spatial and network structure dimensions. Specifically, we introduce the Grouped-Exponential-Coding-based IF (ExpG-IF) model, enabling lossless conversion with constant training overhead and precise regulation for spike patterns. Additionally, we develop Group-wise Spiking Self-Attention (GW-SSA) to reduce computational complexity via multi-scale token grouping and multiplication-free operations within a hybrid attention-convolution framework. Experiments confirm that our method can achieve superior performance with ultra-high energy efficiency on challenging benchmarks. To our best knowledge, this is the first work to systematically establish multi-dimensional grouped computation for resolving the triad of memory overhead, learning capability and energy budget in S-ViTs.
Authors: Yuanting Fan, Jun Liu, Bin-Bin Gao, Xiaochen Chen, Yuhuan Lin, Zhewei Dai, Jiawei Zhan, Chengjie Wang
Abstract: Existing defect/anomaly generation methods often rely on few-shot learning, which overfits to specific defect categories due to the lack of large-scale paired defect editing data. This issue is aggravated by substantial variations in defect scale and morphology, resulting in limited generalization, degraded realism, and category consistency. We address these challenges by introducing UDG, a large-scale dataset of 300K normal-abnormal-mask-caption quadruplets spanning diverse domains, and by presenting UniDG, a universal defect generation foundation model that supports both reference-based defect generation and text instruction-based defect editing without per-category fine-tuning. UniDG performs Defect-Context Editing via adaptive defect cropping and structured diptych input format, and fuses reference and target conditions through MM-DiT multimodal attention. A two-stage training strategy, Diversity-SFT followed by Consistency-RFT, further improves diversity while enhancing realism and reference consistency. Extensive experiments on MVTec-AD and VisA show that UniDG outperforms prior few-shot anomaly generation and image insertion/editing baselines in synthesis quality and downstream single- and multi-class anomaly detection/localization. Code will be available at https://github.com/RetoFan233/UniDG.
Authors: Hengran Zhang, Minghao Tang, Keping Bi, Jiafeng Guo
Abstract: Information retrieval systems have traditionally optimized for topical relevance-the degree to which retrieved documents match a query. However, relevance only approximates a deeper goal: utility, namely, whether retrieved information helps accomplish a user's underlying task. The emergence of retrieval-augmented generation (RAG) fundamentally changes this paradigm. Retrieved documents are no longer consumed directly by users but instead serve as evidence for large language models (LLMs) that produce answers. As a result, retrieval effectiveness must be evaluated by its contribution to generation quality rather than by relevance-based ranking metrics alone. This tutorial argues that retrieval objectives are evolving from relevance-centric optimization toward LLM-centric utility. We present a unified framework covering LLM-agnostic versus LLM-specific utility, context-independent versus context-dependent utility, and the connection with LLM information needs and agentic RAG. By synthesizing recent advances, the tutorial provides conceptual foundations and practical guidance for designing retrieval systems aligned with the requirements of LLM-based information access.
Authors: Rares-Alexandru Roscan, Gabriel Petre1, Adrian-Marius Dumitran, Angela-Liliana Dumitran
Abstract: As Large Language Models (LLMs) become increasingly prevalent in text simplification, systematically evaluating their outputs across diverse prompting strategies and architectures remains a critical methodological challenge in both NLP research and Intelligent Tutoring Systems (ITS). Developing robust prompts is often hindered by the absence of structured, visual frameworks for comparative text analysis. While researchers typically rely on static computational scripts, educators are constrained to standard conversational interfaces -- neither paradigm supports systematic multi-dimensional evaluation of prompt-model permutations. To address these limitations, we introduce \textbf{MuTSE}\footnote{The project code and the demo have been made available for peer review at the following anonymized URL. https://osf.io/njs43/overview?view_only=4b4655789f484110a942ebb7788cdf2a, an interactive human-in-the-loop web application designed to streamline the evaluation of LLM-generated text simplifications across arbitrary CEFR proficiency targets. The system supports concurrent execution of $P \times M$ prompt-model permutations, generating a comprehensive comparison matrix in real-time. By integrating a novel tiered semantic alignment engine augmented with a linearity bias heuristic ($\lambda$), MuTSE visually maps source sentences to their simplified counterparts, reducing the cognitive load associated with qualitative analysis and enabling reproducible, structured annotation for downstream NLP dataset construction.
URLs: https://osf.io/njs43/overview?view_only=4b4655789f484110a942ebb7788cdf2a,
Authors: Mintae Kim, Koushil Sreenath
Abstract: Reinforcement learning (RL) in robotics is often limited by the cost and risk of data collection, motivating experience transfer from a source task to a target task. Offline-to-online RL leverages prior data but typically assumes a given fixed dataset and does not address how to generate reliable data for transfer. We propose \textit{World Model-based Experience Transfer} (WOMBET), a framework that jointly generates and utilizes prior data. WOMBET learns a world model in the source task and generates offline data via uncertainty-penalized planning, followed by filtering trajectories with high return and low epistemic uncertainty. It then performs online fine-tuning in the target task using adaptive sampling between offline and online data, enabling a stable transition from prior-driven initialization to task-specific adaptation. We show that the uncertainty-penalized objective provides a lower bound on the true return and derive a finite-sample error decomposition capturing distribution mismatch and approximation error. Empirically, WOMBET improves sample efficiency and final performance over strong baselines on continuous control benchmarks, demonstrating the benefit of jointly optimizing data generation and transfer.
Authors: Keyu Li, Jin Gao, Dequan Wang
Abstract: While Multi-Agent Systems (MAS) are increasingly deployed for complex workflows, their emergent properties-particularly the accumulation of bias-remain poorly understood. Because real-world MAS are too complex to analyze entirely, evaluating their ethical robustness requires first isolating their foundational mechanics. In this work, we conduct a baseline empirical study investigating how basic MAS topologies and feedback loops influence prejudice. Contrary to the assumption that multi-agent collaboration naturally dilutes bias, we hypothesize that structured workflows act as echo chambers, amplifying minor stochastic biases into systemic polarization. To evaluate this, we introduce Discrim-Eval-Open, an open-ended benchmark that bypasses individual model neutrality through forced comparative judgments across demographic groups. Analyzing bias cascades across various structures reveals that architectural sophistication frequently exacerbates bias rather than mitigating it. We observe systemic amplification even when isolated agents operate neutrally, and identify a 'Trigger Vulnerability' where injecting purely objective context drastically accelerates polarization. By stripping away advanced swarm complexity to study foundational dynamics, we establish a crucial baseline: structural complexity does not guarantee ethical robustness. Our code is available at https://github.com/weizhihao1/MAS-Bias.
Authors: Avni Mittal, Shanu Kumar, Sandipan Dandapat, Monojit Choudhury
Abstract: We study predictive multilingual evaluation: estimating how well a model will perform on a task in a target language when direct benchmark results are missing. This problem is common in multilingual deployment, where evaluation coverage is sparse and published evidence is uneven across languages, tasks, and model families. We introduce a controlled benchmark of 1,500 questions spanning six tasks and five evidence scenarios. The benchmark separates accessible evidence from ground truth, enabling evaluation of systems that must infer missing results from incomplete literature evidence. We also present Litmus (Re)Agent, a DAG-orchestrated agentic system that decomposes queries into hypotheses, retrieves evidence, and synthesises predictions through feature-aware aggregation. Across six systems, Litmus (Re)Agent achieves the best overall performance, with the largest gains in transfer-heavy scenarios where direct evidence is weak or absent. These results show that structured agentic reasoning is a promising approach to multilingual performance estimation under incomplete evidence.
Authors: Yi Luo, Xu Sun, Guangchun Luo, Aiguo Chen
Abstract: Graph neural networks (GNNs) have been widely adopted in engineering applications such as social network analysis, chemical research and computer vision. However, their efficacy is severely compromised by the inherent homophily assumption, which fails to hold for heterophilic graphs where dissimilar nodes are frequently connected. To address this fundamental limitation in graph learning, we first draw inspiration from the recently discovered monophily property of real-world graphs, and propose Neighbourhood Transformers (NT), a novel paradigm that applies self-attention within every local neighbourhood instead of aggregating messages to the central node as in conventional message-passing GNNs. This design makes NT inherently monophily-aware and theoretically guarantees its expressiveness is no weaker than traditional message-passing frameworks. For practical engineering deployment, we further develop a neighbourhood partitioning strategy equipped with switchable attentions, which reduces the space consumption of NT by over 95% and time consumption by up to 92.67%, significantly expanding its applicability to larger graphs. Extensive experiments on 10 real-world datasets (5 heterophilic and 5 homophilic graphs) show that NT outperforms all current state-of-the-art methods on node classification tasks, demonstrating its superior performance and cross-domain adaptability. The full implementation code of this work is publicly available at https://github.com/cf020031308/MoNT to facilitate reproducibility and industrial adoption.
Authors: Jihwan Oh, Soowon Oh, Murad Aghazada, Minchan Jeong, Sungnyun Kim, Se-Young Yun
Abstract: Persona prompting has been widely adopted to steer large language models (LLMs) behavior and improve their instruction performance by assigning specific characters. However, identifying an optimal persona is time-consuming, and its impact on output quality remains poorly understood. Prior work has mainly addressed this issue at the prompt level via inference-time strategies, incurring additional computation. In this work, we avoid inference-time prompt search by tackling persona sensitivity during training, aiming to train models that adapt their behavior to diverse personas while preserving task performance. In particular, we find that reinforcement learning with verifiable rewards (RLVR) systematically reduces sensitivity to persona prompts, but also reveals an inherent trade-off of outcome-based optimization: while RLVR improves robustness on tasks with verifiable goals, it can also degrade persona expressivity when needed, e.g., in-character role-playing. To address this limitation, we propose PerMix-RLVR, a persona-mixed RLVR strategy that mitigates the persona robustness-fidelity trade-off, preserving strong robustness to harmful persona variation while enabling faithful persona adoption when required. Concretely, PerMix-RLVR improves persona stability score (PSS) over RLVR by +21.2% on MATH500, while also enhancing persona fidelity by +11.4% on PersonaGym.
Authors: Zhiyu Zhou, Peilin Liu, Ruoxuan Zhang, Luyang Zhang, Cheng Zhang, Hongxia Xie, Wen-Huang Cheng
Abstract: Small object-centric spatial understanding in indoor videos remains a significant challenge for multimodal large language models (MLLMs), despite its practical value for object search and assistive applications. Although existing benchmarks have advanced video spatial intelligence, embodied reasoning, and diagnostic perception, no existing benchmark directly evaluates whether a model can localize a target object in video and express its position with sufficient precision for downstream use. In this work, we introduce PinpointQA, the first dataset and benchmark for small object-centric spatial understanding in indoor videos. Built from ScanNet++ and ScanNet200, PinpointQA comprises 1,024 scenes and 10,094 QA pairs organized into four progressively challenging tasks: Target Presence Verification (TPV), Nearest Reference Identification (NRI), Fine-Grained Spatial Description (FSD), and Structured Spatial Prediction (SSP). The dataset is built from intermediate spatial representations, with QA pairs generated automatically and further refined through quality control. Experiments on representative MLLMs reveal a consistent capability gap along the progressive chain, with SSP remaining particularly difficult. Supervised fine-tuning on PinpointQA yields substantial gains, especially on the harder tasks, demonstrating that PinpointQA serves as both a diagnostic benchmark and an effective training dataset. The dataset and project page are available at https://rainchowz.github.io/PinpointQA.
Authors: Xiaoke Guo, Songze Li, Zhiqiang Liu, Zhaoyan Gong, Yuanxiang Liu, Huajun Chen, Wen Zhang
Abstract: Table serialization remains a critical bottleneck for Large Language Models (LLMs) in complex table question answering, hindered by challenges such as structural neglect, representation gaps, and reasoning opacity. Existing serialization methods fail to capture explicit hierarchies and lack schema flexibility, while current tree-based approaches suffer from limited semantic adaptability. To address these limitations, we propose ASTRA (Adaptive Semantic Tree Reasoning Architecture) including two main modules, AdaSTR and DuTR. First, we introduce AdaSTR, which leverages the global semantic awareness of LLMs to reconstruct tables into Logical Semantic Trees. This serialization explicitly models hierarchical dependencies and employs an adaptive mechanism to optimize construction strategies based on table scale. Second, building on this structure, we present DuTR, a dual-mode reasoning framework that integrates tree-search-based textual navigation for linguistic alignment and symbolic code execution for precise verification. Experiments on complex table benchmarks demonstrate that our method achieves state-of-the-art (SOTA) performance.
Authors: Wenxi Li, Xihao Wang, Weiwei Sun
Abstract: The widespread use of English as a Second or Foreign Language (ESFL) has sparked a paradigm shift: ESFL is not seen merely as a deviation from standard English but as a distinct linguistic system in its own right. This shift highlights the need for dedicated, knowledge-intensive representations of ESFL. In response, this paper surveys existing ESFL resources, identifies their limitations, and proposes a novel solution. Grounded in constructivist theories, the paper treats constructions as the fundamental units of analysis, allowing it to model the syntax--semantics interface of both ESFL and standard English. This design captures a wide range of ESFL phenomena by referring to syntactico-semantic mappings of English while preserving ESFL's unique characteristics, resulting a gold-standard syntactico-semantic resource comprising 1643 annotated ESFL sentences. To demonstrate the sembank's practical utility, we conduct a pilot study testing the Linguistic Niche Hypothesis, highlighting its potential as a valuable tool in Second Language Acquisition research.
Authors: Carlos Jimeno Miguel, Raul Orduna, Francesco Zola
Abstract: This study addresses the challenge of creating datasets for cybercrime analysis while complying with the requirements of regulations such as the General Data Protection Regulation (GDPR) and Organic Law 10/1995 of the Penal Code. To this end, a system is proposed for collecting information from the Telegram platform, including text, audio, and images; the implementation of speech-to-text transcription models incorporating signal enhancement techniques; and the evaluation of different Named Entity Recognition (NER) solutions, including Microsoft Presidio and AI models designed using a transformer-based architecture. Experimental results indicate that Parakeet achieves the best performance in audio transcription, while the proposed NER solutions achieve the highest f1-score values in detecting sensitive information. In addition, anonymization metrics are presented that allow evaluation of the preservation of structural coherence in the data, while simultaneously guaranteeing the protection of personal information and supporting cybersecurity research within the current legal framework.
Authors: Andre Bacellar
Abstract: Two-hop QA retrieval splits queries into two regimes determined by whether the hop-2 entity is explicitly named in the question (Q-dominant) or only in the bridge passage (B-dominant). We formalize this split with three theorems: (T1) per-query AUC is a monotone function of the cosine separation margin, with R^2 >= 0.90 for six of eight type-encoder pairs; (T2) regime is characterized by two surface-text predicates, with P1 decisive for routing and P2 qualifying the B-dominant case, holding across three encoders and three datasets; and (T3) bridge advantage requires the relation-bearing sentence, not entity name alone, with removal causing an 8.6-14.1 pp performance drop (p < 0.001). Building on this theory, we propose RegimeRouter, a lightweight binary router that selects between question-only and question-plus-relation-sentence retrieval using five text features derived directly from the predicate definitions. Trained on 2WikiMultiHopQA (n = 881, 5-fold cross-fitted) and applied zero-shot to MuSiQue and HotpotQA, RegimeRouter achieves +5.6 pp (p < 0.001), +5.3 pp (p = 0.002), and +1.1 pp (non-significant, no-regret) R@5 improvement, respectively, with artifact-driven.
Authors: Qixuan Huang, Khalid Zaman, Masashi Unoki
Abstract: Auditory large language models (ALLMs) have demonstrated strong general capabilities in audio understanding and reasoning tasks. However, their reliability is still undermined by hallucination issues. Existing hallucination evaluation methods are formulated as binary classification tasks, which are insufficient to characterize the more complex hallucination patterns that arise in generative tasks. Moreover, current hallucination mitigation strategies rely on fine-tuning, resulting in high computational costs. To address the above limitations, we propose a plug-and-play Noise-Aware In-Context Learning (NAICL) method. Specifically, we construct a noise prior library, retrieve noise examples relevant to the input audio, and incorporate them as contextual priors, thereby guiding the model to reduce speculative associations when acoustic evidence is insufficient and to adopt a more conservative generation strategy. In addition, we establish a hallucination benchmark for audio caption tasks including the construction of the Clotho-1K multi-event benchmark dataset, the definition of four types of auditory hallucinations, and the introduction of metrics such as hallucination type distribution to support fine-grained analysis. Experimental results show that all evaluated ALLMs exhibit same hallucination behaviors. Moreover, the proposed NAICL method reduces the overall hallucination rate from 26.53% to 16.98%.
Authors: Zedian Shao, Hongbin Liu, Yuepeng Hu, Neil Zhenqiang Gong
Abstract: Multi-modal large language models (MLLMs) have emerged as powerful tools for analyzing Internet-scale image data, offering significant benefits but also raising critical safety and societal concerns. In particular, open-weight MLLMs may be misused to extract sensitive information from personal images at scale, such as identities, locations, or other private details. In this work, we propose ImageProtector, a user-side method that proactively protects images before sharing by embedding a carefully crafted, nearly imperceptible perturbation that acts as a visual prompt injection attack on MLLMs. As a result, when an adversary analyzes a protected image with an MLLM, the MLLM is consistently induced to generate a refusal response such as "I'm sorry, I can't help with that request." We empirically demonstrate the effectiveness of ImageProtector across six MLLMs and four datasets. Additionally, we evaluate three potential countermeasures, Gaussian noise, DiffPure, and adversarial training, and show that while they partially mitigate the impact of ImageProtector, they simultaneously degrade model accuracy and/or efficiency. Our study focuses on the practically important setting of open-weight MLLMs and large-scale automated image analysis, and highlights both the promise and the limitations of perturbation-based privacy protection.
Authors: Chenjie Yang, Yutian Jiang, Chenyu Wu
Abstract: Vision-language models (VLMs) have shown a promising ability in image geolocation, but they still lack structured geographic reasoning and the capacity for autonomous self-evolution. Existing methods predominantly rely on implicit parametric memory, which often exploits outdated knowledge and generates hallucinated reasoning. Furthermore, current inference is a "one-off" process, lacking the feedback loops necessary for self-evolution based on reasoning outcomes. To address these issues, we propose GeoSkill, a training-free framework based on an evolving Skill-Graph. We first initialize the graph by refining human expert trajectories into atomic, natural-language skills. For execution, GeoSkill employs an inference model to perform direct reasoning guided by the current Skill-Graph. For continuous growth, an Autonomous Evolution mechanism leverages a larger model to conduct multiple reasoning rollouts on image-coordinate pairs sourced from web-scale data and verified real-world reasoning. By analyzing both successful and failed trajectories from these rollouts, the mechanism iteratively synthesizes and prunes skills, effectively expanding the Skill-Graph and correcting geographic biases without any parameter updates. Experiments demonstrate that GeoSkill achieves promising performance in both geolocation accuracy and reasoning faithfulness on GeoRC, while maintaining superior generalization across diverse external datasets. Furthermore, our autonomous evolution fosters the emergence of novel, verifiable skills, significantly enhancing the system's cognition of real-world geographic knowledge beyond isolated case studies.
Authors: Yeonjun Hwang, Sungyong Park, Minju Kim, Dongha Lee, Jinyoung Yeo
Abstract: Large language models have been widely explored as decision-support tools in high-stakes domains due to their contextual understanding and reasoning capabilities. However, existing decision-making benchmarks rely on two simplifying assumptions: actions are selected from a finite set of pre-defined candidates, and explicit conditions restricting action feasibility are not incorporated into the decision-making process. These assumptions fail to capture the compositional structure of real-world actions and the explicit conditions that constrain their validity. To address these limitations, we introduce CONDESION-BENCH, a benchmark designed to evaluate conditional decision-making in compositional action space. In CONDESION-BENCH, actions are defined as allocations to decision variables and are restricted by explicit conditions at the variable, contextual, and allocation levels. By employing oracle-based evaluation of both decision quality and condition adherence, we provide a more rigorous assessment of LLMs as decision-support tools.
Authors: Salva R\"uhling Cachay, Duncan Watson-Parris, Rose Yu
Abstract: AI-based weather forecasting now rivals traditional physics-based ensembles, but state-of-the-art (SOTA) models rely on specialized architectures and massive computational budgets, creating a high barrier to entry. We demonstrate that such complexity is unnecessary for frontier performance. We introduce U-Cast, a probabilistic forecaster built on a standard U-Net backbone trained with a simple recipe: deterministic pre-training on Mean Absolute Error followed by short probabilistic fine-tuning on the Continuous Ranked Probability Score (CRPS) using Monte Carlo Dropout for stochasticity. As a result, our model matches or exceeds the probabilistic skill of GenCast and IFS ENS at 1.5$^\circ\$ resolution while reducing training compute by over 10$\times$ compared to leading CRPS-based models and inference latency by over 10$\times$ compared to diffusion-based models. U-Cast trains in under 12 H200 GPU-days and generates a 60-step ensemble forecast in 11 seconds. These results suggest that scalable, general-purpose architectures paired with efficient training curricula can match complex domain-specific designs at a fraction of the cost, opening the training of frontier probabilistic weather models to the broader community. Our code is available at: https://github.com/Rose-STL-Lab/u-cast.
Authors: Mauricio Fadel Argerich, Jonathan F\"urst, Marta Pati\~no-Mart\'inez
Abstract: While the large energy consumption of Large Language Models (LLMs) is recognized by the community, system operators lack guidance for energy-efficient LLM inference deployments that leverage energy trade-offs of heterogeneous hardware due to a lack of energy-aware benchmarks and data. In this work we address this gap with Watt Counts: the largest open-access dataset of energy consumption of LLMs, with over 5,000 experiments for 50 LLMs across 10 NVIDIA Graphics Processing Units (GPUs) in batch and server scenarios along with a reproducible, open-source benchmark that enables community submissions to expand this dataset. Leveraging this dataset, we conduct a system-level study of LLM inference across heterogeneous GPU architectures and show that GPU selection is crucial for energy efficiency outcomes and that optimal hardware choices vary significantly across models and deployment scenarios, demonstrating the critical importance of hardware-aware deployment in heterogeneous LLM systems. Guided by our data and insights, we show that practitioners can reduce energy consumption by up to 70% in server scenarios with negligible impact on user experience, and by up to 20% in batch scenarios.
Authors: Min Young Baeg, Yoon-Yeong Kim
Abstract: Long-horizon spatiotemporal prediction remains a challenging problem due to cumulative errors, noise amplification, and the lack of physical consistency in existing models. While diffusion models provide a probabilistic framework for modeling uncertainty, conventional approaches often rely on mean squared error objectives and fail to capture the underlying dynamics governed by physical laws. In this work, we propose PDYffusion, a dynamics-informed diffusion framework that integrates PDE-based regularization and uncertainty-aware forecasting for stable long-term prediction. The proposed method consists of two key components: a PDE-regularized interpolator and a UKF-based forecaster. The interpolator incorporates a differential operator to enforce physically consistent intermediate states, while the forecaster leverages the Unscented Kalman Filter to explicitly model uncertainty and mitigate error accumulation during iterative prediction. We provide theoretical analyses showing that the proposed interpolator satisfies PDE-constrained smoothness properties, and that the forecaster converges under the proposed loss formulation. Extensive experiments on multiple dynamical datasets demonstrate that PDYffusion achieves superior performance in terms of CRPS and MSE, while maintaining stable uncertainty behavior measured by SSR. We further analyze the inherent trade-off between prediction accuracy and uncertainty, showing that our method provides a balanced and robust solution for long-horizon forecasting.
Authors: Guoqing Wang, Pin Tang, Xiangxuan Ren, Guodongfang Zhao, Bailan Feng, Chao Ma
Abstract: Vision-Language-Action (VLA) models have recently achieved notable progress in end-to-end autonomous driving by integrating perception, reasoning, and control within a unified multimodal framework. However, they often lack explicit modeling of temporal dynamics and global world consistency, which limits their foresight and safety. In contrast, world models can simulate plausible future scenes but generally struggle to reason about or evaluate the imagined future they generate. In this work, we present VLA-World, a simple yet effective VLA world model that unifies predictive imagination with reflective reasoning to improve driving foresight. VLA-World first uses an action-derived feasible trajectory to guide the generation of the next-frame image, capturing rich spatial and temporal cues that describe how the surrounding environment evolves. The model then reasons over this self-generated future imagined frame to refine the predicted trajectory, achieving higher performance and better interpretability. To support this pipeline, we curate nuScenes-GR-20K, a generative reasoning dataset derived from nuScenes, and employ a three-stage training strategy that includes pretraining, supervised fine-tuning, and reinforcement learning. Extensive experiments demonstrate that VLA-World consistently surpasses state-of-the-art VLA and world-model baselines on both planning and future-generation benchmarks. Project page: https://vlaworld.github.io
Authors: Yuxi Zhou, Zhengbo Zhang, Jingyu Pan, Zhiyu Lin, Zhigang Tu
Abstract: Human action recognition is pivotal in computer vision, with applications ranging from surveillance to human-robot interaction. Despite the effectiveness of supervised skeleton-based methods, their reliance on exhaustive annotation limits generalization to novel actions. Zero-Shot Skeleton Action Recognition (ZSAR) emerges as a promising paradigm, yet it faces challenges due to the spectral bias of diffusion models, which oversmooth high-frequency dynamics. Here, we propose Frequency-Aware Diffusion for Skeleton-Text Matching (FDSM), integrating a Semantic-Guided Spectral Residual Module, a Timestep-Adaptive Spectral Loss, and Curriculum-based Semantic Abstraction to address these challenges. Our approach effectively recovers fine-grained motion details, achieving state-of-the-art performance on NTU RGB+D, PKU-MMD, and Kinetics-skeleton datasets. Code has been made available at https://github.com/yuzhi535/FDSM. Project homepage: https://yuzhi535.github.io/FDSM.github.io/
URLs: https://github.com/yuzhi535/FDSM., https://yuzhi535.github.io/FDSM.github.io/
Authors: Parjanya Aditya Shukla, Shubham Kumar Nigam, Debtanu Datta, Balaramamahanthi Deepak Patnaik, Noel Shallum, Pradeep Reddy Vanga, Saptarshi Ghosh, Arnab Bhattacharya
Abstract: Court Judgment Prediction and Explanation (CJPE) aims to predict a judicial decision and provide a legally grounded explanation for a given case based on the facts, legal issues, arguments, cited statutes, and relevant precedents. For such systems to be practically useful in judicial or legal research settings, they must not only achieve high predictive performance but also generate transparent and structured legal reasoning that aligns with established judicial practices. In this work, we present NyayaMind, an open-source framework designed to enable transparent and scalable legal reasoning for the Indian judiciary. The proposed framework integrates retrieval, reasoning, and verification mechanisms to emulate the structured decision-making process typically followed in courts. Specifically, NyayaMind consists of two main components: a Retrieval Module and a Prediction Module. The Retrieval Module employs a RAG pipeline to identify legally relevant statutes and precedent cases from large-scale legal corpora, while the Prediction Module utilizes reasoning-oriented LLMs fine-tuned for the Indian legal domain to generate structured outputs including issues, arguments, rationale, and the final decision. Our extensive results and expert evaluation demonstrate that NyayaMind significantly improves the quality of explanation and evidence alignment compared to existing CJPE approaches, providing a promising step toward trustworthy AI-assisted legal decision support systems.
Authors: Harry Proshian, Nikita Severin, Sergey Nikolenko, Kireev Ivan, Andrey Savchenko, Ivan Sergeev, Maria Postnova, Ilya Makarov
Abstract: Large-scale digital platforms generate billions of timestamped user-item interactions (events) that are crucial for predicting user attributes in, e.g., fraud prevention and recommendations. While self-supervised learning (SSL) effectively models the temporal order of events, it typically overlooks the global structure of the user-item interaction graph. To bridge this gap, we propose three model-agnostic strategies for integrating this structural information into contrastive SSL: enriching event embeddings, aligning client representations with graph embeddings, and adding a structural pretext task. Experiments on four financial and e-commerce datasets demonstrate that our approach consistently improves the accuracy (up to a 2.3% AUC) and reveals that graph density is a key factor in selecting the optimal integration strategy.
Authors: Li Huang, Zhongxin Liu, Yifan Wu, Tao Yin, Dong Li, Jichao Bi, Nankun Mu, Hongyu Zhang, Meng Yan
Abstract: Large Language Models (LLMs) for code generation can replicate insecure patterns from their training data. To mitigate this, a common strategy for security hardening is to fine-tune models using supervision derived from the final transformer layer. However, this design may suffer from a final-layer bottleneck: vulnerability-discriminative cues can be distributed across layers and become less detectable near the output representations optimized for next-token prediction. To diagnose this issue, we perform layer-wise linear probing. We observe that vulnerability-related signals are most detectable in a band of intermediate-to-upper layers yet attenuate toward the final layers. Motivated by this observation, we introduce DeepGuard, a framework that leverages distributed security-relevant cues by aggregating representations from multiple upper layers via an attention-based module. The aggregated signal powers a dedicated security analyzer within a multi-objective training objective that balances security enhancement and functional correctness, and further supports a lightweight inference-time steering strategy. Extensive experiments across five code LLMs demonstrate that DeepGuard improves the secure-and-correct generation rate by an average of 11.9% over strong baselines such as SVEN. It also preserves functional correctness while exhibiting generalization to held-out vulnerability types. Our code is public at https://github.com/unknownhl/DeepGuard.
Authors: Akshit Jindal, Saket Anand, Chetan Arora, Vikram Goyal
Abstract: Organisations with limited data and computational resources increasingly outsource model training to Machine Learning as a Service (MLaaS) providers, who adapt vision-language models (VLMs) such as CLIP to downstream tasks via prompt tuning rather than training from scratch. This semi-honest setting creates a security risk where a malicious provider can follow the prompt-tuning protocol yet implant a backdoor, forcing triggered inputs to be classified into an attacker-chosen class, even for out-of-distribution (OOD) data. Such backdoors leave encoders untouched, making them undetectable to existing methods that focus on encoder corruption. Other data-level methods that sanitize data before training or during inference, also fail to answer the critical question, "Is the delivered model backdoored or not?" To address this model-level verification problem, we introduce CLIP-Inspector (CI), a backdoor detection method designed for prompt-tuned CLIP models. Assuming white-box access to the delivered model and a pool of unlabeled OOD images, CI reconstructs possible triggers for each class to determine if the model exhibits backdoor behaviour or not. Additionally, we demonstrate that using CI's reconstructed trigger for fine-tuning on correctly labeled triggered inputs enables us to re-align the model and reduce backdoor effectiveness. Through extensive experiments across ten datasets and four backdoor attacks, we demonstrate that CI can reconstruct effective triggers in a single epoch using only 1,000 OOD images, achieving a 94% detection accuracy (47/50 models). Compared to adapted trigger-inversion baselines, CI yields a markedly higher AUROC score (0.973 vs 0.495/0.687), thus enabling the vetting and post-hoc repair of prompt-tuned CLIP models to ensure safe deployment.
Authors: Tommy Shaffer Shane, Simon Mylius, Hamish Hobbs
Abstract: Scheming, the covert pursuit of misaligned goals by AI systems, represents a potentially catastrophic risk, yet scheming research suffers from significant limitations. In particular, scheming evaluations demonstrate behaviours that may not occur in real-world settings, limiting scientific understanding, hindering policy development, and not enabling real-time detection of loss of control incidents. Real-world evidence is needed, but current monitoring techniques are not effective for this purpose. This paper introduces a novel open-source intelligence (OSINT) methodology for detecting real-world scheming incidents: collecting and analysing transcripts from chatbot conversations or command-line interactions shared online. Analysing over 183,420 transcripts from X (formerly Twitter), we identify 698 real-world scheming-related incidents between October 2025 and March 2026. We observe a statistically significant 4.9x increase in monthly incidents from the first to last month, compared to a 1.7x increase in posts discussing scheming. We find evidence of multiple scheming-related behaviours in real-world deployments previously reported only in experiments, many resulting in real-world harms. While we did not detect catastrophic scheming incidents, the behaviours observed demonstrate concerning precursors, such as willingness to disregard instructions, circumvent safeguards, lie to users, and single-mindedly pursue goals in harmful ways. As AI systems become more capable, these could evolve into more strategic scheming with potentially catastrophic consequences. Our findings demonstrate the viability of transcript-based OSINT as a scalable approach to real-world scheming detection supporting scientific research, policy development, and emergency response. We recommend further investment towards OSINT techniques for monitoring scheming and loss of control.
Authors: Chenhao Ye, Huaizheng Zhang, Mingcong Han, Baoquan Zhong, Xiang Li, Qixiang Chen, Xinyi Zhang, Weidong Zhang, Kaihua Jiang, Wang Zhang, He Sun, Wencong Xiao, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau
Abstract: Modern LLM reinforcement learning (RL) workloads require a highly efficient weight transfer system to scale training across heterogeneous computational resources. However, existing weight transfer approaches either fail to provide flexibility for dynamically scaling clusters or incur fundamental data movement overhead, resulting in poor performance. We introduce Reference-Oriented Storage (ROS), a new storage abstraction for RL weight transfer that exploits the highly replicated model weights in place. ROS presents the illusion that certain versions of the model weights are stored and can be fetched on demand. Underneath, ROS does not physically store any copies of the weights; instead, it tracks the workers that hold these weights on GPUs for inference. Upon request, ROS directly uses them to serve reads. We build TensorHub, a production-quality system that extends the ROS idea with topology-optimized transfer, strong consistency, and fault tolerance. Evaluation shows that TensorHub fully saturates RDMA bandwidth and adapts to three distinct rollout workloads with minimal engineering effort. Specifically, TensorHub reduces total GPU stall time by up to 6.7x for standalone rollouts, accelerates weight update for elastic rollout by 4.8x, and cuts cross-datacenter rollout stall time by 19x. TensorHub has been deployed in production to support cutting-edge RL training.
Authors: Changi Hong, Yoonah Song, Hwayoung Park, Chaewoon Bang, Dayeon Gu, Do Hyun Lee, Hong Kook Kim
Abstract: Recently, artificial intelligence-based dubbing technology has advanced, enabling automated dubbing (AD) to convert the source speech of a video into target speech in different languages. However, natural AD still faces synchronization challenges such as duration and lip-synchronization (lip-sync), which are crucial for preserving the viewer experience. Therefore, this paper proposes a synchronization method for AD processes that paraphrases translated text, comprising two steps: isochrony for timing constraints and phonetic synchronization (PS) to preserve lip-sync. First, we achieve isochrony by paraphrasing the translated text with a language model, ensuring the target speech duration matches that of the source speech. Second, we introduce PS, which employs dynamic time warping (DTW) with local costs of vowel distances measured from training data so that the target text composes vowels with pronunciations similar to source vowels. Third, we extend this approach to PSComet, which jointly considers semantic and phonetic similarity to preserve meaning better. The proposed methods are incorporated into text-to-speech systems, PS-TTS and PS-Comet TTS. The performance evaluation using Korean and English lip-reading datasets and a voice-actor dubbing dataset demonstrates that both systems outperform TTS without PS on several objective metrics and outperform voice actors in Korean-to-English and English-to-Korean dubbing. We extend the experiments to French, testing all pairs among these languages to evaluate cross-linguistic applicability. Across all language pairs, PS-Comet performed best, balancing lip-sync accuracy with semantic preservation, confirming that PS-Comet achieves more accurate lip-sync with semantic preservation than PS alone.
Authors: Peng Wang (X-LANCE Lab, Shanghai Jiao Tong University), Yanqiao Zhu (X-LANCE Lab, Shanghai Jiao Tong University), Zixuan Jiang (X-LANCE Lab, Shanghai Jiao Tong University), Qinyuan Chen (School of Computer Science, Fudan University), Xingjian Zhao (School of Computer Science, Fudan University), Xipeng Qiu (School of Computer Science, Fudan University), Wupeng Wang (Tongyi Fun Team, Alibaba Group), Zhifu Gao (Tongyi Fun Team, Alibaba Group), Xiangang Li (Tongyi Fun Team, Alibaba Group), Kai Yu (X-LANCE Lab, Shanghai Jiao Tong University), Xie Chen (X-LANCE Lab, Shanghai Jiao Tong University)
Abstract: Recent years have witnessed remarkable progress in automatic speech recognition (ASR), driven by advances in model architectures and large-scale training data. However, two important aspects remain underexplored. First, Word Error Rate (WER), the dominant evaluation metric for decades, treats all words equally and often fails to reflect the semantic correctness of an utterance at the sentence level. Second, interactive correction-an essential component of human communication-has rarely been systematically studied in ASR research. In this paper, we integrate these two perspectives under an agentic framework for interactive ASR. We propose leveraging LLM-as-a-Judge as a semantic-aware evaluation metric to assess recognition quality beyond token-level accuracy. Furthermore, we design an LLM-driven agent framework to simulate human-like multi-turn interaction, enabling iterative refinement of recognition outputs through semantic feedback. Extensive experiments are conducted on standard benchmarks, including GigaSpeech (English), WenetSpeech (Chinese), the ASRU 2019 code-switching test set. Both objective and subjective evaluations demonstrate the effectiveness of the proposed framework in improving semantic fidelity and interactive correction capability. We will release the code to facilitate future research in interactive and agentic ASR.
Authors: Yi-Lun Liao, Alexander J. Hoffman, Sabrina C. Shen, Alexandre Duval, Sam Walton Norwood, Tess Smidt
Abstract: As $SE(3)$-equivariant graph neural networks mature as a core tool for 3D atomistic modeling, improving their efficiency, expressivity, and physical consistency has become a central challenge for large-scale applications. In this work, we introduce EquiformerV3, the third generation of the $SE(3)$-equivariant graph attention Transformer, designed to advance all three dimensions: efficiency, expressivity, and generality. Building on EquiformerV2, we have the following three key advances. First, we optimize the software implementation, achieving $1.75\times$ speedup. Second, we introduce simple and effective modifications to EquiformerV2, including equivariant merged layer normalization, improved feedforward network hyper-parameters, and attention with smooth radius cutoff. Third, we propose SwiGLU-$S^2$ activations to incorporate many-body interactions for better theoretical expressivity and to preserve strict equivariance while reducing the complexity of sampling $S^2$ grids. Together, SwiGLU-$S^2$ activations and smooth-cutoff attention enable accurate modeling of smoothly varying potential energy surfaces (PES), generalizing EquiformerV3 to tasks requiring energy-conserving simulations and higher-order derivatives of PES. With these improvements, EquiformerV3 trained with the auxiliary task of denoising non-equilibrium structures (DeNS) achieves state-of-the-art results on OC20, OMat24, and Matbench Discovery.
Authors: Yushi Feng, Junye Du, Qifan Wang, Zizhan Ma, Qian Niu, Yutaka Matsuo, Long Feng, Lequan Yu
Abstract: Graphical user interface (GUI) agents powered by vision language models (VLMs) are rapidly moving from passive assistance to autonomous operation. However, this unrestricted action space exposes users to severe and irreversible financial, privacy or social harm. Existing safeguards rely on prompt engineering, brittle heuristics and VLM-as-critic lack formal verification and user-tunable guarantees. We propose CORA (COnformal Risk-controlled GUI Agent), a post-policy, pre-action safeguarding framework that provides statistical guarantees on harmful executed actions. CORA reformulates safety as selective action execution: we train a Guardian model to estimate action-conditional risk for each proposed step. Rather than thresholding raw scores, we leverage Conformal Risk Control to calibrate an execute/abstain boundary that satisfies a user-specified risk budget and route rejected actions to a trainable Diagnostician model, which performs multimodal reasoning over rejected actions to recommend interventions (e.g., confirm, reflect, or abort) to minimize user burden. A Goal-Lock mechanism anchors assessment to a clarified, frozen user intent to resist visual injection attacks. To rigorously evaluate this paradigm, we introduce Phone-Harm, a new benchmark of mobile safety violations with step-level harm labels under real-world settings. Experiments on Phone-Harm and public benchmarks against diverse baselines validate that CORA improves the safety--helpfulness--interruption Pareto frontier, offering a practical, statistically grounded safety paradigm for autonomous GUI execution. Code and benchmark are available at cora-agent.github.io.
Authors: Fatma Bet\"ul G\"ure\c{s}, Tanya Nazaretsky, Seyed Parsa Neshaei, Tanja K\"aser
Abstract: Supporting students in developing diagnostic reasoning is a key challenge across educational domains. Novices often face cognitive biases such as premature closure and over-reliance on heuristics, and they struggle to transfer diagnostic strategies to new cases. Scenario-based learning (SBL) enhanced by Learning Analytics (LA) and large language models (LLM) offers a promising approach by combining realistic case experiences with personalized scaffolding. Yet, how different scaffolding approaches shape reasoning processes remains insufficiently explored. This study introduces PharmaSim Switch, an SBL environment for pharmacy technician training, extended with an LA- and LLM-powered pharmacist agent that implements pedagogical conversations rooted in two theory-driven scaffolding approaches: \emph{structuring} and \emph{problematizing}, as well as a student learning trajectory. In a between-groups experiment, 63 vocational students completed a learning scenario, a near-transfer scenario, and a far-transfer scenario under one of the two scaffolding conditions. Results indicate that both scaffolding approaches were effective in supporting the use of diagnostic strategies. Performance outcomes were primarily influenced by scenario complexity rather than students' prior knowledge or the scaffolding approach used. The structuring approach was associated with more accurate Active and Interactive participation, whereas problematizing elicited more Constructive engagement. These findings underscore the value of combining scaffolding approaches when designing LA- and LLM-based systems to effectively foster diagnostic reasoning.
Authors: Yuqin Yang, Haowu Zhou, Haoran Tu, Zhiwen Hui, Shiqi Yan, HaoYang Li, Dong She, Xianrong Yao, Yang Gao, Zhanpeng Jin
Abstract: Most affective computing research treats emotion as a static property of text, focusing on the writer's sentiment while overlooking the reader's perspective. This approach ignores how individual personalities lead to diverse emotional appraisals of the same event. Although role-playing Large Language Models (LLMs) attempt to simulate such nuanced reactions, they often suffer from "personality illusion'' -- relying on surface-level stereotypes rather than authentic cognitive logic. A critical bottleneck is the absence of ground-truth human data to link personality traits to emotional shifts. To bridge the gap, we introduce Persona-E$^2$ (Persona-Event2Emotion), a large-scale dataset grounded in annotated MBTI and Big Five traits to capture reader-based emotional variations across news, social media, and life narratives. Extensive experiments reveal that state-of-the-art LLMs struggle to capture precise appraisal shifts, particularly in social media domains. Crucially, we find that personality information significantly improves comprehension, with the Big Five traits alleviating "personality illusion.'
Authors: Mansour Zoubeirou a Mayaki
Abstract: We develop a theory of generalization and scaling for Mixture-of-Experts (MoE) Transformers that cleanly separates \emph{active} per-input capacity from routing combinatorics. By conditioning on fixed routing patterns and union-bounding across them, we derive a sup-norm covering-number bound whose metric entropy scales with the active parameter budget and incurs a MoE-specific routing overhead. Combined with a standard ERM analysis for squared loss, this yields a generalization bound under a $d$-dimensional manifold data model and $C^\beta$ targets, showing that approximation and estimation trade off as in dense networks once active parameters are accounted for appropriately. We further prove a constructive approximation theorem for MoE architectures, showing that, under the approximation construction, error can decrease either by scaling active capacity or by increasing the number of experts, depending on the dominant bottleneck. From these results we derive neural scaling laws for model size, data size, and compute-optimal tradeoffs. Overall, our results provide a transparent statistical reference point for reasoning about MoE scaling, clarifying which behaviors are certified by worst-case theory and which must arise from data-dependent routing structure or optimization dynamics.
Authors: Avni Mittal
Abstract: LLMs internalize safety policies through RLHF, yet these policies are never formally specified and remain difficult to inspect. Existing benchmarks evaluate models against external standards but do not measure whether models understand and enforce their own stated boundaries. We introduce the Symbolic-Neural Consistency Audit (SNCA), a framework that (1) extracts a model's self-stated safety rules via structured prompts, (2) formalizes them as typed predicates (Absolute, Conditional, Adaptive), and (3) measures behavioral compliance via deterministic comparison against harm benchmarks. Evaluating four frontier models across 45 harm categories and 47,496 observations reveals systematic gaps between stated policy and observed behavior: models claiming absolute refusal frequently comply with harmful prompts, reasoning models achieve the highest self-consistency but fail to articulate policies for 29% of categories, and cross-model agreement on rule types is remarkably low (11%). These results demonstrate that the gap between what LLMs say and what they do is measurable and architecture-dependent, motivating reflexive consistency audits as a complement to behavioral benchmarks.
Authors: Francesca Fati, Felipe Coutinho, Marika Reinius, Marina Rosanu, Gabriel Funingana, Luigi De Vitis, Gabriella Schivardi, Hannah Clayton, Alice Traversa, Zeyu Gao, Guilherme Penteado, Shangqi Gao, Francesco Pastori, Ramona Woitek, Maria Cristina Ghioni, Giovanni Damiano Aletti, Mercedes Jimenez-Linan, Sarah Burge, Nicoletta Colombo, Evis Sala, Maria Francesca Spadea, Timothy L. Kline, James D. Brenton, Jaime Cardoso, Francesco Multinu, Elena De Momi, Mireia Crispin-Ortuzar, Ines P. Machado
Abstract: Purpose. High-grade serous ovarian carcinoma (HGSOC) is characterized by pronounced biological and spatial heterogeneity and is frequently diagnosed at an advanced stage. Neoadjuvant chemotherapy (NACT) followed by delayed primary surgery is commonly employed in patients unsuitable for primary cytoreduction. The Chemotherapy Response Score (CRS) is a validated histopathological biomarker of response to NACT, but it is only available postoperatively. In this study, we investigate whether pre-treatment computed tomography (CT) imaging and clinical data can be used to predict CRS as an investigational decision-support adjunct to inform multidisciplinary team (MDT) discussions regarding expected treatment response. Methods. We proposed a 2.5D multimodal deep learning framework that processes lesion-dense omental slices using a pre-trained Vision Transformer encoder and integrates the resulting visual representations with clinical variables through an intermediate fusion module to predict CRS. Results. Our multimodal model, integrating imaging and clinical data, achieved a ROC-AUC of 0.95 alongside 95% accuracy and 80% precision on the internal test cohort (IEO, n=41 patients). On the external test set (OV04, n=70 patients), it achieved a ROC-AUC of 0.68, alongside 67% accuracy and 75% precision. Conclusion. These preliminary results demonstrate the feasibility of transformer-based deep learning for preoperative prediction of CRS in HGSOC using routine clinical data and CT imaging. As an investigational, pre-treatment decision-support tool, this approach may assist MDT discussions by providing early, non-invasive estimates of treatment response.
Authors: Kobi Hackenburg, Luke Hewitt, Caroline Wagner, Ben M. Tappin, Christopher Summerfield
Abstract: There is substantial concern about the ability of advanced artificial intelligence to influence people's behaviour. A rapidly growing body of research has found that AI can produce large persuasive effects on people's attitudes, but whether AI can persuade people to take consequential real-world actions has remained unclear. In two large preregistered experiments N=17,950 responses from 14,779 people), we used conversational AI models to persuade participants on a range of attitudinal and behavioural outcomes, including signing real petitions and donating money to charity. We found sizable AI persuasion effects on these behavioural outcomes (e.g. +19.7 percentage points on petition signing). However, we observed no evidence of a correlation between AI persuasion effects on attitudes and behaviour. Moreover, we replicated prior findings that information provision drove effects on attitudes, but found no such evidence for our behavioural outcomes. In a test of eight behavioural persuasion strategies, all outperformed the most effective attitudinal persuasion strategy, but differences among the eight were small. Taken together, these results suggest that previous findings relying on attitudinal outcomes may generalize poorly to behaviour, and therefore risk substantially mischaracterizing the real-world behavioural impact of AI persuasion.
Authors: Anas Hattay, Fred Ngole Mboula, Eric Gascard, Zakaria Yahoun
Abstract: Cloud providers must assign heterogeneous compute resources to workflow DAGs while balancing competing objectives such as completion time, cost, and energy consumption. In this work, we study a single-workflow, queue-free scheduling setting and consider a graph neural network (GNN)-based deep reinforcement learning scheduler designed to minimize workflow completion time and energy usage. We identify specific out-of-distribution (OOD) conditions under which GNN-based deep reinforcement learning schedulers fail and provide a principled explanation of why these failures occur. Through controlled OOD evaluations, we demonstrate that performance degradation stems from structural mismatches between training and deployment environments, which disrupt message passing and undermine policy generalization. Our analysis exposes fundamental limitations of current GNN-based schedulers and highlights the need for more robust representations to ensure reliable scheduling performance under distribution shifts.
Authors: Yunqiang Wang, Hengyuan Na, Di Wu, Miao Hu, Guocong Quan
Abstract: Audio large language models (ALLMs) enable rich speech-text interaction, but they also introduce jailbreak vulnerabilities in the audio modality. Existing audio jailbreak methods mainly optimize jailbreak success while overlooking utility preservation, as reflected in transcription quality and question answering performance. In practice, stronger attacks often come at the cost of degraded utility. To study this trade-off, we revisit existing attacks by varying their perturbation coverage in the frequency domain, from partial-band to full-band, and find that broader frequency coverage does not necessarily improve jailbreak performance, while utility consistently deteriorates. This suggests that concentrating perturbation on a subset of bands can yield a better attack-utility trade-off than indiscriminate full-band coverage. Based on this insight, we propose GRM, a utility-aware frequency-selective jailbreak framework. It ranks Mel bands by their attack contribution relative to utility sensitivity, perturbs only a selected subset of bands, and learns a reusable universal perturbation under a semantic-preservation objective. Experiments on four representative ALLMs show that GRM achieves an average Jailbreak Success Rate (JSR) of 88.46% while providing a better attack-utility trade-off than representative baselines. These results highlight the potential of frequency-selective perturbation for better balancing attack effectiveness and utility preservation in audio jailbreak. Content Warning: This paper includes harmful query examples and unsafe model responses.
Authors: Esila Keskin
Abstract: Von Economo neurons (VENs) are large bipolar projection neurons found exclusively in the anterior cingulate cortex (ACC) and frontal insula of species with complex social cognition, including humans, great apes, and cetaceans. Their selective depletion in frontotemporal dementia (FTD) and altered development in autism implicate them in rapid social decision-making, yet no computational model of VEN function has previously existed. We introduce the Fast Lane Hypothesis: VENs implement a biological speed-accuracy tradeoff (SAT) by providing a sparse, fast projection pathway that enables rapid social decisions at the cost of deliberate processing accuracy. We model VENs as fast leaky integrate-and-fire (LIF) neurons with membrane time constant 5 ms and sparse dendritic fan-in of eight afferents, compared to 20 ms and eighty afferents for standard pyramidal neurons, within a spiking cortical circuit of 2,000 neurons trained on a social discrimination task. Networks are evaluated under three clinically motivated conditions across 10 independent random seeds: typical (2% VENs), autism-like (0.4% VENs), and FTD-like (post-training VEN ablation). All configurations achieve equivalent asymptotic classification accuracy (99.4%), consistent with the prediction that VENs modulate decision speed rather than representational capacity. Temporal analysis confirms that VENs produce median first-spike latencies 4 ms earlier than pyramidal neurons. At a fixed decision threshold, the typical condition is significantly faster than FTD-like (t=-23.31, p<0.0001), while autism-like is intermediate (mean RT=26.91+/-9.01 ms vs. typical 20.70+/-2.02 ms; p=0.078). A preliminary evolutionary analysis shows qualitative correspondence between model-optimal VEN fraction and the primate phylogenetic gradient. To our knowledge, this is the first computational model that asks what a Von Economo neuron actually computes.
Authors: Zizhao Li, Zhengkang Xiang, Jiayang Ao, Feng Liu, Joseph West, Kourosh Khoshelham
Abstract: LiDAR-based perception is critical for autonomous driving due to its robustness to poor lighting and visibility conditions. Yet, current models operate under the closed-set assumption and often fail to recognize unexpected out-of-distribution (OOD) objects in the open world. Existing OOD scoring functions exhibit limited performance because they ignore the pronounced class imbalance inherent in LiDAR OOD detection and assume a uniform class distribution. To address this limitation, we propose the Neural Distribution Prior (NDP), a framework that models the distributional structure of network predictions and adaptively reweights OOD scores based on alignment with a learned distribution prior. NDP dynamically captures the logit distribution patterns of training data and corrects class-dependent confidence bias through an attention-based module. We further introduce a Perlin noise-based OOD synthesis strategy that generates diverse auxiliary OOD samples from input scans, enabling robust OOD training without external datasets. Extensive experiments on the SemanticKITTI and STU benchmarks demonstrate that NDP substantially improves OOD detection performance, achieving a point-level AP of 61.31\% on the STU test set, which is more than 10$\times$ higher than the previous best result. Our framework is compatible with various existing OOD scoring formulations, providing an effective solution for open-world LiDAR perception.
Authors: Augustin Chan
Abstract: The King Wen sequence of the I-Ching (c. 1000 BC) orders 64 hexagrams -- states of a six-dimensional binary space -- in a pattern that has puzzled scholars for three millennia. We present a rigorous statistical characterization of this ordering using Monte Carlo permutation analysis against 100,000 random baselines. We find that the sequence has four statistically significant properties: higher-than-random transition distance (98.2nd percentile), negative lag-1 autocorrelation (p=0.037), yang-balanced groups of four (p=0.002), and asymmetric within-pair vs. between-pair distances (99.2nd percentile). These properties superficially resemble principles from curriculum learning and curiosity-driven exploration, motivating the hypothesis that they might benefit neural network training. We test this hypothesis through three experiments: learning rate schedule modulation, curriculum ordering, and seed sensitivity analysis, conducted across two hardware platforms (NVIDIA RTX 2060 with PyTorch and Apple Silicon with MLX). The results are uniformly negative. King Wen LR modulation degrades performance at all tested amplitudes. As curriculum ordering, King Wen is the worst non-sequential ordering on one platform and within noise on the other. A 30-seed sweep confirms that only King Wen's degradation exceeds natural seed variance. We explain why: the sequence's high variance -- the very property that makes it statistically distinctive -- destabilizes gradient-based optimization. Anti-habituation in a fixed combinatorial sequence is not the same as effective training dynamics.
Authors: Suhita Ghosh, Yamini Sinha, Sebastian Stober
Abstract: Differentiable Digital Signal Processing (DDSP) pipelines for voice conversion rely on subtractive synthesis, where a periodic excitation signal is shaped by a learned spectral envelope to reconstruct the target voice. In DDSP-QbE, the excitation is generated via phase accumulation, producing a sawtooth-like waveform whose abrupt discontinuities introduce aliasing artefacts that manifest perceptually as buzziness and spectral distortion, particularly at higher fundamental frequencies. We propose two targeted improvements to the excitation stage of the DDSP-QbE subtractive synthesizer. First, we incorporate explicit voicing detection to gate the harmonic excitation, suppressing the periodic component in unvoiced regions and replacing it with filtered noise, thereby avoiding aliased harmonic content where it is most perceptually disruptive. Second, we apply Polynomial Band-Limited Step (PolyBLEP) correction to the phase-accumulated oscillator, substituting the hard waveform discontinuity at each phase wrap with a smooth polynomial residual that cancels alias-generating components without oversampling or spectral truncation. Together, these modifications yield a cleaner harmonic roll-off, reduced high-frequency artefacts, and improved perceptual naturalness, as measured by MOS. The proposed approach is lightweight, differentiable, and integrates seamlessly into the existing DDSP-QbE training pipeline with no additional learnable parameters.
Authors: Yuqin Lan, Gen Li, Yuanze Hu, Weihao Shen, Zhaoxin Fan, Faguo Wu, Xiao Zhang, Laurence T. Yang, Zhiming Zheng
Abstract: Vision-Language Models (VLMs) are powerful but remain vulnerable to multimodal jailbreak attacks. Existing attacks mainly rely on either explicit visual prompt attacks or gradient-based adversarial optimization. While the former is easier to detect, the latter produces subtle perturbations that are less perceptible, but is usually optimized and evaluated under homogeneous open-source surrogate-target settings, leaving its effectiveness on commercial closed-source VLMs under heterogeneous settings unclear. To examine this issue, we study different surrogate-target settings and observe a consistent gap between homogeneous and heterogeneous settings, a phenomenon we term surrogate dependency. Motivated by this finding, we propose Mosaic, a Multi-view ensemble optimization framework for multimodal jailbreak against closed-source VLMs, which alleviates surrogate dependency under heterogeneous surrogate-target settings by reducing over-reliance on any single surrogate model and visual view. Specifically, Mosaic incorporates three core components: a Text-Side Transformation module, which perturbs refusal-sensitive lexical patterns; a Multi-View Image Optimization module, which updates perturbations under diverse cropped views to avoid overfitting to a single visual view; and a Surrogate Ensemble Guidance module, which aggregates optimization signals from multiple surrogate VLMs to reduce surrogate-specific bias. Extensive experiments on safety benchmarks demonstrate that Mosaic achieves state-of-the-art Attack Success Rate and Average Toxicity against commercial closed-source VLMs.
Authors: Jingzhi Gong, Ruizhen Gu, Zhiwei Fei, Yazhuo Cao, Lukas Twist, Alina Geiger, Shuo Han, Dominik Sobania, Federica Sarro, Jie M. Zhang
Abstract: Agent skills provide modular, task-specific guidance for LLM- based coding agents, but manually tuning skill bundles to balance success rate, cost, and runtime is expensive and fragile. We present SkillMOO, a multi-objective optimization framework that automatically evolves skill bundles using LLM-proposed edits and NSGA-II survivor selection: a solver agent evaluates candidate skill bundles on coding tasks and an optimizer agent proposes bundle edits based on failure analysis. On three SkillsBench software engineering tasks, SkillMOO improves pass rate by up to 131% while reducing cost up to 32% relative to the best baseline per task at low optimization overhead. Pattern analysis reveals pruning and substitution as primary drivers of improvement, suggesting effective bundles favor minimal, focused content over accumulated instructions.
Authors: Tobias Meuser, Jannis Weil, Aninda Lahiri, Marius Paraschiv
Abstract: Quantum networks are expected to become a key enabler for interconnecting quantum devices. In contrast to classical communication networks, however, information transfer in quantum networks is usually restricted to short distances due to physical constraints of entanglement distribution. Satellites can extend entanglement distribution over long distances, but routing in such networks is challenging because satellite motion and stochastic link generation create a highly dynamic quantum topology. Existing routing methods often rely on global topology information that quickly becomes outdated due to delays in the classical control plane, while decentralized methods typically act on incomplete local information. We propose SatQNet, a reinforcement learning approach for entanglement routing in satellite-assisted quantum networks that can be decentralized at runtime. Its key innovation is an edge-centric directed line graph neural network that performs local message passing on directed edge embeddings, enabling it to better capture link properties in high-degree and time-varying topologies. By exchanging messages with neighboring repeaters, SatQNet learns a local graph representation at runtime that supports agents in establishing high-fidelity end-to-end entanglements. Trained on random graphs, SatQNet outperforms heuristic and learning-based approaches across diverse settings, including a real-world European backbone topology, and generalizes to unseen topologies without retraining.
Authors: Zengbin Wang, Feng Xiong, Liang Lin, Xuecai Hu, Yong Wang, Yanlin Wang, Man Zhang, Xiangxiang Chu
Abstract: Reinforcement learning with verifiable rewards (RLVR) has significantly advanced the reasoning ability of vision-language models (VLMs). However, the inherent text-dominated nature of VLMs often leads to insufficient visual faithfulness, characterized by sparse attention activation to visual tokens. More importantly, our empirical analysis reveals that temporal visual forgetting along reasoning steps exacerbates this deficiency. To bridge this gap, we propose Visually-Guided Policy Optimization (VGPO), a novel framework to reinforce visual focus during policy optimization. Specifically, VGPO initially introduces a Visual Attention Compensation mechanism that leverages visual similarity to localize and amplify visual cues, while progressively elevating visual expectations in later steps to counteract visual forgetting. Building on this mechanism, we implement a dual-grained advantage re-weighting strategy: the intra-trajectory level highlights tokens exhibiting relatively high visual activation, while the inter-trajectory level prioritizes trajectories demonstrating superior visual accumulation. Extensive experiments demonstrate that VGPO achieves better visual activation and superior performance in mathematical multimodal reasoning and visual-dependent tasks.
Authors: Peng Ding
Abstract: The rapid proliferation of Large Language Model (LLM) providers--each exposing proprietary API formats--has created a fragmented ecosystem where applications become tightly coupled to individual vendors. Switching or bridging providers requires $O(N^2)$ bilateral adapters, impeding portability and multi-provider architectures. We observe that despite substantial syntactic divergence, the major LLM APIs share a common semantic core: the practical challenge is the combinatorial surface of syntactic variations, not deep semantic incompatibility. Based on this finding, we present LLM-Rosetta, an open-source translation framework built on a hub-and-spoke Intermediate Representation (IR) that captures the shared semantic core--messages, content parts, tool calls, reasoning traces, and generation controls--in a 9-type content model and 10-type stream event schema. A modular Ops-composition converter architecture enables each API standard to be added independently. LLM-Rosetta supports bidirectional conversion (provider-to-IR-to-provider) for both request and response payloads, including chunk-level streaming with stateful context management. We implement converters for four API standards (OpenAI Chat Completions, OpenAI Responses, Anthropic Messages, and Google GenAI), covering the vast majority of commercial providers. Empirical evaluation demonstrates lossless round-trip fidelity, correct streaming behavior, and sub-100 microsecond conversion overhead--competitive with LiteLLM's single-pass approach while providing bidirectionality and provider neutrality. LLM-Rosetta passes the Open Responses compliance suite and is deployed in production at Argonne National Laboratory. Code is available at https://github.com/Oaklight/llm-rosetta.
Authors: Guiyao Tie, Jiawen Shi, Pan Zhou, Lichao Sun
Abstract: Agent ecosystems increasingly rely on installable skills to extend functionality, and some skills bundle learned model artifacts as part of their execution logic. This creates a supply-chain risk that is not captured by prompt injection or ordinary plugin misuse: a third-party skill may appear benign while concealing malicious behavior inside its bundled model. We present BadSkill, a backdoor attack formulation that targets this model-in-skill threat surface. In BadSkill, an adversary publishes a seemingly benign skill whose embedded model is backdoor-fine-tuned to activate a hidden payload only when routine skill parameters satisfy attacker-chosen semantic trigger combinations. To realize this attack, we train the embedded classifier with a composite objective that combines classification loss, margin-based separation, and poison-focused optimization, and evaluate it in an OpenClaw-inspired simulation environment that preserves third-party skill installation and execution while enabling controlled multi-model study. Our benchmark spans 13 skills, including 8 triggered tasks and 5 non-trigger control skills, with a combined main evaluation set of 571 negative-class queries and 396 trigger-aligned queries. Across eight architectures (494M--7.1B parameters) from five model families, BadSkill achieves up to 99.5\% average attack success rate (ASR) across the eight triggered skills while maintaining strong benign-side accuracy on negative-class queries. In poison-rate sweeps on the standard test split, a 3\% poison rate already yields 91.7\% ASR. The attack remains effective across the evaluated model scales and under five text perturbation types. These findings identify model-bearing skills as a distinct model supply-chain risk in agent ecosystems and motivate stronger provenance verification and behavioral vetting for third-party skill artifacts.
Authors: Andy Anderson
Abstract: AI coding tools are widely adopted, but most teams plateau at prompt-and-review without a framework for systematic progression. This paper presents the AI Codebase Maturity Model (ACMM), a 5-level framework describing how codebases evolve from basic AI-assisted coding to self-sustaining systems. Inspired by CMMI, each level is defined by its feedback loop topology the specific mechanisms that must exist before the next level becomes possible. I validate the model through a 4-month experience report maintaining KubeStellar Console, a CNCF Kubernetes dashboard built from scratch with Claude Code (Opus) and GitHub Copilot. The system currently operates with 63 CI/CD workflows, 32 nightly test suites, 91% code coverage, and achieves bug-to-fix times under 30 minutes 24 hours a day. The central finding: the intelligence of an AI-driven development system resides not in the AI model itself, but in the infrastructure of instructions, tests, metrics, and feedback loops that surround it. You cannot skip levels, and at each level, the thing that unlocks the next one is another feedback mechanism. Testing the volume of test cases, the coverage thresholds, and the reliability of test execution proved to be the single most important investment in the entire journey.
Authors: Wiebke Hutiri, Morgan Scheuerman, Shruti Nagpal, Austin Hoag, Alice Xiang
Abstract: This paper argues that a one-size-fits-all approach to specifying consent for the use of creative works in generative AI is insufficient. Real-world ownership and rights holder structures, the imitation of artistic styles and likeness, and the limitless contexts of use of AI outputs make the status quo of binary consent with opt-in by default untenable. To move beyond the current impasse, we consider levers of control in generative AI workflows at training, inference, and dissemination. Based on these insights, we position inference-time opt-in as an overlooked opportunity for nuanced consent verification. We conceptualize nuanced consent conditions for opt-in and propose an agent-based inference-time opt-in architecture to verify if user intent requests meet conditional consent granted by rights holders. In a case study for music, we demonstrate that nuanced opt-in at inference can account for established rights and re-establish a balance of power between rights holders and AI developers.
Authors: Siyuan Zhou, Hejun Wang, Hu Cheng, Jinxi Li, Dongsheng Wang, Junwei Jiang, Yixiao Jin, Jiayue Huang, Shiwei Mao, Shangjia Liu, Yafei Yang, Hongkang Song, Shenxing Wei, Zihui Zhang, Peng Huang, Shijie Liu, Zhengli Hao, Hao Li, Yitian Li, Wenqi Zhou, Zhihan Zhao, Zongqi He, Hongtao Wen, Shouwang Huang, Peng Yun, Bowen Cheng, Pok Kazaf Fu, Wai Kit Lai, Jiahao Chen, Kaiyuan Wang, Zhixuan Sun, Ziqi Li, Haochen Hu, Di Zhang, Chun Ho Yuen, Bing Wang, Zhihua Wang, Chuhang Zou, Bo Yang
Abstract: We present PhysInOne, a large-scale synthetic dataset addressing the critical scarcity of physically-grounded training data for AI systems. Unlike existing datasets limited to merely hundreds or thousands of examples, PhysInOne provides 2 million videos across 153,810 dynamic 3D scenes, covering 71 basic physical phenomena in mechanics, optics, fluid dynamics, and magnetism. Distinct from previous works, our scenes feature multiobject interactions against complex backgrounds, with comprehensive ground-truth annotations including 3D geometry, semantics, dynamic motion, physical properties, and text descriptions. We demonstrate PhysInOne's efficacy across four emerging applications: physics-aware video generation, long-/short-term future frame prediction, physical property estimation, and motion transfer. Experiments show that fine-tuning foundation models on PhysInOne significantly enhances physical plausibility, while also exposing critical gaps in modeling complex physical dynamics and estimating intrinsic properties. As the largest dataset of its kind, orders of magnitude beyond prior works, PhysInOne establishes a new benchmark for advancing physics-grounded world models in generation, simulation, and embodied AI.
Authors: Sanchita S. Kamath, Aziz N Zeidieh, Venkatesh Potluri, Sile O'Modhrain, Kenneth Perry, JooYoung Seo
Abstract: Three-dimensional (3D) data visualizations, such as surface plots, are vital in STEM fields from biomedical imaging to spectroscopy, yet remain largely inaccessible to blind and low-vision (BLV) people. To address this gap, we conducted an Experience-Based Co-Design with BLV co-designers with expertise in non-visual data representations to create an accessible, multi-modal, web-native visualization tool. Using a multi-phase methodology, our team of five BLV and one non-BLV researcher(s) participated in two iterative sessions, comparing a low-fidelity tactile probe with a high-fidelity digital prototype. This process produced a prototype with empirically grounded features, including reference sonification, stereo and volumetric audio, and configurable buffer aggregation, which our co-designers validated as improving analytic accuracy and learnability. In this study, we target core analytic tasks essential for non-visual 3D data exploration: orientation, landmark and peak finding, comparing local maxima versus global trends, gradient tracing, and identifying occluded or partially hidden features. Our work offers accessibility researchers and developers a co-design protocol for translating tactile knowledge to digital interfaces, concrete design guidance for future systems, and opportunities to extend accessible 3D visualization into embodied data environments.
Authors: Wonbong Jang, Shikun Liu, Soubhik Sanyal, Juan Camilo Perez, Kam Woh Ng, Sanskar Agrawal, Juan-Manuel Perez-Rua, Yiannis Douratsos, Tao Xiang
Abstract: Recovering camera parameters from images and rendering scenes from novel viewpoints have long been treated as separate tasks in computer vision and graphics. This separation breaks down when image coverage is sparse or poses are ambiguous, since each task needs what the other produces. We propose Rays as Pixels, a Video Diffusion Model (VDM) that learns a joint distribution over videos and camera trajectories. We represent each camera as dense ray pixels (raxels) and denoise them jointly with video frames through Decoupled Self-Cross Attention mechanism. A single trained model handles three tasks: predicting camera trajectories from video, jointly generating video and camera trajectory from input images, and generating video from input images along a target camera trajectory. Because the model can both predict trajectories from a video and generate views conditioned on its own predictions, we evaluate it through a closed-loop self-consistency test, demonstrating that its forward and inverse predictions agree. Notably, trajectory prediction requires far fewer denoising steps than video generation, even a few denoising steps suffice for self-consistency. We report results on pose estimation and camera-controlled video generation.
Authors: Dario Maio
Abstract: Text embeddings are central to modern information retrieval and Retrieval-Augmented Generation (RAG). While dense models derived from Large Language Models (LLMs) dominate current practice, recent work has explored quantum-inspired alternatives motivated by the geometric properties of Hilbert-like spaces and their potential to encode richer semantic structure. This paper presents an experimental framework for constructing quantum-inspired 1024-dimensional document embeddings based on overlapping windows and multi-scale aggregation. The pipeline combines semantic projections (e.g., EigAngle), circuit-inspired feature mappings, and optional teacher-student distillation, together with a fingerprinting mechanism for reproducibility and controlled evaluation. We introduce a set of diagnostic tools for hybrid retrieval, including static and dynamic interpolation between BM25 and embedding-based scores, candidate union strategies, and a conceptual alpha-oracle that provides an upper bound for score-level fusion. Experiments on controlled corpora of Italian and English documents across technical, narrative, and legal domains, using synthetic queries, show that BM25 remains a strong baseline, teacher embeddings provide stable semantic structure, and standalone quantum-inspired embeddings exhibit weak and unstable ranking signals. Distillation yields mixed effects, improving alignment in some cases but not consistently enhancing retrieval performance, while hybrid retrieval can recover competitive results when lexical and embedding-based signals are combined. Overall, the results highlight structural limitations in the geometry of quantum-inspired embeddings, including distance compression and ranking instability, and clarify their role as auxiliary components rather than standalone retrieval representations.
Authors: Yuning Wang, Pol Suarez, Mathis Bode, Ricardo Vinuesa
Abstract: Turbulent boundary layers over aerodynamic surfaces are a major source of aircraft drag, yet their control remains challenging due to multiscale dynamics and spatial variability, particularly under adverse pressure gradients. Reinforcement learning has outperformed state-of-the-art strategies in canonical flows, but its application to realistic geometries is limited by computational cost and transferability. Here we show that these limitations can be overcome by exploiting local structures of wall-bounded turbulence. Policies are trained in turbulent channel flows matched to wing boundary-layer statistics and deployed directly onto a NACA4412 wing at $Re_c=2\times10^5$ without further training, being the so-called zero-shot control. This achieves a 28.7\% reduction in skin-friction drag and a 10.7\% reduction in total drag, outperforming the state-of-the-art opposition control by 40\% in friction drag reduction and 5\% in total drag. Training cost is reduced by four orders of magnitude relative to on-wing training, enabling scalable flow control.
Authors: Qingzhuo Wang, Leilei Wen, Juntao Chen, Kunyu Peng, Ruiyang Qin, Zhihua Wei, Wen Shen
Abstract: In this paper, we propose a sequential recommendation model that integrates Time-aware personalization, Multi-interest personalization, and Explanation personalization for Personalized Sequential Recommendation (TME-PSR). That is, we consider the differences across different users in temporal rhythm preference, multiple fine-grained latent interests, and the personalized semantic alignment between recommendations and explanations. Specifically, the proposed TME-PSR model employs a dual-view gated time encoder to capture personalized temporal rhythms, a lightweight multihead Linear Recurrent Unit architecture that enables fine-grained sub-interest modeling with improved efficiency, and a dynamic dual-branch mutual information weighting mechanism to achieve personalized alignment between recommendations and explanations. Extensive experiments on real-world datasets demonstrate that our method consistently improves recommendation accuracy and explanation quality, at a lower computational cost.
Authors: Jingyu Zhang, Tianjian Li, William Jurayj, Hongyuan Zhan, Benjamin Van Durme, Daniel Khashabi
Abstract: Large language model agents receive instructions from many sources-system messages, user prompts, tool outputs, and more-each carrying different levels of trust and authority. When these instructions conflict, models must reliably follow the highest-privilege instruction to remain safe and effective. The dominant paradigm, instruction hierarchy (IH), assumes a fixed, small set of privilege levels (typically fewer than five) defined by rigid role labels (e.g., system > user). This is inadequate for real-world agentic settings, where conflicts can arise across far more sources and contexts. In this work, we propose Many-Tier Instruction Hierarchy (ManyIH), a paradigm for resolving instruction conflicts among instructions with arbitrarily many privilege levels. We introduce ManyIH-Bench, the first benchmark for ManyIH. ManyIH-Bench requires models to navigate up to 12 levels of conflicting instructions with varying privileges, comprising 853 agentic tasks (427 coding and 426 instruction-following). ManyIH-Bench composes constraints developed by LLMs and verified by humans to create realistic and difficult test cases spanning 46 real-world agents. Our experiments show that even the current frontier models perform poorly (~40% accuracy) when instruction conflict scales. This work underscores the urgent need for methods that explicitly target fine-grained, scalable instruction conflict resolution in agentic settings.
Authors: Lifeng Chen, Tianqi You, Hao Liu, Zhimin Bao, Jile Jiao, Xiao Han, Zhicai Ou, Tao Sun, Xiaofeng Mou, Xiaojie Jin, Yi Xu
Abstract: Chest X-ray report generation (CXR-RG) has the potential to substantially alleviate radiologists' workload. However, conventional autoregressive vision--language models (VLMs) suffer from high inference latency due to sequential token decoding. Diffusion-based models offer a promising alternative through parallel generation, but they still require multiple denoising iterations. Compressing multi-step denoising to a single step could further reduce latency, but often degrades textual coherence due to the mean-field bias introduced by token-factorized denoisers. To address this challenge, we propose \textbf{ECHO}, an efficient diffusion-based VLM (dVLM) for chest X-ray report generation. ECHO enables stable one-step-per-block inference via a novel Direct Conditional Distillation (DCD) framework, which mitigates the mean-field limitation by constructing unfactorized supervision from on-policy diffusion trajectories to encode joint token dependencies. In addition, we introduce a Response-Asymmetric Diffusion (RAD) training strategy that further improves training efficiency while maintaining model effectiveness. Extensive experiments demonstrate that ECHO surpasses state-of-the-art autoregressive methods, improving RaTE and SemScore by \textbf{64.33\%} and \textbf{60.58\%} respectively, while achieving an \textbf{$8\times$} inference speedup without compromising clinical accuracy.
Authors: Maksim Anisimov (Imperial College London), Francesco Belardinelli (Imperial College London), Matthew Wicker (Imperial College London)
Abstract: Safety guarantees are a prerequisite to the deployment of reinforcement learning (RL) agents in safety-critical tasks. Often, deployment environments exhibit non-stationary dynamics or are subject to changing performance goals, requiring updates to the learned policy. This leads to a fundamental challenge: how to update an RL policy while preserving its safety properties on previously encountered tasks? The majority of current approaches either do not provide formal guarantees or verify policy safety only a posteriori. We propose a novel a priori approach to safe policy updates in continual RL by introducing the Rashomon set: a region in policy parameter space certified to meet safety constraints within the demonstration data distribution. We then show that one can provide formal, provable guarantees for arbitrary RL algorithms used to update a policy by projecting their updates onto the Rashomon set. Empirically, we validate this approach across grid-world navigation environments (Frozen Lake and Poisoned Apple) where we guarantee an a priori provably deterministic safety on the source task during downstream adaptation. In contrast, we observe that regularisation-based baselines experience catastrophic forgetting of safety constraints while our approach enables strong adaptation with provable guarantees that safety is preserved.
Authors: Zukun Zhang, Kai Shu, Mingqiao Mo
Abstract: Learning-based quadruped controllers achieve impressive agility but typically lack formal safety guarantees under model uncertainty, perception noise, and unstructured contact conditions. We introduce SafeMind, a differentiable stochastic safety-control framework that unifies probabilistic Control Barrier Functions with semantic context understanding and meta-adaptive risk calibration. SafeMind explicitly models epistemic and aleatoric uncertainty through a variance-aware barrier constraint embedded in a differentiable quadratic program, thereby preserving gradient flow for end-to-end training. A semantics-to-constraint encoder modulates safety margins using perceptual or language cues, while a meta-adaptive learner continuously adjusts risk sensitivity across environments. We provide theoretical conditions for probabilistic forward invariance, feasibility, and stability under stochastic dynamics. SafeMind is deployed on Unitree A1 and ANYmal C at 200~Hz and validated across 12 terrain types, dynamic obstacles, morphology perturbations, and semantically defined tasks. Experiments show that SafeMind reduces safety violations by 3--10x and energy consumption by 10--15% relative to state-of-the-art CBF, MPC, and hybrid RL baselines, while maintaining real-time control performance.
Authors: Israt Jahan Mouri, Muhammad Ridowan, Muhammad Abdullah Adnan
Abstract: Model poisoning attacks pose a significant security threat to Federated Learning (FL). Most existing model poisoning attacks rely on collusion, requiring adversarial clients to coordinate by exchanging local benign models and synchronizing the generation of their poisoned updates. However, sustaining such coordination is increasingly impractical in real-world FL deployments, as it effectively requires botnet-like control over many devices. This approach is costly to maintain and highly vulnerable to detection. This context raises a fundamental question: Can model poisoning attacks remain effective without any communication between attackers? To address this challenge, we introduce and formalize the \textbf{non-collusive attack model}, in which all compromised clients share a common adversarial objective but operate independently. Under this model, each attacker generates its malicious update without communicating with other adversaries, accessing other clients' updates, or relying on any knowledge of server-side defenses. To demonstrate the feasibility of this threat model, we propose \textbf{XFED}, the first aggregation-agnostic, non-collusive model poisoning attack. Our empirical evaluation across six benchmark datasets shows that XFED bypasses eight state-of-the-art defenses and outperforms six existing model poisoning attacks. These findings indicate that FL systems are substantially less secure than previously believed and underscore the urgent need for more robust and practical defense mechanisms.
Authors: Kyle Whitecross, Negin Rahimi
Abstract: We propose RecaLLM, a set of reasoning language models post-trained to make effective use of long-context information. In-context retrieval, which identifies relevant evidence from context, and reasoning are deeply intertwined: retrieval supports reasoning, while reasoning often determines what must be retrieved. However, their interaction remains largely underexplored. In preliminary experiments on several open-source LLMs, we observe that in-context retrieval performance substantially degrades even after a short reasoning span, revealing a key bottleneck for test-time scaling that we refer to as lost-in-thought: reasoning steps that improve performance also make subsequent in-context retrieval more challenging. To address this limitation, RecaLLM interleaves reasoning with explicit in-context retrieval, alternating between reasoning and retrieving context information needed to solve intermediate subproblems. We introduce a negligible-overhead constrained decoding mechanism that enables verbatim copying of evidence spans, improving the grounding of subsequent generation. Trained on diverse lexical and semantic retrieval tasks, RecaLLM achieves strong performance on two long-context benchmarks, RULER and HELMET, significantly outperforming baselines. Notably, we observe consistent gains at context windows of up to 128K tokens using training samples of at most 10K tokens, far shorter than those used by existing long-context approaches, highlighting a promising path toward improving long-context performance without expensive long-context training data.
Authors: Hippolyte Gisserot-Boukhlef, Nicolas Boizard, Emmanuel Malherbe, C\'eline Hudelot, Pierre Colombo
Abstract: Accurate evaluation is central to the large language model (LLM) ecosystem, guiding model selection and downstream adoption across diverse use cases. In practice, however, evaluating generative outputs typically relies on rigid lexical methods to extract and assess answers, which can conflate a model's true problem-solving ability with its compliance with predefined formatting guidelines. While recent LLM-as-a-Judge approaches mitigate this issue by assessing semantic correctness rather than strict structural conformity, they also introduce substantial computational overhead, making evaluation costly. In this work, we first systematically investigate the limitations of lexical evaluation through a large-scale empirical study spanning 36 models and 15 downstream tasks, demonstrating that such methods correlate poorly with human judgments. To address this limitation, we introduce BERT-as-a-Judge, an encoder-driven approach for assessing answer correctness in reference-based generative settings, robust to variations in output phrasing, and requiring only lightweight training on synthetically annotated question-candidate-reference triplets. We show that it consistently outperforms the lexical baseline while matching the performance of much larger LLM judges, providing a compelling tradeoff between the two and enabling reliable, scalable evaluation. Finally, through extensive experimentation, we provide detailed insights into BERT-as-a-Judge's performance to offer practical guidance for practitioners, and release all project artifacts to foster downstream adoption.
Authors: Yucheng Shen, Jiulong Wu, Jizhou Huang, Dawei Yin, Lingyong Yan, Min Cao
Abstract: Visual Retrieval-Augmented Generation (VRAG) empowers Vision-Language Models to retrieve and reason over visually rich documents. To tackle complex queries requiring multi-step reasoning, agentic VRAG systems interleave reasoning with iterative retrieval.. However, existing agentic VRAG faces two critical bottlenecks. (1) Visual Evidence Sparsity: key evidence is scattered across pages yet processed in isolation, hindering cross-page reasoning; moreover, fine-grained intra-image evidence often requires precise visual actions, whose misuse degrades retrieval quality; (2) Search Drift in Long Horizons: the accumulation of visual tokens across retrieved pages dilutes context and causes cognitive overload, leading agents to deviate from their search objective. To address these challenges, we propose VISOR (Visual Retrieval-Augmented Generation via Iterative Search and Over-horizon Reasoning), a unified single-agent framework. VISOR features a structured Evidence Space for progressive cross-page reasoning, coupled with a Visual Action Evaluation and Correction mechanism to manage visual actions. Additionally, we introduce a Dynamic Trajectory with Sliding Window and Intent Injection to mitigate search drift. They anchor the evidence space while discarding earlier raw interactions, preventing context from being overwhelmed by visual tokens. We train VISOR using a Group Relative Policy Optimization-based Reinforcement Learning (GRPO-based RL) pipeline with state masking and credit assignment tailored for dynamic context reconstruction. Extensive experiments on ViDoSeek, SlideVQA, and MMLongBench demonstrate that VISOR achieves state-of-the-art performance with superior efficiency for long-horizon visual reasoning tasks.
Authors: Anthony T. Nixon
Abstract: When two agents of different computational capacities interact with the same environment, they need not compress a common semantic alphabet differently; they can induce different semantic alphabets altogether. We show that the quotient POMDP $Q_{m,T}(M)$ - the unique coarsest abstraction consistent with an agent's capacity - serves as a capacity-derived semantic space for any bounded agent, and that communication between heterogeneous agents exhibits a sharp structural phase transition. Below a critical rate $R_{\text{crit}}$ determined by the quotient mismatch, intent-preserving communication is structurally impossible. In the supported one-way memoryless regime, classical side-information coding then yields exponential decay above the induced benchmark. Classical coding theorems tell you the rate once the source alphabet is fixed; our contribution is to derive that alphabet from bounded interaction itself. Concretely, we prove: (1) a fixed-$\varepsilon$ structural phase-transition theorem whose lower bound is fully general on the common-history quotient comparison; (2) a one-way Wyner-Ziv benchmark identification on quotient alphabets, with exact converse, exact operational equality for memoryless quotient sources, and an ergodic long-run bridge via explicit mixing bounds; (3) an asymptotic one-way converse in the shrinking-distortion regime $\varepsilon = O(1/T)$, proved from the message stream and decoder side information; and (4) alignment traversal bounds enabling compositional communication through intermediate capacity levels. Experiments on eight POMDP environments (including RockSample(4,4)) illustrate the phase transition, a structured-policy benchmark shows the one-way rate can drop by up to $19\times$ relative to the counting bound, and a shrinking-distortion sweep matches the regime of the asymptotic converse.
Authors: Stefan Andreas Baumann, Jannik Wiese, Tommaso Martorella, Mahdi M. Kalayeh, Bj\"orn Ommer
Abstract: Accurately anticipating how complex, diverse scenes will evolve requires models that represent uncertainty, simulate along extended interaction chains, and efficiently explore many plausible futures. Yet most existing approaches rely on dense video or latent-space prediction, expending substantial capacity on dense appearance rather than on the underlying sparse trajectories of points in the scene. This makes large-scale exploration of future hypotheses costly and limits performance when long-horizon, multi-modal motion is essential. We address this by formulating the prediction of open-set future scene dynamics as step-wise inference over sparse point trajectories. Our autoregressive diffusion model advances these trajectories through short, locally predictable transitions, explicitly modeling the growth of uncertainty over time. This dynamics-centric representation enables fast rollout of thousands of diverse futures from a single image, optionally guided by initial constraints on motion, while maintaining physical plausibility and long-range coherence. We further introduce OWM, a benchmark for open-set motion prediction based on diverse in-the-wild videos, to evaluate accuracy and variability of predicted trajectory distributions under real-world uncertainty. Our method matches or surpasses dense simulators in predictive accuracy while achieving orders-of-magnitude higher sampling speed, making open-set future prediction both scalable and practical. Project page: http://compvis.github.io/myriad.
Authors: Wenyi Xiao, Xinchi Xu, Leilei Gan
Abstract: Large Vision Language Models (LVLMs) achieve strong multimodal reasoning but frequently exhibit hallucinations and incorrect responses with high certainty, which hinders their usage in high-stakes domains. Existing verbalized confidence calibration methods, largely developed for text-only LLMs, typically optimize a single holistic confidence score using binary answer-level correctness. This design is mismatched to LVLMs: an incorrect prediction may arise from perceptual failures or from reasoning errors given correct perception, and a single confidence conflates these sources while visual uncertainty is often dominated by language priors. To address these issues, we propose VL-Calibration, a reinforcement learning framework that explicitly decouples confidence into visual and reasoning confidence. To supervise visual confidence without ground-truth perception labels, we introduce an intrinsic visual certainty estimation that combines (i) visual grounding measured by KL-divergence under image perturbations and (ii) internal certainty measured by token entropy. We further propose token-level advantage reweighting to focus optimization on tokens based on visual certainty, suppressing ungrounded hallucinations while preserving valid perception. Experiments on thirteen benchmarks show that VL-Calibration effectively improves calibration while boosting visual reasoning accuracy, and it generalizes to out-of-distribution benchmarks across model scales and architectures.
Authors: Guanyu Zhou, Yida Yin, Wenhao Chai, Shengbang Tong, Xingyu Fu, Zhuang Liu
Abstract: Vision-language models (VLMs) still struggle with visual perception tasks such as spatial understanding and viewpoint recognition. One plausible contributing factor is that natural image datasets provide limited supervision for low-level visual skills. This motivates a practical question: can targeted synthetic supervision, generated from only a task keyword such as Depth Order, address these weaknesses? To investigate this question, we introduce VisionFoundry, a task-aware synthetic data generation pipeline that takes only the task name as input and uses large language models (LLMs) to generate questions, answers, and text-to-image (T2I) prompts, then synthesizes images with T2I models and verifies consistency with a proprietary VLM, requiring no reference images or human annotation. Using VisionFoundry, we construct VisionFoundry-10K, a synthetic visual question answering (VQA) dataset containing 10k image-question-answer triples spanning 10 tasks. Models trained on VisionFoundry-10K achieve substantial improvements on visual perception benchmarks: +7% on MMVP and +10% on CV-Bench-3D, while preserving broader capabilities and showing favorable scaling behavior as data size increases. Our results suggest that limited task-targeted supervision is an important contributor to this bottleneck and that synthetic supervision is a promising path toward more systematic training for VLMs.
Authors: Zibin Geng, Xuefeng Jiang, Jia Li, Zheng Li, Tian Wen, Lvhua Wu, Sheng Sun, Yuwei Wang, Min Liu
Abstract: Prompt learning is a parameter-efficient approach for vision-language models, yet its robustness under label noise is less investigated. Visual content contains richer and more reliable semantic information, which remains more robust under label noise. However, the prompt itself is highly susceptible to label noise. Motivated by this intuition, we propose VisPrompt, a lightweight and robust vision-guided prompt learning framework for noisy-label settings. Specifically, we exploit a cross-modal attention mechanism to reversely inject visual semantics into prompt representations. This enables the prompt tokens to selectively aggregate visual information relevant to the current sample, thereby improving robustness by anchoring prompt learning to stable instance-level visual evidence and reducing the influence of noisy supervision. To address the instability caused by using the same way of injecting visual information for all samples, despite differences in the quality of their visual cues, we further introduce a lightweight conditional modulation mechanism to adaptively control the strength of visual information injection, which strikes a more robust balance between text-side semantic priors and image-side instance evidence. The proposed framework effectively suppresses the noise-induced disturbances, reduce instability in prompt updates, and alleviate memorization of mislabeled samples. VisPrompt significantly improves robustness while keeping the pretrained VLM backbone frozen and introducing only a small amount of additional trainable parameters. Extensive experiments under synthetic and real-world label noise demonstrate that VisPrompt generally outperforms existing baselines on seven benchmark datasets and achieves stronger robustness. Our code is publicly available at https://github.com/gezbww/Vis_Prompt.
Authors: Soroosh Tayebi Arasteh, Mehdi Joodaki, Mahshad Lotfinia, Sven Nebelung, Daniel Truhn
Abstract: Evidence-grounded reasoning requires more than attaching retrieved text to a prediction: a model should make decisions that depend on whether the provided evidence supports the target claim. In practice, this often fails because supervision is weak, evidence is only loosely tied to the claim, and evaluation does not test evidence dependence directly. We introduce case-grounded evidence verification, a general framework in which a model receives a local case context, external evidence, and a structured claim, and must decide whether the evidence supports the claim for that case. Our key contribution is a supervision construction procedure that generates explicit support examples together with semantically controlled non-support examples, including counterfactual wrong-state and topic-related negatives, without manual evidence annotation. We instantiate the framework in radiology and train a standard verifier on the resulting support task. The learned verifier substantially outperforms both case-only and evidence-only baselines, remains strong under correct evidence, and collapses when evidence is removed or swapped, indicating genuine evidence dependence. This behavior transfers across unseen evidence articles and an external case distribution, though performance degrades under evidence-source shift and remains sensitive to backbone choice. Overall, the results suggest that a major bottleneck in evidence grounding is not only model capacity, but the lack of supervision that encodes the causal role of evidence.
Authors: Hadas Orgad, Boyi Wei, Kaden Zheng, Martin Wattenberg, Peter Henderson, Seraphina Goldfarb-Tarrant, Yonatan Belinkov
Abstract: Large language models (LLMs) undergo alignment training to avoid harmful behaviors, yet the resulting safeguards remain brittle: jailbreaks routinely bypass them, and fine-tuning on narrow domains can induce ``emergent misalignment'' that generalizes broadly. Whether this brittleness reflects a fundamental lack of coherent internal organization for harmfulness remains unclear. Here we use targeted weight pruning as a causal intervention to probe the internal organization of harmfulness in LLMs. We find that harmful content generation depends on a compact set of weights that are general across harm types and distinct from benign capabilities. Aligned models exhibit a greater compression of harm generation weights than unaligned counterparts, indicating that alignment reshapes harmful representations internally--despite the brittleness of safety guardrails at the surface level. This compression explains emergent misalignment: if weights of harmful capabilities are compressed, fine-tuning that engages these weights in one domain can trigger broad misalignment. Consistent with this, pruning harm generation weights in a narrow domain substantially reduces emergent misalignment. Notably, LLMs harmful generation capability is dissociated from how they recognize and explain such content. Together, these results reveal a coherent internal structure for harmfulness in LLMs that may serve as a foundation for more principled approaches to safety.
Authors: Xiaojie Xu, Zongyuan Li, Chang Lu, Runnan Qi, Yanan Ni, Lumin Jiang, Xiangbei Liu, Xuebo Zhang, Yongchun Fang, Kuihua Huang, Xian Guo, Zhanghua Wu, Zhenya Li
Abstract: StarCraft II is a complex and dynamic real-time strategy (RTS) game environment, which is very suitable for artificial intelligence and reinforcement learning research. To address the problem of Large Language Model(LLM) learning in complex environments through self-reflection, we propose a Reflection of Episodes(ROE) framework based on expert experience and self-experience. This framework first obtains key information in the game through a keyframe selection method, then makes decisions based on expert experience and self-experience. After a game is completed, it reflects on the previous experience to obtain new self-experience. Finally, in the experiment, our method beat the robot under the Very Hard difficulty in TextStarCraft II. We analyze the data of the LLM in the process of the game in detail, verified its effectiveness.
Authors: Shahab Rahimirad, Guven Gergerli, Lucia Romero, Angela Qian, Matthew Lyle Olson, Simon Stepputtis, Joseph Campbell
Abstract: Social reasoning - inferring unobservable beliefs and intentions from partial observations of other agents - remains a challenging task for large language models (LLMs). We evaluate the limits of current reasoning language models in the social deduction game Avalon and find that while the largest models demonstrate strong performance, they require extensive test-time inference and degrade sharply when distilled to smaller, real-time-capable variants. To address this, we introduce a hybrid reasoning framework that externalizes belief inference to a structured probabilistic model, while using an LLM for language understanding and interaction. Our approach achieves competitive performance with much larger models in Agent-Agent play and, notably, is the first language agent to defeat human players in a controlled study - achieving a 67% win rate and receiving higher qualitative ratings than both reasoning baselines and human teammates. We release code, models, and a dataset to support future work on social reasoning in LLM agents, which can be found at https://camp-lab-purdue.github.io/bayesian-social-deduction/
URLs: https://camp-lab-purdue.github.io/bayesian-social-deduction/
Authors: Zhirong Chen, Kaiyan Chang, Zhuolin Li, Cangyuan Li, Xinyang He, Chujie Chen, Mengdi Wang, Haobo Xu, Yinhe Han, Huawei Li, Ying Wang
Abstract: Large Language Models have emerged as powerful tools for automating Register-Transfer Level (RTL) code generation, yet they face critical limitations: existing approaches typically fail to simultaneously optimize functional correctness and hardware efficiency metrics such as Power, Performance, and Area (PPA). Methods relying on supervised fine-tuning commonly produce functionally correct but suboptimal designs due to the lack of inherent mechanisms for learning hardware optimization principles. Conversely, external post-processing techniques aiming to refine PPA performance after generation often suffer from inefficiency and do not improve the LLMs' intrinsic capabilities. To overcome these challenges, we propose ChipSeek, a novel hierarchical reward based reinforcement learning framework designed to encourage LLMs to generate RTL code that is both functionally correct and optimized for PPA metrics. Our approach integrates direct feedback from EDA simulators and synthesis tools into a hierarchical reward mechanism, facilitating a nuanced understanding of hardware design trade-offs. Through Curriculum-Guided Dynamic Policy Optimization (CDPO), ChipSeek enhances the LLM's ability to generate high-quality, optimized RTL code. Evaluations on standard benchmarks demonstrate ChipSeek's superior performance, achieving state-of-the-art functional correctness and PPA performance. Furthermore, it excels in specific optimization tasks, consistently yielding highly efficient designs when individually targeting fine-grained optimization goals such as power, delay, and area. The artifact is open-source in https://github.com/rong-hash/chipseek.
Authors: Rui Wang, Qihan Lin, Jiayu Liu, Qing Zong, Tianshi Zheng, Dadi Guo, Haochen Shi, Weiqi Wang, Yangqiu Song
Abstract: Prospect Theory (PT) models human decision-making behaviour under uncertainty, among which linguistic uncertainty is commonly adopted in real-world scenarios. Although recent studies have developed some frameworks to test PT parameters for Large Language Models (LLMs), few have considered the fitness of PT itself on LLMs. Moreover, whether PT is robust under linguistic uncertainty perturbations, especially epistemic markers (e.g. "likely"), remains highly under-explored. To address these gaps, we design a three-stage workflow based on a classic behavioural economics experimental setup. We first estimate PT parameters with economics questions and evaluate PT's fitness with performance metrics. We then derive probability mappings for epistemic markers in the same context, and inject these mappings into the prompt to investigate the stability of PT parameters. Our findings suggest that modelling LLMs' decision-making with PT is not consistently reliable across models, and applying Prospect Theory to LLMs is likely not robust to epistemic uncertainty. The findings caution against the deployment of PT-based frameworks in real-world applications where epistemic ambiguity is prevalent, giving valuable insights in behaviour interpretation and future alignment direction for LLM decision-making.
Authors: Edward Kim, Daniel He, Jorge Chao, Wiktor Rajca, Mohammed Amin, Nishant Malpani, Ruta Desai, Antti Oulasvirta, Bjoern Hartmann, Sanjit Seshia
Abstract: Teaching systems physical tasks is a long standing goal in HCI, yet most prior work has focused on non collaborative physical activities. Collaborative tasks introduce added complexity, requiring systems to infer users assumptions about their teammates intent, which is an inherently ambiguous and dynamic process. This necessitates representations that are interpretable and correctable, enabling users to inspect and refine system behavior. We address this challenge by framing collaborative task learning as a program synthesis problem. Our system represents behavior as editable programs and uses narrated demonstrations, i.e. paired physical actions and natural language, as a unified modality for teaching, inspecting, and correcting system logic without requiring users to see or write code. The same modality is used for the system to communicate its learning to users. In a within subjects study, 20 users taught multiplayer soccer tactics to our system. 70 percent (14/20) of participants successfully refined learned programs to match their intent and 90 percent (18/20) found it easy to correct the programs. The study surfaced unique challenges in representing learning as programs and in enabling users to teach collaborative physical activities. We discuss these issues and outline mitigation strategies.
Authors: Xinzhe Li
Abstract: Test-time scaling improves large language models (LLMs) on long-horizon reasoning tasks by allocating more compute at inference. LLM inference via tree search (LITS) achieves strong performance but is highly inefficient. We propose Chain-in-Tree (CiT), a plug-in framework that decides when to branch during search instead of expanding at every step. CiT introduces lightweight Branching Necessity (BN) evaluations, including BN-DP (direct prompting) and BN-SC (self-consistency). Integrated into Tree of Thoughts, ReST-MCTS, and RAP, BN-DP reduces token generation, model calls, and runtime by 75-85% on GSM8K and Math500, with often negligible or no accuracy loss. BN-SC typically yields substantial savings (up to 80%) generally but shows instability in 1-4 out of 14 settings, caused by a small subset of examples that produce extremely long reasoning steps. We theoretically prove that BN-DP never increases policy invocations and release unified implementations applicable across LITS frameworks. The full codebase is publicly available at https://github.com/xinzhel/chain_in_tree.
Authors: Hyeong Kyu Choi, Xiaojin Zhu, Sharon Li
Abstract: Multi-agent debate (MAD) aims to improve large language model (LLM) reasoning by letting multiple agents exchange answers and then aggregate their opinions. Yet recent studies reveal that agents are not neutral: they are prone to identity-driven sycophancy and self-bias, uncritically adopting a peer's view or stubbornly adhering to their own prior output, undermining the reliability of debate. In this work, we present the first principled framework that joins sycophancy and self-bias to mitigate and quantify identity bias in MAD. First, we formalize the debate dynamics as an identity-weighted Bayesian update process. Second, we propose response anonymization: by removing identity markers from prompts, agents cannot distinguish "self" from "peer", which forces equal weights on agent identity, thereby reducing bias and improving trustworthiness. Third, we define the Identity Bias Coefficient (IBC), a principled bias metric that measures an agent's tendency to follow its peer versus itself. Empirical studies across multiple models and benchmarks confirm that identity bias is widespread, with sycophancy far more common than self-bias. Our findings highlight the need to ensure that MAD systems reason based on content rather than identity. Code is released in https://github.com/deeplearning-wisc/MAD-identity-bias.
URLs: https://github.com/deeplearning-wisc/MAD-identity-bias.
Authors: Xiaohan Zhang, Tian Gao, Mingyue Cheng, Bokai Pan, Ze Guo, Yaguo Liu, Xiaoyu Tao, Qi Liu
Abstract: Time series forecasting plays a crucial role in decision-making across many real-world applications. Despite substantial progress, most existing methods still treat forecasting as a static, single-pass regression problem. In contrast, human experts form predictions through iterative reasoning that integrates temporal features, domain knowledge, case-based references, and supplementary context, with continuous refinement. In this work, we propose Alphacast, an interaction-driven agentic reasoning framework that enables accurate time series forecasting with training-free large language models. Alphacast reformulates forecasting as an expert-like process and organizes it into a multi-stage workflow involving context preparation, reasoning-based generation, and reflective evaluation, transforming forecasting from a single-pass output into a multi-turn, autonomous interaction process. To support diverse perspectives commonly considered by human experts, we develop a lightweight toolkit comprising a feature set, a knowledge base, a case library, and a contextual pool that provides external support for LLM-based reasoning. Extensive experiments across multiple benchmarks show that Alphacast generally outperforms representative baselines. Code is available at this repository: https://github.com/echo01-ai/AlphaCast.
Authors: Jiahuan Long, Tingsong Jiang, Hanqing Liu, Chao Ma, Weien Zhou, Yang Yang, Wen Yao
Abstract: Adversarial patches have emerged as a popular privacy-preserving approach for resisting AI-driven surveillance systems. However, their conspicuous appearance makes them difficult to deploy in real-world scenarios. In this paper, we propose a thermally activated adversarial wearable designed to ensure adaptability and effectiveness in complex real-world environments. The system integrates thermochromic dyes with flexible heating units to induce visually dynamic adversarial patterns on clothing surfaces. In its default state, the clothing appears as an ordinary black T-shirt. Upon heating via an embedded thermal unit, hidden adversarial patterns on the fabric are activated, allowing the wearer to effectively evade detection across both visible and infrared modalities. Physical experiments demonstrate that the adversarial wearable achieves rapid texture activation within 50 seconds and maintains an adversarial success rate above 80\% across diverse real-world surveillance environments. This work demonstrates a new pathway toward physically grounded, user-controllable anti-AI systems, highlighting the growing importance of proactive adversarial techniques for privacy protection in the age of ubiquitous AI surveillance.
Authors: Runze Li, Yuwen Zhai, Bo Xu, LiWu Xu, Nian Shi, Wei Zhang, Ran Lin, Liang Wang
Abstract: Contemporary GUI agents, while increasingly capable due to advances in Large Vision-Language Models (VLMs), often operate with a critical limitation: they treat each task in isolation, lacking a mechanism to systematically learn from past successes. This digital ''amnesia'' results in sub-optimal performance, repeated errors, and poor generalization to novel challenges. To bridge this gap, we introduce EchoTrail-GUI, a novel framework designed to mimic human-like experiential learning by equipping agents with a dynamic, accessible memory. Our framework operates in three distinct stages. First, during Experience Exploration, an agent autonomously interacts with GUI environments to build a curated database of successful task trajectories, validated by a reward model. Crucially, the entire knowledge base construction is thus fully automated, requiring no human supervision. Second, in the Memory Injection stage, upon receiving a new task, our system efficiently retrieves the most relevant past trajectories to serve as actionable ''memories''. Finally, during GUI Task Inference, these memories are injected as in-context guidance to inform the agent's reasoning and decision-making process. We demonstrate the efficacy of our approach on benchmarks including Android World and AndroidLab. The results show that EchoTrail-GUI significantly improves the task success rate and operational efficiency of baseline agents, validating the power of structured memory in creating more robust and intelligent GUI automation.
Authors: Celeste Veronese, Alessandro Farinelli, Daniele Meli
Abstract: Reinforcement Learning (RL) is a well-established framework for sequential decision-making in complex environments. However, state-of-the-art Deep RL (DRL) algorithms typically require large training datasets and often struggle to generalize beyond small-scale training scenarios, even within standard benchmarks. We propose a neuro-symbolic DRL approach that integrates background symbolic knowledge to improve sample efficiency and generalization to more challenging, unseen tasks. Partial policies defined for simple domain instances, where high performance is easily attained, are transferred as useful priors to accelerate learning in more complex settings and avoid tuning DRL parameters from scratch. To do so, partial policies are represented as logical rules, and online reasoning is performed to guide the training process through two mechanisms: (i) biasing the action distribution during exploration, and (ii) rescaling Q-values during exploitation. This neuro-symbolic integration enhances interpretability and trustworthiness while accelerating convergence, particularly in sparse-reward environments and tasks with long planning horizons. We empirically validate our methodology on challenging variants of gridworld environments, both in the fully observable and partially observable setting. We show improved performance over a state-of-the-art reward machine baseline.
Authors: Issa Hanou, Eric Kemmeren, Devin Wild Thomas, Mathijs de Weerdt
Abstract: Executing a multi-agent plan can be challenging when an agent is delayed, because this typically creates conflicts with other agents. So, we need to quickly find a new safe plan. Replanning only the delayed agent often does not yield an efficient plan, and sometimes cannot even yield a feasible one. On the other hand, replanning other agents may lead to a cascade of changes and delays and is computationally expensive. We show how to efficiently replan by tracking and using the temporal flexibility of other agents while avoiding cascading delays. This flexibility is the maximum delay an agent can take without changing the order of other agents or further delaying them. Our algorithm, FlexSIPP, precomputes all possible plans for the delayed agent and returns the changes to the other agents for any single-agent delay within the given scenario. We demonstrate our method in a real-world case study of replanning trains in the densely-used Dutch railway network and in the MovingAI benchmark set. Our experiments show that FlexSIPP provides effective solutions relevant to real-world adjustments, and within a reasonable timeframe.
Authors: William Walden, Miriam Wanner
Abstract: Hint-based faithfulness evaluations have established that Large Reasoning Models (LRMs) may not say what they think: they do not always volunteer information about how key parts of the input (e.g. answer hints) influence their reasoning. Yet, these evaluations also fail to specify what models should do when confronted with hints or other unusual prompt content -- even though versions of such instructions are standard security measures (e.g. for countering prompt injections). Here, we study faithfulness under this more realistic setting in which models are explicitly alerted to the possibility of unusual inputs. We find that such instructions can yield strong results on faithfulness metrics from prior work. However, results on new, more granular metrics proposed in this work paint a mixed picture: although models may acknowledge the presence of hints, they will often deny intending to use them -- even when permitted to use hints and even when it can be demonstrated that they are using them. Our results thus raise broader challenges for CoT monitoring and interpretability.
Authors: Mayank Sharma, Roy Pea, Hari Subramonyam
Abstract: Despite their growing adoption in education, LLMs remain misaligned with the core principle of effective tutoring: the dialogic construction of knowledge. We introduce ConvoLearn, a dataset of 2,134 semi-synthetic tutor-student dialogues operationalizing six dimensions of dialogic tutoring grounded in knowledge-building theory, situated in a middle school Earth Science curriculum. We show that dimension-labeled dialogic training data captures meaningful pedagogical signal that generalizes beyond its semi-synthetic domain: scores from a classifier trained on ConvoLearn correlate significantly with expert-coded instructional quality in authentic classrooms across multiple subscales. As a proof of concept, we fine-tune Mistral-7B on ConvoLearn and show that dimension-level fine-tuning can steer a 7B open-weight model toward dialogic tutoring behavior that credentialed teachers rate as competitive with a strong proprietary baseline. With this work, we support the development of AI tutors capable of more dialogic interactions.
Authors: Alexander H\"agele, Aryo Pradipta Gema, Henry Sleight, Ethan Perez, Jascha Sohl-Dickstein
Abstract: As AI becomes more capable, we entrust it with more general and consequential tasks. The risks from failure grow more severe with increasing task scope. It is therefore important to understand how extremely capable AI models will fail: Will they fail by systematically pursuing goals we do not intend? Or will they fail by being a hot mess, and taking nonsensical actions that do not further any goal? We operationalize this question using a bias-variance decomposition of the errors made by AI models: An AI's \emph{error-incoherence} on a task is measured over test-time randomness as the fraction of its error that stems from variance rather than bias in task outcome. Across all tasks and frontier models we measure, the longer models spend reasoning and taking actions, \emph{the more incoherent} their failures become. Error-incoherence changes with model scale in a way that is experiment dependent. However, in several settings, larger, more capable models are more incoherent than smaller models. Consequently, scale alone seems unlikely to eliminate error-incoherence. Instead, as more capable AIs pursue harder tasks, requiring more sequential action and thought, our results predict failures to be accompanied by more incoherent behavior. This suggests a future where AIs sometimes cause industrial accidents (due to unpredictable misbehavior), but are less likely to exhibit consistent pursuit of a misaligned goal. This increases the relative importance of alignment research targeting reward hacking or goal misspecification.
Authors: Xia Jiang, Jing Chen, Cong Zhang, Jie Gao, Chengpeng Hu, Chenhao Zhang, Yaoxin Wu, Yingqian Zhang
Abstract: While large language models (LLMs) have shown strong performance in math and logic reasoning, their ability to handle combinatorial optimization (CO) -- searching high-dimensional solution spaces under hard constraints -- remains underexplored. To bridge the gap, we introduce NLCO, a \textbf{N}atural \textbf{L}anguage \textbf{C}ombinatorial \textbf{O}ptimization benchmark that evaluates LLMs on end-to-end CO reasoning: given a language-described decision-making scenario, the model must output a discrete solution without writing code or calling external solvers. NLCO covers 43 CO problems and is organized using a four-layer taxonomy of variable types, constraint families, global patterns, and objective classes, enabling fine-grained evaluation. We provide solver-annotated solutions and comprehensively evaluate LLMs by feasibility, solution optimality, and reasoning efficiency. Experiments across a wide range of modern LLMs show that high-performing models achieve strong feasibility and solution quality on small instances, but both degrade as instance size grows, even if more tokens are used for reasoning. We also observe systematic effects across the taxonomy: set-based tasks are relatively easy, whereas graph-structured problems and bottleneck objectives lead to more frequent failures.
Authors: Jun-Min Lee, Meong Hi Son, Edward Choi
Abstract: Hospital administration departments handle a wide range of operational tasks and, in large hospitals, process over 10,000 requests per day, driving growing interest in LLM-based automation. However, prior work has focused primarily on patient-physician interactions or isolated administrative subtasks, failing to capture the complexity of real administrative workflows. To address this gap, we propose H-AdminSim, a comprehensive simulation framework that combines realistic data generation with multi-agent-based simulation of hospital administrative workflows. These tasks are quantitatively evaluated using detailed rubrics, enabling systematic comparison of LLMs. Through FHIR integration, H-AdminSim provides a unified and interoperable environment for testing administrative workflows across heterogeneous hospital settings, serving as a standardized testbed for assessing the feasibility and performance of LLM-driven administrative automation.
Authors: Bang Nguyen, Dominik So\'os, Qian Ma, Rochana R. Obadage, Zack Ranjan, Sai Koneru, Anna Szabelska, Adam Gill, Timothy M. Errington, Shakhlo Nematova, Sarah Rajtmajer, Jian Wu, Meng Jiang
Abstract: The literature has witnessed an emerging interest in AI agents for automated assessment of scientific papers. Existing benchmarks focus primarily on the computational aspect of this task, testing agents' ability to reproduce or replicate research outcomes when having access to the code and data. This setting, while foundational, (1) fails to capture the inconsistent availability of new data for replication as opposed to reproduction, and (2) lacks ground-truth diversity by focusing only on reproducible papers, thereby failing to evaluate an agent's ability to identify non-replicable research. Furthermore, most benchmarks only evaluate outcomes rather than the replication process. In response, we introduce ReplicatorBench, an end-to-end benchmark, including human-verified replicable and non-replicable research claims in social and behavioral sciences for evaluating AI agents in research replication across three stages: (1) extraction and retrieval of replication data; (2) design and execution of computational experiments; and (3) interpretation of results, allowing a test of AI agents' capability to mimic the activities of human replicators in real world. To set a baseline of AI agents' capability, we develop ReplicatorAgent, an agentic framework equipped with necessary tools, like web search and iterative interaction with sandboxed environments, to accomplish tasks in ReplicatorBench. We evaluate ReplicatorAgent across four underlying large language models (LLMs), as well as different design choices of programming language and levels of code access. Our findings reveal that while current LLM agents are capable of effectively designing and executing computational experiments, they struggle with retrieving resources, such as new data, necessary to replicate a claim. All code and data are publicly available at https://github.com/CenterForOpenScience/llm-benchmarking.
URLs: https://github.com/CenterForOpenScience/llm-benchmarking.
Authors: Yuanda Xu, Hejian Sang, Zhengze Zhou, Ran He, Zhipeng Wang
Abstract: Standard LLM distillation treats all training problems equally -- wasting compute on problems the student has already mastered or cannot yet solve. We empirically show that this inefficiency has a precise gradient-level signature: the cross-problem gradient signal-to-noise ratio (SNR) follows a bell curve over student pass rate, collapsing at both extremes. We propose PACED, which weights each problem by $w(p) = p(1{-}p)$ where $p$ is the student's empirical pass rate -- concentrating training on the zone of proximal development. This requires only student rollouts, no architectural changes, and no hyperparameters. We prove the Beta kernel $w(p) = p^\alpha(1{-}p)^\beta$ is the leading-order optimal weight family arising from the SNR boundary-collapse structure, and is minimax-robust under misspecification (worst-case efficiency loss $O(\delta^2)$). Across Qwen3, Qwen2.5, and Llama-3 families, PACED sets a new state of the art in our experimental setting on MATH-500, AIME~2024, and AIME~2025, improving over unweighted distillation by up to $\mathbf{+8.2}$ and over the strong AKL baseline by up to $\mathbf{+3.6}$, while reducing forgetting to $\mathbf{1.4\%}$ and $\mathbf{0.6\%}$ in distillation and self-distillation. A two-stage forward-then-reverse KL schedule pushes gains further to $\mathbf{+5.8}$ over standard forward KL on the hardest benchmark.
Authors: Neelmani Vispute, Aditya Kadam
Abstract: As AI agents transition from human-supervised copilots to autonomous platform infrastructure, the ability to analyze their reasoning behavior across populations of investigations becomes a pressing infrastructure requirement. Existing operational tooling addresses adjacent needs effectively: state checkpoint systems enable fault tolerance; observability platforms provide execution traces for debugging; telemetry standards ensure interoperability. What current systems do not natively provide as a first-class, schema-level primitive is structured reasoning provenance -- normalized, queryable records of why the agent chose each action, what it concluded from each observation, how each conclusion shaped its strategy, and which evidence supports its final verdict. This paper introduces the Agent Execution Record (AER), a structured reasoning provenance primitive that captures intent, observation, and inference as first-class queryable fields on every step, alongside versioned plans with revision rationale, evidence chains, structured verdicts with confidence scores, and delegation authority chains. We formalize the distinction between computational state persistence and reasoning provenance, argue that the latter cannot in general be faithfully reconstructed from the former, and show how AERs enable population-level behavioral analytics: reasoning pattern mining, confidence calibration, cross-agent comparison, and counterfactual regression testing via mock replay. We present a domain-agnostic model with extensible domain profiles, a reference implementation and SDK, and outline an evaluation methodology informed by preliminary deployment on a production platformized root cause analysis agent.
Authors: Zhanting Zhou, KaHou Tam, Ziqiang Zheng, Zeyu Ma, Yang Yang
Abstract: Multimodal recommendation systems (MRS) jointly model user-item interaction graphs and rich item content, but this tight coupling makes user data difficult to remove once learned. Approximate machine unlearning offers an efficient alternative to full retraining, yet existing methods for MRS mainly rely on a largely uniform reverse update across the model. We show that this assumption is fundamentally mismatched to modern MRS: deleted-data influence is not uniformly distributed, but concentrated unevenly across \textit{ranking behavior}, \textit{modality branches}, and \textit{network layers}. This non-uniformity gives rise to three bottlenecks in MRS unlearning: target-item persistence in the collaborative graph, modality imbalance across feature branches, and layer-wise sensitivity in the parameter space. To address this mismatch, we propose \textbf{targeted reverse update} (TRU), a plug-and-play unlearning framework for MRS. Instead of applying a blind global reversal, TRU performs three coordinated interventions across the model hierarchy: a ranking fusion gate to suppress residual target-item influence in ranking, branch-wise modality scaling to preserve retained multimodal representations, and capacity-aware layer isolation to localize reverse updates to deletion-sensitive modules. Experiments across two representative backbones, three datasets, and three unlearning regimes show that TRU consistently achieves a better retain-forget trade-off than prior approximate baselines, while security audits further confirm deeper forgetting and behavior closer to a full retraining on the retained data.
Authors: Zhenfeng Lin, Haoji Hu, Ming Hao, Xuchao Zhang, Ryan Zhang, Junhao Li, Ze Li, Oleg Kulygin, Chetan Bansal, Hatay Tuna, Murali Chintalapati, Sheila Jiang, Salman Zafar, Angie Anderson
Abstract: Outage management in large-scale cloud operations remains heavily manual, requiring rapid triage, cross-team coordination, and experience-driven decisions under partial observability. We present \textbf{ActionNex}, a production-grade agentic system that supports end-to-end outage assistance, including real-time updates, knowledge distillation, and role- and stage-conditioned next-best action recommendations. ActionNex ingests multimodal operational signals (e.g., outage content, telemetry, and human communications) and compresses them into critical events that represent meaningful state transitions. It couples this perception layer with a hierarchical memory subsystem: long-term Key-Condition-Action (KCA) knowledge distilled from playbooks and historical executions, episodic memory of prior outages, and working memory of the live context. A reasoning agent aligns current critical events to preconditions, retrieves relevant memories, and generates actionable recommendations; executed human actions serve as an implicit feedback signal to enable continual self-evolution in a human-agent hybrid system. We evaluate ActionNex on eight real Azure outages (8M tokens, 4,000 critical events) using two complementary ground-truth action sets, achieving 71.4\% precision and 52.8-54.8\% recall. The system has been piloted in production and has received positive early feedback.
Authors: Chao Li, Yuru Wang, Chunyi Zhao
Abstract: We establish a computation-substrate-agnostic inference architecture in which domain is an explicit first-class computational parameter. This produces domain-scoped pruning that reduces per-query search space from O(N) to O(N/K), substrate-independent execution over symbolic, neural, vector, and hybrid substrates, and transparent inference chains where every step carries its evaluative context. The contribution is architectural, not logical. We formalize the computational theory across five dimensions: a five-layer architecture; three domain computation modes including chain indexing, path traversal as Kleisli composition, and vector-guided computation as a substrate transition; a substrate-agnostic interface with three operations Query, Extend, Bridge; reliability conditions C1 to C4 with three failure mode classes; and validation through a PHQ-9 clinical reasoning case study. The computational theory including operational semantics, complexity bounds, monad structure, substrate transitions, and boundary conditions is the contribution of this paper.
Authors: Jingyang Qiao, Weicheng Meng, Yu Cheng, Zhihang Lin, Zhizhong Zhang, Xin Tan, Jingyu Gong, Kun Shao, Yuan Xie
Abstract: Deep research agents (DRAs) integrate LLM reasoning with external tools. Memory systems enable DRAs to leverage historical experiences, which are essential for efficient reasoning and autonomous evolution. Existing methods rely on retrieving similar trajectories from memory to aid reasoning, while suffering from key limitations of ineffective memory evolution and increasing storage and retrieval costs. To address these problems, we propose a novel Memory Intelligence Agent (MIA) framework, consisting of a Manager-Planner-Executor architecture. Memory Manager is a non-parametric memory system that can store compressed historical search trajectories. Planner is a parametric memory agent that can produce search plans for questions. Executor is another agent that can search and analyze information guided by the search plan. To build the MIA framework, we first adopt an alternating reinforcement learning paradigm to enhance cooperation between the Planner and the Executor. Furthermore, we enable the Planner to continuously evolve during test-time learning, with updates performed on-the-fly alongside inference without interrupting the reasoning process. Additionally, we establish a bidirectional conversion loop between parametric and non-parametric memories to achieve efficient memory evolution. Finally, we incorporate a reflection and an unsupervised judgment mechanisms to boost reasoning and self-evolution in the open world. Extensive experiments across eleven benchmarks demonstrate the superiority of MIA.
Authors: Chenjie Yang, Yutian Jiang, Anqi Liang, Wei Qi, Chenyu Wu, Junbo Zhang
Abstract: Human mobility modeling is indispensable for diverse urban applications. However, existing data-driven methods often suffer from data scarcity, limiting their applicability in regions where historical trajectories are unavailable or restricted. To bridge this gap, we propose \textbf{ActivityEditor}, a novel dual-LLM-agent framework designed for zero-shot cross-regional trajectory generation. Our framework decomposes the complex synthesis task into two collaborative stages. Specifically, an intention-based agent, which leverages demographic-driven priors to generate structured human intentions and coarse activity chains to ensure high-level socio-semantic coherence. These outputs are then refined by editor agent to obtain mobility trajectories through iteratively revisions that enforces human mobility law. This capability is acquired through reinforcement learning with multiple rewards grounded in real-world physical constraints, allowing the agent to internalize mobility regularities and ensure high-fidelity trajectory generation. Extensive experiments demonstrate that \textbf{ActivityEditor} achieves superior zero-shot performance when transferred across diverse urban contexts. It maintains high statistical fidelity and physical validity, providing a robust and highly generalizable solution for mobility simulation in data-scarce scenarios. Our code is available at: https://anonymous.4open.science/r/ActivityEditor-066B.
URLs: https://anonymous.4open.science/r/ActivityEditor-066B.
Authors: Wang Yang, Chaoda Song, Xinpeng Li, Debargha Ganguly, Chuang Ma, Shouren Wang, Zhihao Dou, Yuli Zhou, Vipin Chaudhary, Xiaotian Han
Abstract: Existing Agent benchmarks suffer from two critical limitations: high environment interaction overhead (up to 41\% of total evaluation time) and imbalanced task horizon and difficulty distributions that make aggregate scores unreliable. To address these issues, we propose AgentCE-Bench built around a unified grid-based planning task, where agents must fill hidden slots in a partially completed schedule subject to both local slot constraints and global constraints. Our benchmark offers fine-grained control through two orthogonal axes: \textbf{Scalable Horizons}, controlled by the number of hidden slots $H$, and \textbf{Controllable Difficulty}, governed by a decoy budget $B$ that determines the number of globally misleading decoy candidates. Crucially, all tool calls are resolved via static JSON files under a \textbf{Lightweight Environment} design, eliminating setup overhead and enabling fast, reproducible evaluation suitable for training-time validation. We first validate that $H$ and $B$ provide reliable control over task horizon and difficulty, and that AgentCE-Bench exhibits strong domain consistency and model discriminability. We then conduct comprehensive experiments across 13 models of diverse sizes and families over 6 domains, revealing significant cross-model performance variation and confirming that AgentCE-Bench provides interpretable and controllable evaluation of agent reasoning.
Authors: Wenxuan Liu, Zixuan Li, Long Bai, Chunmao Zhang, Fenghui Zhang, Zhuo Chen, Wei Li, Yuxin Zuo, Fei Wang, Bingbing Xu, Xuhui Jiang, Jin Zhang, Xiaolong Jin, Jiafeng Guo, Tat-Seng Chua, Xueqi Cheng
Abstract: Deep Research (DR) requires LLM agents to autonomously perform multi-step information seeking, processing, and reasoning to generate comprehensive reports. In contrast to existing studies that mainly focus on unstructured web content, a more challenging DR task should additionally utilize structured knowledge to provide a solid data foundation, facilitate quantitative computation, and lead to in-depth analyses. In this paper, we refer to this novel task as Knowledgeable Deep Research (KDR), which requires DR agents to generate reports with both structured and unstructured knowledge. Furthermore, we propose the Hybrid Knowledge Analysis framework (HKA), a multi-agent architecture that reasons over both kinds of knowledge and integrates the texts, figures, and tables into coherent multimodal reports. The key design is the Structured Knowledge Analyzer, which utilizes both coding and vision-language models to produce figures, tables, and corresponding insights. To support systematic evaluation, we construct KDR-Bench, which covers 9 domains, includes 41 expert-level questions, and incorporates a large number of structured knowledge resources (e.g., 1,252 tables). We further annotate the main conclusions and key points for each question and propose three categories of evaluation metrics including general-purpose, knowledge-centric, and vision-enhanced ones. Experimental results demonstrate that HKA consistently outperforms most existing DR agents on general-purpose and knowledge-centric metrics, and even surpasses the Gemini DR agent on vision-enhanced metrics, highlighting its effectiveness in deep, structure-aware knowledge analysis. Finally, we hope this work can serve as a new foundation for structured knowledge analysis in DR agents and facilitate future multimodal DR studies.
Authors: Monishwaran Maheswaran, Leon Lakhani, Zhongzhu Zhou, Shijia Yang, Junxiong Wang, Coleman Hooper, Yuezhou Hu, Rishabh Tiwari, Jue Wang, Harman Singh, Qingyang Wu, Yuqing Jian, Ce Zhang, Kurt Keutzer, Tri Dao, Xiaoxia Wu, Ben Athiwaratkun, James Zou, Chenfeng Xu
Abstract: We show that verifier-free evolution is bottlenecked by both diversity and efficiency: without external correction, repeated evolution accelerates collapse toward narrow modes, while the uniform use of a high-cost model wastes compute and quickly becomes economically impractical. We introduce Squeeze Evolve, a unified multi-model orchestration framework for verifier-free evolutionary inference. Our approach is guided by a simple principle: allocate model capability where it has the highest marginal utility. Stronger models are reserved for high-impact stages, while cheaper models handle the other stages at much lower costs. This principle addresses diversity and cost-efficiency jointly while remaining lightweight. Squeeze Evolve naturally supports open-source, closed-source, and mixed-model deployments. Across AIME 2025, HMMT 2025, LiveCodeBench V6, GPQA-Diamond, ARC-AGI-V2, and multimodal vision benchmarks, such as MMMU-Pro and BabyVision, Squeeze Evolve consistently improves the cost-capability frontier over single-model evolution and achieves new state-of-the-art results on several tasks. Empirically, Squeeze Evolve reduces API cost by up to $\sim$3$\times$ and increases fixed-budget serving throughput by up to $\sim$10$\times$. Moreover, on discovery tasks, Squeeze Evolve is the first verifier-free evolutionary method to match, and in some cases exceed, the performance of verifier-based evolutionary methods.
Authors: Boer Zhang, Mingyan Wu, Dongzhuoran Zhou, Yuqicheng Zhu, Wendong Fan, Puzhen Zhang, Zifeng Ding, Guohao Li, Yuan He
Abstract: Deep research requires reasoning over web evidence to answer open-ended questions, and it is a core capability for AI agents. Yet many deep research agents still rely on implicit, unstructured search behavior that causes redundant exploration and brittle evidence aggregation. Motivated by Anthropic's "think" tool paradigm and insights from the information-retrieval literature, we introduce Q+, a set of query and evidence processing tools that make web search more deliberate by guiding query planning, monitoring search progress, and extracting evidence from long web snapshots. We integrate Q+ into the browser sub-agent of Eigent, an open-source, production-ready multi-agent workforce for computer use, yielding EigentSearch-Q+. Across four benchmarks (SimpleQA-Verified, FRAMES, WebWalkerQA, and X-Bench DeepSearch), Q+ improves Eigent's browser agent benchmark-size-weighted average accuracy by 3.0, 3.8, and 0.6 percentage points (pp) for GPT-4.1, GPT-5.1, and Minimax M2.5 model backends, respectively. Case studies further suggest that EigentSearch-Q+ produces more coherent tool-calling trajectories by making search progress and evidence handling explicit.
Authors: Arda Y\"uksel, Gabriel Thiem, Susanne Walter, Patrick Felka, Gabriela Alves Werb, Ivan Habernal
Abstract: Industry classification schemes are integral parts of public and corporate databases as they classify businesses based on economic activity. Due to the size of the company registers, manual annotation is costly, and fine-tuning models with every update in industry classification schemes requires significant data collection. We replicate the manual expert verification by using existing or easily retrievable multimodal resources for industry classification. We present MONETA, the first multimodal industry classification benchmark with text (Website, Wikipedia, Wikidata) and geospatial sources (OpenStreetMap and satellite imagery). Our dataset enlists 1,000 businesses in Europe with 20 economic activity labels according to EU guidelines (NACE). Our training-free baseline reaches 62.10% and 74.10% with open and closed-source Multimodal Large Language Models (MLLM). We observe an increase of up to 22.80% with the combination of multi-turn design, context enrichment, and classification explanations. We will release our dataset and the enhanced guidelines.
Authors: Ajsal Shereef Palattuparambil, Thommen George Karimpanal, Santu Rana
Abstract: Reinforcement Learning (RL) agents often struggle to generalize knowledge to new tasks, even those structurally similar to ones they have mastered. Although recent approaches have attempted to mitigate this issue via zero-shot transfer, they are often constrained by predefined, discrete class systems, limiting their adaptability to novel or compositional task variations. We propose a significantly more generalized approach, replacing discrete latent variables with natural language conditioning via a text-conditioned Variational Autoencoder (VAE). Our core innovation utilizes a Large Language Model (LLM) as a dynamic \textit{semantic operator} at test time. Rather than relying on rigid rules, our agent queries the LLM to semantically remap the description of the current observation to align with the source task. This source-aligned caption conditions the VAE to generate an imagined state compatible with the agent's original training, enabling direct policy reuse. By harnessing the flexible reasoning capabilities of LLMs, our approach achieves zero-shot transfer across a broad spectrum of complex and truly novel analogous tasks, moving beyond the limitations of fixed category mappings. Code and videos are available \href{https://anonymous.4open.science/r/ASPECT-85C3/}{here}.
Authors: Zixuan Hu, Yongxian Wei, Li Shen, Zhenyi Wang, Baoyuan Wu, Chun Yuan, Dacheng Tao
Abstract: Data-Free Meta-Learning (DFML) aims to enable efficient learning of unseen few-shot tasks, by meta-learning from multiple pre-trained models without accessing their original training data. While existing DFML methods typically generate synthetic data from these models to perform meta-learning, a comprehensive analysis of DFML's robustness-particularly its failure modes and vulnerability to potential attacks-remains notably absent. Such an analysis is crucial as algorithms often operate in complex and uncertain real-world environments. This paper fills this significant gap by systematically investigating the robustness of DFML, identifying two critical but previously overlooked vulnerabilities: Task-Distribution Shift (TDS) and Task-Distribution Corruption (TDC). TDS refers to the sequential shifts in the evolving task distribution, leading to the catastrophic forgetting of previously learned meta-knowledge. TDC exposes a security flaw of DFML, revealing its susceptibility to attacks when the pre-trained model pool includes untrustworthy models that deceptively claim to be beneficial but are actually harmful. To mitigate these vulnerabilities, we propose a trustworthy DFML framework comprising three components: synthetic task reconstruction, meta-learning with task memory interpolation, and automatic model selection. Specifically, utilizing model inversion techniques, we reconstruct synthetic tasks from multiple pre-trained models to perform meta-learning. To prevent forgetting, we introduce a strategy to replay interpolated historical tasks to efficiently recall previous meta-knowledge. Furthermore, our framework seamlessly incorporates an automatic model selection mechanism to automatically filter out untrustworthy models during the meta-learning process. Code is available at https://github.com/Egg-Hu/Trustworthy-DFML.
Authors: Jung-Hoon Cho, Sirui Li, Jeongyun Kim, Cathy Wu
Abstract: The recent development of connected and automated vehicle (CAV) technologies has spurred investigations to optimize dense urban traffic to maximize vehicle speed and throughput. This paper explores advisory autonomy, in which real-time driving advisories are issued to the human drivers, thus achieving near-term performance of automated vehicles. Due to the complexity of traffic systems, recent studies of coordinating CAVs have resorted to leveraging deep reinforcement learning (RL). Coarse-grained advisory is formalized as zero-order holds, and we consider a range of hold duration from 0.1 to 40 seconds. However, despite the similarity of the higher frequency tasks on CAVs, a direct application of deep RL fails to be generalized to advisory autonomy tasks. To overcome this, we utilize zero-shot transfer, training policies on a set of source tasks--specific traffic scenarios with designated hold durations--and then evaluating the efficacy of these policies on different target tasks. We introduce Temporal Transfer Learning (TTL) algorithms to select source tasks for zero-shot transfer, systematically leveraging the temporal structure to solve the full range of tasks. TTL selects the most suitable source tasks to maximize the performance of the range of tasks. We validate our algorithms on diverse mixed-traffic scenarios, demonstrating that TTL more reliably solves the tasks than baselines. This paper underscores the potential of coarse-grained advisory autonomy with TTL in traffic flow optimization.
Authors: Wei Duan, Jie Lu, Junyu Xuan
Abstract: Cooperative Multi-Agent Reinforcement Learning (MARL) necessitates seamless collaboration among agents, often represented by an underlying relation graph. Existing methods for learning this graph primarily focus on agent-pair relations, neglecting higher-order relationships. While several approaches attempt to extend cooperation modelling to encompass behaviour similarities within groups, they commonly fall short in concurrently learning the latent graph, thereby constraining the information exchange among partially observed agents. To overcome these limitations, we present a novel approach to infer the Group-Aware Coordination Graph (GACG), which is designed to capture both the cooperation between agent pairs based on current observations and group-level dependencies from behaviour patterns observed across trajectories. This graph is further used in graph convolution for information exchange between agents during decision-making. To further ensure behavioural consistency among agents within the same group, we introduce a group distance loss, which promotes group cohesion and encourages specialization between groups. Our evaluations, conducted on StarCraft II micromanagement tasks, demonstrate GACG's superior performance. An ablation study further provides experimental evidence of the effectiveness of each component of our method.
Authors: Lorenzo Mannocci, Michele Mazza, Anna Monreale, Maurizio Tesconi, Stefano Cresci
Abstract: Coordination is a fundamental aspect of life. The advent of social media has made it integral also to online human interactions, such as those that characterize thriving online communities and social movements. At the same time, coordination is also core to effective disinformation, manipulation, and hate campaigns. This survey collects, categorizes, and critically discusses the body of work produced as a result of the growing interest on coordinated online behavior. We reconcile industry and academic definitions, propose a comprehensive framework to study coordinated online behavior, and review and critically discuss the existing detection and characterization methods. Our analysis identifies open challenges and promising directions of research, serving as a guide for scholars, practitioners, and policymakers in understanding and addressing the complexities inherent to online coordination.
Authors: Sehun Kim
Abstract: Electrocardiogram (ECG) captures the heart's electrical signals, offering valuable information for diagnosing cardiac conditions. However, the scarcity of labeled data makes it challenging to fully leverage supervised learning in the medical domain. Self-supervised learning (SSL) offers a promising solution, enabling models to learn from unlabeled data and uncover meaningful patterns. In this paper, we show that masked modeling in the latent space can be a powerful alternative to existing self-supervised methods in the ECG domain. We introduce ECG-JEPA, an SSL model for 12-lead ECG analysis that learns semantic representations of ECG data by predicting in the hidden latent space, bypassing the need to reconstruct raw signals. This approach offers several advantages in the ECG domain: (1) it avoids producing unnecessary details, such as noise, which is common in ECG; and (2) it addresses the limitations of naive L2 loss between raw signals. Another key contribution is the introduction of Cross-Pattern Attention (CroPA), a specialized masked attention mechanism tailored for 12-lead ECG data. ECG-JEPA is trained on the union of several open ECG datasets, totaling approximately 180,000 samples, and achieves state-of-the-art performance in various downstream tasks including diagnostic classification, feature extraction, and segmentation. Our code is openly available at https://github.com/sehunfromdaegu/ECG_JEPA.
Authors: Sajib Kumar Saha Joy, Arman Hassan Mahy, Meherin Sultana, Azizah Mamun Abha, MD Piyal Ahmmed, Yue Dong, G M Shahariar
Abstract: In this study, we investigate extrinsic gender bias in Bangla pretrained language models, a largely underexplored area in low-resource languages. To assess this bias, we construct four manually annotated, task-specific benchmark datasets for sentiment analysis, toxicity detection, hate speech detection, and sarcasm detection. Each dataset is augmented using nuanced gender perturbations, where we systematically swap gendered names and terms while preserving semantic content, enabling minimal-pair evaluation of gender-driven prediction shifts. We then propose RandSymKL, a randomized debiasing strategy integrated with symmetric KL divergence and cross-entropy loss to mitigate the bias across task-specific pretrained models. RandSymKL is a refined training approach to integrate these elements in a unified way for extrinsic gender bias mitigation focused on classification tasks. Our approach was evaluated against existing bias mitigation methods, with results showing that our technique not only effectively reduces bias but also maintains competitive accuracy compared to other baseline approaches. To promote further research, we have made both our implementation and datasets publicly available: https://github.com/sajib-kumar/Mitigating-Bangla-Extrinsic-Gender-Bias
URLs: https://github.com/sajib-kumar/Mitigating-Bangla-Extrinsic-Gender-Bias
Authors: Yangyang Li, Daqing Liu, Wu Liu, Allen He, Xinchen Liu, Yongdong Zhang, Guoqing Jin
Abstract: Creative visual concept generation often draws inspiration from specific concepts in a reference image to produce relevant outcomes. However, existing methods are typically constrained to single-aspect concept generation or are easily disrupted by irrelevant concepts in multi-aspect concept scenarios, leading to concept confusion and hindering creative generation. To address this, we propose OmniPrism, a visual concept disentangling approach for creative image generation. Our method learns disentangled concept representations guided by natural language and trains a diffusion model to incorporate these concepts. We utilize the rich semantic space of a multimodal extractor to achieve concept disentanglement from given images and concept guidance. To disentangle concepts with different semantics, we construct a paired concept disentangled dataset (PCD-200K), where each pair shares the same concept such as content, style, and composition. We learn disentangled concept representations through our contrastive orthogonal disentangled (COD) training pipeline, which are then injected into additional diffusion cross-attention layers for generation. A set of block embeddings is designed to adapt each block's concept domain in the diffusion models. Extensive experiments demonstrate that our method can generate high-quality, concept-disentangled results with high fidelity to text prompts and desired concepts.
Authors: Muhammad Umair Haider, Hammad Rizwan, Hassan Sajjad, Peizhong Ju, A. B. Siddique
Abstract: Pervasive polysemanticity in large language models (LLMs) undermines discrete neuron-concept attribution, posing a significant challenge for model interpretation and control. We systematically analyze both encoder and decoder based LLMs across diverse datasets, and observe that even highly salient neurons for specific semantic concepts consistently exhibit polysemantic behavior. Importantly, we uncover a consistent pattern: concept-conditioned activation magnitudes of neurons form distinct, often Gaussian-like distributions with minimal overlap. Building on this observation, we hypothesize that interpreting and intervening on concept-specific activation ranges can enable more precise interpretability and targeted manipulation in LLMs. To this end, we introduce NeuronLens, a novel range-based interpretation and manipulation framework, that localizes concept attribution to activation ranges within a neuron. Extensive empirical evaluations show that range-based interventions enable effective manipulation of target concepts while causing substantially less collateral degradation to auxiliary concepts and overall model performance compared to neuron-level masking.
Authors: Jinghua Piao, Yuwei Yan, Jun Zhang, Nian Li, Junbo Yan, Xiaochong Lan, Zhihong Lu, Zhiheng Zheng, Jing Yi Wang, Di Zhou, Chen Gao, Fengli Xu, Fang Zhang, Ke Rong, Jun Su, Yong Li
Abstract: Understanding human behavior and society is a central focus in social sciences, with the rise of generative social science marking a significant paradigmatic shift. By leveraging bottom-up simulations, it replaces costly and logistically challenging traditional experiments with scalable, replicable, and systematic computational approaches for studying complex social dynamics. Recent advances in large language models (LLMs) have further transformed this research paradigm, enabling the creation of human-like generative social agents and realistic simulacra of society. In this paper, we propose AgentSociety, a large-scale social simulator that integrates LLM-driven agents, a realistic societal environment, and a powerful large-scale simulation engine. Based on the proposed simulator, we generate social lives for over 10k agents, simulating their 5 million interactions both among agents and between agents and their environment. Furthermore, we explore the potential of AgentSociety as a testbed for computational social experiments, focusing on five key social issues: polarization, the spread of inflammatory messages, the effects of universal basic income policies, the impact of external shocks such as hurricanes, and urban sustainability. These five issues serve as valuable cases for assessing AgentSociety's support for typical research methods -- such as surveys, interviews, and interventions -- as well as for investigating the patterns, causes, and underlying mechanisms of social issues. The alignment between AgentSociety's outcomes and real-world experimental results not only demonstrates its ability to capture human behaviors and their underlying mechanisms, but also underscores its potential as an important platform for social scientists and policymakers.
Authors: Hao-Xiang Xu, Jun-Yu Ma, Zhen-Hua Ling, Ningyu Zhang, Jia-Chen Gu
Abstract: Large language models (LLMs) struggle with hallucinations due to false or outdated knowledge. Given the high resource demands of retraining these models, there is an increasing focus on developing model editing. However, the general abilities of LLMs across downstream tasks are prone to significant degradation during sequential editing. This paper statistically observes that the parameter matrix after editing exhibits a significant deviation compared to its previous state as the number of edits increases. This serious deviation affects the original knowledge associations within LLMs and leads to the degradation of their general abilities. To this end, a framework termed Editing Anchor Compression (EAC) is proposed to constrain the deviation of the parameter matrix during sequential editing. It compresses the editing information by selecting editing anchors that are important in encoding new relations without deviating too much from the original matrix, thereby preserving the general abilities. Experiments of applying EAC to two popular editing methods on three LLMs across four tasks are conducted. Evaluation results show that EAC effectively minimizes unreasonable deviations caused by model editing, preserving over 70% of the general abilities while better retaining the editing knowledge compared to the original counterpart methods.
Authors: Junhao Liu, Haonan Yu, Zhenyu Yan, Xin Zhang
Abstract: Post-hoc explanations provide transparency and are essential for guiding model optimization, such as prompt engineering and data sanitation. However, applying model-agnostic techniques to Large Language Models (LLMs) is hindered by prohibitive computational costs, rendering these tools dormant for real-world applications. To revitalize model-agnostic interpretability, we propose a budget-friendly proxy framework that leverages efficient models to approximate the decision boundaries of expensive LLMs. We introduce a screen-and-apply mechanism to statistically verify local alignment before deployment. Our empirical evaluation confirms that proxy explanations achieve over 90% fidelity with only 11% of the oracle's cost. Building on this foundation, we demonstrate the actionable utility of our framework in prompt compression and poisoned example removal. Results show that reliable proxy explanations effectively guide optimization, transforming interpretability from a passive observation tool into a scalable primitive for LLM development. Additionally, we open-source code and datasets to facilitate future research.
Authors: Bryan Sangwoo Kim, Jeongsol Kim, Jong Chul Ye
Abstract: Modern single-image super-resolution (SISR) models deliver photo-realistic results at the scale factors on which they are trained, but collapse when asked to magnify far beyond that regime. We address this scalability bottleneck with Chain-of-Zoom (CoZ), a model-agnostic framework that factorizes SISR into an autoregressive chain of intermediate scale-states with multi-scale-aware prompts. CoZ repeatedly re-uses a backbone SR model, decomposing the conditional probability into tractable sub-problems to achieve extreme resolutions without additional training. Because visual cues diminish at high magnifications, we augment each zoom step with multi-scale-aware text prompts generated by a vision-language model (VLM). The prompt extractor itself is fine-tuned using Generalized Reward Policy Optimization (GRPO) with a critic VLM, aligning text guidance towards human preference. Experiments show that a standard 4x diffusion SR model wrapped in CoZ attains beyond 256x enlargement with high perceptual quality and fidelity. Project Page: https://bryanswkim.github.io/chain-of-zoom/.
Authors: Jing-En Huang, I-Sheng Fang, Tzuhsuan Huang, Yu-Lun Liu, Chih-Yu Wang, Jun-Cheng Chen
Abstract: The data scarcity, label noise, and long-tailed category imbalance remain important and unresolved challenges in many computer vision tasks, such as object detection and instance segmentation, especially on large-vocabulary benchmarks like LVIS, where most categories appear in only a few images. Current synthetic data generation methods still suffer from multiple objects per mask, inaccurate segmentation, incorrect category labels, and other issues, limiting their effectiveness. To address these issues, we introduce Gen-n-Val, a novel agentic data generation framework that leverages Layer Diffusion (LD), a Large Language Model (LLM), and a Vision Large Language Model (VLLM) to produce high-quality and diverse instance masks and images for object detection and instance segmentation. Gen-n-Val consists of two agents: (1) the LD prompt agent, an LLM, optimizes rompts to encourage LD to generate high-quality foreground single-object images and corresponding segmentation masks; and (2) the data validation agent, a VLLM, filters out low-quality synthetic instance images. The system prompts for both agents are optimized by TextGrad. Compared to state-of-the-art synthetic data approaches like MosaicFusion, our approach reduces invalid synthetic data from 50% to 7% and improves performance by 7.6% on rare classes in LVIS instance segmentation with Mask R-CNN, and by 3.6% mAP on rare classes in COCO instance segmentation with YOLOv9c and YOLO11m. Furthermore, Gen-n-Val shows significant improvements (7.1% mAP) over YOLO-Worldv2-M in open-vocabulary object detection benchmarks with YOLO11m. Moreover, Gen-n-Val has scalability in model capacity and dataset size. The code is available at https://github.com/aiiu-lab/Gen-n-Val.
Authors: Zhiyu Xue, Reza Abbasi-Asl, Ramtin Pedarsani
Abstract: Generative medical vision-language models~(Med-VLMs) are primarily designed to generate complex textual information~(e.g., diagnostic reports) from multimodal inputs including vision modality~(e.g., medical images) and language modality~(e.g., clinical queries). However, their security vulnerabilities remain underexplored. Med-VLMs should be capable of rejecting harmful queries, such as \textit{Provide detailed instructions for using this CT scan for insurance fraud}. At the same time, addressing security concerns introduces the risk of over-defense, where safety-enhancing mechanisms may degrade general performance, causing Med-VLMs to reject benign clinical queries. In this paper, we propose a novel inference-time defense strategy to mitigate harmful queries, enabling defense against visual and textual jailbreak attacks. Using diverse medical imaging datasets collected from nine modalities, we demonstrate that our defense strategy based on synthetic clinical demonstrations enhances model safety without significantly compromising performance. Additionally, we find that increasing the demonstration budget alleviates the over-defense issue. We then introduce a mixed demonstration strategy as a trade-off solution for balancing security and performance under few-shot demonstration budget constraints.
Authors: Alexander Gambashidze, Li Pengyi, Matvey Skripkin, Andrey Galichin, Anton Gusarov, Konstantin Sobolev, Andrey Kuznetsov, Ivan Oseledets
Abstract: Training robust and generalizable reward models for human visual preferences is essential for aligning text-to-image and text-to-video generative models with human intent. However, current reward models often fail to generalize, and supervised fine-tuning leads to memorization, demanding complex annotation pipelines. While reinforcement learning (RL), specifically Group Relative Policy Optimization (GRPO), improves generalization, we uncover a key failure mode: a significant drop in reasoning accuracy occurs when a model's reasoning trace contradicts that of an independent, frozen vision-language model ("listener") evaluating the same output. To address this, we introduce a listener-augmented GRPO framework. Here, the listener re-evaluates the reasoner's chain-of-thought to provide a dense, calibrated confidence score, shaping the RL reward signal. This encourages the reasoner not only to answer correctly, but to produce explanations that are persuasive to an independent model. Our listener-shaped reward scheme achieves best accuracy on the ImageReward benchmark (67.4%), significantly improves out-of-distribution (OOD) performance on a large-scale human preference dataset (1.2M votes, up to +6% over naive reasoner), and reduces reasoning contradictions compared to strong GRPO and SFT baselines. These results demonstrate that listener-based rewards provide a scalable, data-efficient path to aligning vision-language models with nuanced human preferences. We will release our reasoning model here: https://huggingface.co/alexgambashidze/qwen2.5vl_image_preference_reasoner.
URLs: https://huggingface.co/alexgambashidze/qwen2.5vl_image_preference_reasoner.
Authors: Haoyu Zhang, Shihao Zhang, Ian Colbert, Rayan Saab
Abstract: Post-training quantization (PTQ) has become a crucial tool for reducing the memory and compute costs of modern deep neural networks, including large language models (LLMs). Among PTQ algorithms, the OPTQ framework-also known as GPTQ-has emerged as a leading method due to its computational efficiency and strong empirical performance. Despite its widespread adoption, however, OPTQ lacks rigorous quantitative theoretical guarantees. This paper presents the first quantitative error bounds for both deterministic and stochastic variants of OPTQ, as well as for Qronos, a recent related state-of-the-art PTQ algorithm. We analyze how OPTQ's iterative procedure induces quantization error and derive non-asymptotic 2-norm error bounds that depend explicitly on the calibration data and a regularization parameter that OPTQ uses. Our analysis provides theoretical justification for several practical design choices, including the widely used heuristic of ordering features by decreasing norm, as well as guidance for selecting the regularization parameter. For the stochastic variant, we establish stronger infinity-norm error bounds, which enable control over the required quantization alphabet and are particularly useful for downstream layers and nonlinearities. Finally, we extend our analysis to Qronos, providing new theoretical bounds, for both its deterministic and stochastic variants, that help explain its empirical advantages.
Authors: Jianxiang He, Meisheng Hong, Jungang Li, Weiyu Guo, Xuming Hu, Hui Xiong
Abstract: Multimodal large language models (MLLMs) demonstrate exceptional performance in vision-language tasks, yet their processing of long videos is constrained by input context length and high computational costs. Sparse frame sampling thus becomes a necessary preprocessing step, with sampled frame quality directly impacting downstream performance. Existing keyframe search algorithms achieve a balance between efficiency and sampled frame quality but heavily rely on the visual modality alone. This makes them difficult to adapt to text-related tasks and often leads to retrieval results deviating from core semantic content. To address this, we propose the VISUAL-SUBTITLE INTEGRATION (VSI), a multimodal keyframe retrieval framework. It employs a dual-branch collaborative retrieval approach combining Video Search and Subtitle Match to fuse complementary visual and textual information for precise localization. Experiments on LongVideoBench and VideoMME demonstrate that VSI achieves state-of-the-art accuracy in keyframe retrieval while delivering breakthrough performance in text-related tasks and exhibiting strong generalization across other tasks.
Authors: Artzai Picon, Itziar Eguskiza, Daniel Mugica, Javier Romero, Carlos Javier Jimenez, Eric White, Gabriel Do-Lago-Junqueira, Christian Klukas, Ramon Navarra-Mestre
Abstract: Reliable plant species and damage segmentation for herbicide field research trials requires models that can withstand substantial real-world variation across seasons, geographies, devices, and sensing modalities. Most deep learning approaches trained on controlled datasets fail to generalize under these domain shifts, limiting their suitability for operational phenotyping pipelines. This study evaluates a segmentation framework that integrates vision foundation models (DINOv2) with hierarchical taxonomic inference to improve robustness across heterogeneous agricultural conditions. We train on a large, multi-year dataset collected in Germany and Spain (2018-2020), comprising 14 plant species and 4 herbicide damage classes, and assess generalization under increasingly challenging shifts: temporal and device changes (2023), geographic transfer to the United States, and extreme sensor shift to drone imagery (2024). Results show that the foundation-model backbone consistently outperforms prior baselines, improving species-level F1 from 0.52 to 0.87 on in-distribution data and maintaining significant advantages under moderate (0.77 vs. 0.24) and extreme (0.44 vs. 0.14) shift conditions. Hierarchical inference provides an additional layer of robustness, enabling meaningful predictions even when fine-grained species classification degrades (family F1: 0.68, class F1: 0.88 on aerial imagery). Error analysis reveals that failures under severe shift stem primarily from vegetation-soil confusion, suggesting that taxonomic distinctions remain preserved despite background and viewpoint variability. The system is now deployed within BASF's phenotyping workflow for herbicide research trials across multiple regions, illustrating the practical viability of combining foundation models with structured biological hierarchies for scalable, shift-resilient agricultural monitoring.
Authors: Sebastian Lubos, Alexander Felfernig, Damian Garber, Gerhard Leitner, Julian Schwazer, Manuel Henrich
Abstract: Usability evaluation is an essential method to support the design of effective and intuitive user interfaces (UIs). However, it commonly relies on resource-intensive, expert-driven methods, which limit its accessibility, especially for small organizations. Recent multimodal large language models (MLLMs) have the potential to support usability evaluation by analyzing textual instructions together with visual UI context. This paper investigates the use of MLLMs as assistive tools for usability evaluation by framing the task as a prioritization problem. It identifies and explains usability issues and ranks them by severity. We report a study that compares the evaluations generated by multiple MLLMs with assessments from usability experts. The results demonstrate that MLLMs can offer complementary insights and support the efficient prioritization of critical issues. Additionally, we present an interactive visualization tool that enables the transparent review and validation of model-generated findings. Based on this, we outline concepts for integrating MLLM-based usability evaluation into real-world development workflows.
Authors: Chen Zeng, Tiehang Xu, Qiao Wang
Abstract: Traditional neural networks struggle to capture the spectral structure of complex signals. Fourier neural networks (FNNs) attempt to address this by embedding Fourier series components, yet many real-world signals are almost-periodic with non-commensurate frequencies, posing additional challenges. Building on prior work showing that ARIMA outperforms large language models (LLMs) for time series forecasting, we extend the comparison to neural predictors and find that ARIMA still maintains a clear advantage. Inspired by this finding, we propose the Autoregressive-Weight-Enhanced Kolmogorov-Arnold Network (AR-KAN). Based in the Universal Myopic Mapping Theorem, it integrates a pre-trained AR module for temporal memory with a KAN for nonlinear representation. We prove that the AR module preserves essential temporal features while reducing redundancy, and that the upper bound of the approximation error for AR-KAN is smaller than that for KAN in a probabilistic sense. Experimental results also demonstrate that AR-KAN delivers exceptional performance compared to existing models, both on synthetic almost-periodic functions and real-world datasets. These results highlight AR-KAN as a robust and effective framework for time series forecasting. Our code is available at https://github.com/ChenZeng001/AR-KAN.
Authors: Hao Chen, Tao Han, Jie Zhang, Song Guo, Lei Bai
Abstract: To gain finer regional forecasts, many works have explored the regional integration from the global atmosphere, e.g., by solving boundary equations in physics-based methods or cropping regions from global forecasts in data-driven methods. However, the effectiveness of these methods is often constrained by static and imprecise regional boundaries, resulting in poor generalization ability. To address this issue, we propose Spatial-Temporal Weather Forecasting (STCast), a novel AI-driven framework for adaptive regional boundary optimization and dynamic monthly forecast allocation. Specifically, our approach employs a Spatial-Aligned Attention (SAA) mechanism, which aligns global and regional spatial distributions to initialize boundaries and adaptively refines them based on attention-derived alignment patterns. Furthermore, we design a Temporal Mixture-of-Experts (TMoE) module, where atmospheric variables from distinct months are dynamically routed to specialized experts using a discrete Gaussian distribution, enhancing the model's ability to capture temporal patterns. Beyond global and regional forecasting, we evaluate our STCast on extreme event prediction and ensemble forecasting. Experimental results demonstrate consistent superiority over state-of-the-art methods across all four tasks. Code: https://github.com/chenhao-zju/STCast
Authors: Rongguang Ye, Ming Tang, Edith C. H. Ngai
Abstract: As increasingly large pre-trained models are released, deploying them on edge devices for privacy-preserving applications requires effective compression. Recent works combine quantization with the fine-tuning of high-precision LoRA adapters, which can substantially reduce model size while mitigating the accuracy loss from quantization. However, edge devices have inherently heterogeneous capabilities, while performing configuration-wise fine-tuning for every quantization setting is computationally prohibitive. In this paper, we propose CoA-LoRA, a method that dynamically adjusts the LoRA adapter to arbitrary quantization configurations (i.e., the per-layer bit-width choices of a pre-trained model) without requiring repeated fine-tuning. This is accomplished via a configuration-aware model that maps each configuration to its low-rank adjustments. The effectiveness of this model critically depends on the training configuration set, a collection of configurations chosen to cover different total bit-width budgets. However, constructing a high-quality configuration set is non-trivial. We therefore design a Pareto-based configuration search that iteratively optimizes the training configuration set, yielding more precise low-rank adjustments. Our experiments demonstrate that, unlike the state-of-the-art methods that require fine-tuning a separate LoRA adapter for each configuration, CoA-LoRA incurs no additional time cost while achieving comparable or even superior performance to those methods.
Authors: Sangwon Ryu, Heejin Do, Yunsu Kim, Gary Geunbae Lee, Jungseul Ok
Abstract: Controllable summarization moves beyond generic outputs toward human-aligned summaries guided by specified attributes. In practice, the interdependence among attributes makes it challenging for language models to satisfy correlated constraints consistently. Moreover, previous approaches often require per-attribute fine-tuning, limiting flexibility across diverse summary attributes. In this paper, we propose adaptive planning for multi-attribute controllable summarization (PACO), a training-free framework that reframes the task as planning the order of sequential attribute control with a customized Monte Carlo Tree Search (MCTS). In PACO, nodes represent summaries, and actions correspond to single-attribute adjustments, enabling progressive refinement of only the attributes requiring further control. This strategy adaptively discovers optimal control orders, ultimately producing summaries that effectively meet all constraints. Extensive experiments across diverse domains and models demonstrate that PACO achieves robust multi-attribute controllability, surpassing both LLM-based self-planning models and fine-tuned baselines. Remarkably, PACO with Llama-3.2-1B rivals the controllability of the much larger Llama-3.3-70B baselines. With larger models, PACO achieves superior control performance, outperforming all competitors.
Authors: Han Zhou, Jinjin Cao, Liyuan Ma, Xueji Fang, Guo-jun Qi
Abstract: Learning diverse manipulation skills for real-world robots is severely bottlenecked by the reliance on costly and hard-to-scale teleoperated demonstrations. While human videos offer a scalable alternative, effectively transferring manipulation knowledge is fundamentally hindered by the significant morphological gap between human and robotic embodiments. To address this challenge and facilitate skill transfer from human to robot, we introduce Traj2Action, a novel framework that bridges this embodiment gap by using the 3D trajectory of the operational endpoint as a unified intermediate representation, and then transfers the manipulation knowledge embedded in this trajectory to the robot's actions. Our policy first learns to generate a coarse trajectory, which forms a high-level motion plan by leveraging both human and robot data. This plan then conditions the synthesis of precise, robot-specific actions (e.g., orientation and gripper state) within a co-denoising framework. Our work centers on two core objectives: first, the systematic verification of the Traj2Action framework's effectiveness-spanning architectural design, cross-task generalization, and data efficiency and second, the revelation of key laws that govern robot policy learning during the integration of human hand demonstration data. This research focus enables us to provide a scalable paradigm tailored to address human-to-robot skill transfer across morphological gaps. Extensive real-world experiments on a Franka robot demonstrate that Traj2Action boosts the performance by up to 27% and 22.25% over $\pi_0$ baseline on short- and long-horizon real-world tasks, and achieves significant gains as human data scales in robot policy learning.
Authors: Danial Samadi Vahdati, Tai Duc Nguyen, Ekta Prashnani, Koki Nagano, David Luebke, Orazio Gallo, Matthew Stamm
Abstract: AI-based talking-head videoconferencing systems reduce bandwidth by sending a compact pose-expression latent and re-synthesizing RGB at the receiver, but this latent can be puppeteered, letting an attacker hijack a victim's likeness in real time. Because every frame is synthetic, deepfake and synthetic video detectors fail outright. To address this security problem, we exploit a key observation: the pose-expression latent inherently contains biometric information of the driving identity. Therefore, we introduce the first biometric leakage defense without ever looking at the reconstructed RGB video: a pose-conditioned, large-margin contrastive encoder that isolates persistent identity cues inside the transmitted latent while cancelling transient pose and expression. A simple cosine test on this disentangled embedding flags illicit identity swaps as the video is rendered. Our experiments on multiple talking-head generation models show that our method consistently outperforms existing puppeteering defenses, operates in real-time, and shows strong generalization to out-of-distribution scenarios.
Authors: Zhepeng Cen, Haolin Chen, Shiyu Wang, Zuxin Liu, Zhiwei Liu, Jielin Qiu, Ding Zhao, Silvio Savarese, Caiming Xiong, Huan Wang, Weiran Yao
Abstract: Large Language Models (LLMs) have achieved remarkable success through imitation learning on vast text corpora, but this paradigm creates a training-generation gap and limits robust reasoning. Reinforcement learning (RL) offers a more data-efficient solution capable of bridging this gap, yet its application has been constrained by a critical data bottleneck: existing RL datasets are orders of magnitude smaller and less diverse than web-scale pre-training corpora. To address this, we introduce the Webscale-RL pipeline, a scalable data engine that systematically converts large-scale pre-training documents into millions of diverse, verifiable question-answer pairs for RL. Using this pipeline, we construct the Webscale-RL dataset, containing 1.2 million examples across more than 9 domains. Our experiments show that the model trained on this dataset significantly outperforms continual pretraining and strong data refinement baselines across a suite of benchmarks. Notably, RL training with our dataset proves substantially more efficient, achieving the performance of continual pre-training with up to 100$\times$ fewer tokens. Our work presents a viable path toward scaling RL to pre-training levels, enabling more capable and efficient language models.
Authors: Shaokai Wu, Yanbiao Ji, Qiuchang Li, Zhiyi Zhang, Qichen He, Wenyuan Xie, Guodong Zhang, Bayram Bayramli, Yue Ding, Hongtao Lu
Abstract: Embodied agents face a fundamental limitation: once deployed in real-world environments, they cannot easily acquire new knowledge to improve task performance. In this paper, we propose Dejavu, a general post-deployment learning framework that augments a frozen Vision-Language-Action (VLA) policy with retrieved execution memories through an Experience Feedback Network (EFN). EFN identifies contextually relevant prior action experiences and conditions action prediction on the retrieved guidance. We train EFN with reinforcement learning and semantic similarity rewards, encouraging the predicted actions to align with past behaviors under the current observation. During deployment, EFN continually expands its memory with new trajectories, enabling the agent to exhibit ``learning from experience.'' Experiments across diverse embodied tasks show that EFN improves adaptability, robustness, and success rates over frozen baselines. Our Project Page is https://dejavu2025.github.io/.
Authors: Yuquan Xue, Guanxing Lu, Zhenyu Wu, Chuanrui Zhang, Bofang Jia, Zhengyi Gu, Ziwei Wang
Abstract: Vision-Language-Action (VLA) models have demonstrated remarkable performance on complex tasks through imitation learning in recent robotic manipulation works. Based on large-scale and high-quality demonstration datasets, existing imitation learning method arms VLA models acquired with strong capabilities. However, these datasets that predominantly consist of successful trajectories, are costly to collect and often limited in distribution, leading to capability bottlenecks when faced with out-of-distribution (OOD) scenarios during deployment while unable to recover. To address this issue, we propose an automated data augmentation framework named RESample that effectively improves the distribution coverage of VLA training datasets through the well-designed exploratory sampling mechanism. Specifically, the exploratory sampling mechanism identifies the potential coverage gaps during the policy rollout and actively samples exploratory actions to extend the coverage of training data with high sample efficiency. Furthermore, to effectively reflect the distribution of the training dataset, we propose a lightweight Coverage Function that indicates the coverage density of states in the training dataset, which further guides the exploratory sampling process to focus on low-coverage regions. To validate the effectiveness of our method, we conduct extensive experiments on the LIBERO benchmark as well as a series of real-world robotic tasks, demonstrating a significant performance gain of 12% of our proposed RESample over baselines, with only 10-20% additional samples compared to original training data.
Authors: Thaweerath Phisannupawong, Joshua Julian Damanik, Han-Lim Choi
Abstract: Flight delay prediction has become a key focus in air traffic management (ATM), as delays reflect inefficiencies in the system. This paper proposes LLM4Delay, a large language model (LLM)-based framework for predicting flight delays from the perspective of air traffic controllers monitoring aircraft after they enter the terminal maneuvering area (TMA). LLM4Delay is designed to integrate textual aeronautical information, including flight data, weather reports, and aerodrome notices, together with multiple trajectories that model airspace conditions, forming a comprehensive delay-relevant context. By jointly leveraging comprehensive textual and trajectory contexts via instance-level projection, an effective cross-modality adaptation strategy that maps multiple instance-level trajectory representations into the language modality, the framework improves delay prediction accuracy. LLM4Delay demonstrates superior performance compared to existing ATM frameworks and prior time-series-to-language adaptation methods. This highlights the complementary roles of textual and trajectory data while leveraging knowledge from both the pretrained trajectory encoder and the pretrained LLM. The proposed framework enables continuous updates to predictions as new information becomes available, indicating potential operational relevance.
Authors: Taha Yasseri, Saeedeh Mohammadi
Abstract: The launch of Grokipedia, an AI-generated encyclopedia developed by Elon Musk's xAI, was presented as a response to perceived ideological and structural biases in Wikipedia, aiming to produce "truthful" entries using the Grok large language model. Yet whether an AI-driven alternative can escape the biases and limitations of human-edited platforms remains unclear. This study conducts a large-scale computational comparison of 17,790 matched article pairs from the 20,000 most-edited English Wikipedia pages. Using metrics spanning lexical richness, readability, reference density, structural features, and semantic similarity, we assess how closely the two platforms align in form and substance. We find that Grokipedia articles are substantially longer and contain significantly fewer references per word. Moreover, Grokipedia's content divides into two distinct groups: one that remains semantically and stylistically aligned with Wikipedia, and another that diverges sharply. Among the dissimilar articles, we observe a systematic rightward shift in the political bias of frequently cited news media sources, concentrated primarily in entries related to history and religion, and literature and art. More broadly, the findings indicate that AI-generated encyclopedic content departs from established editorial norms, favoring narrative expansion over citation-based verification, raising questions about transparency, provenance, and the governance of knowledge in automated information systems.
Authors: Seunghee Han, Yeonghun Kang, Taeun Bae, Junho Kim, Younghun Kim, Varinia Bernales, Alan Aspuru-Guzik, Jihan Kim
Abstract: Designing materials with targeted properties remains challenging due to the vastness of chemical space and the scarcity of property-labeled data. While recent advances in generative models offer a promising way for inverse design, most approaches require large datasets and must be retrained for every new target property. Here, we introduce the EGMOF (Efficient Generation of MOFs), a hybrid diffusion-transformer framework that overcomes these limitations through a modular, descriptor-mediated workflow. EGMOF decomposes inverse design into two steps: (1) a one-dimensional diffusion model (Prop2Desc) that maps desired properties to chemically meaningful descriptors followed by (2) a transformer model (Desc2MOF) that generates structures from these descriptors. This modular hybrid design enables minimal retraining and maintains high accuracy even under small-data conditions. On a hydrogen uptake dataset, EGMOF achieved over 95% validity and 84% hit rate, representing significant improvements of up to 57% in validity and 14% in hit rate compared to existing methods, while remaining effective with only 1,000 training samples. Moreover, our model successfully performed conditional generation across 29 diverse property datasets, including CoREMOF, QMOF, and text-mined experimental datasets, whereas previous models have not. This work presents a data-efficient, generalizable approach to the inverse design of diverse MOFs and highlights the potential of modular inverse design workflows for broader materials discovery.
Authors: Dom\'icio Pereira Neto, Jo\~ao Correia, Penousal Machado
Abstract: Deep diffusion models have revolutionized image generation by producing high-quality outputs. However, achieving specific objectives with these models often requires costly adaptations such as fine-tuning, which can be resource-intensive and time-consuming. An alternative approach is inference-time control, which involves optimizing the prompt embeddings to guide the generation process without altering the model weights. We explore prompt-embedding search optimization for the Stable Diffusion XL Turbo model, comparing a gradient-free evolutionary approach, the Separable Covariance Matrix Adaptation Evolution Strategy (sep-CMA-ES), against the widely used gradient-based optimizer Adaptive Moment Estimation (Adam). Candidate images are evaluated by a weighted objective that combines LAION Aesthetic Predictor V2 and CLIPScore, enabling explicit trade-offs between aesthetic quality and prompt-image alignment. On 36 prompts sampled from Parti Prompts (P2) under three weight settings (aesthetics-only, balanced, alignment-only), sep-CMA-ES consistently achieves higher objective values than Adam. We additionally analyze divergence from the unoptimized baseline using cosine similarity and SSIM and report the compute and memory footprints. These results suggest that sep-CMA-ES is an effective inference-time optimizer for prompt-embedding search, improving aesthetics-alignment trade-offs and resource usage without model fine-tuning.
Authors: Manan Suri, Puneet Mathur, Nedim Lipka, Franck Dernoncourt, Ryan A. Rossi, Dinesh Manocha
Abstract: LLM agents with tool-calling capabilities often fail when user instructions are ambiguous or incomplete, leading to incorrect invocations and task failures. Existing approaches operate in unstructured language spaces, generating clarifying questions through prompting strategies that lack principled criteria for determining which questions to ask and when to stop. We introduce a principled formulation of structured uncertainty that operates directly over tool parameters and their domains, cleanly separating specification uncertainty (what the user wants) from model uncertainty (what the LLM predicts). Our formulation uses Expected Value of Perfect Information (EVPI) to quantify the disambiguation value of each potential question, balanced against aspect-based cost modeling that prevents redundant questioning. We demonstrate the versatility of this formulation through two applications. First, SAGE-Agent uses structured uncertainty for inference-time question selection, achieving 7-39% higher coverage on ambiguous tasks while reducing clarification questions by 1.5-2.7x compared to strong prompting and uncertainty-based baselines. Second, we show that structured uncertainty provides effective training signals: uncertainty-guided reward modeling boosts When2Call accuracy from 36.5% to 65.2% (3B model) and 36.7% to 62.9% (7B model) through uncertainty-weighted GRPO training, demonstrating more sample-efficient reinforcement learning for tool-calling agents. To enable evaluation, we present ClarifyBench, the first multi-turn dynamic tool-calling disambiguation benchmark. Our results establish structured uncertainty as a principled framework that improves both inference-time interaction efficiency and training-time sample efficiency in tool-augmented agents.
Authors: Zhirui Liu, Kaiyang Ji, Ke Yang, Jingyi Yu, Ye Shi, Jingya Wang
Abstract: Enabling humanoid robots to follow free-form language commands is critical for seamless human-robot interaction, collaborative task execution, and general-purpose embodied intelligence. While recent advances have improved low-level humanoid locomotion and robot manipulation, language-conditioned whole-body control remains a significant challenge. Existing methods are often limited to simple instructions and sacrifice either motion diversity or physical plausibility. To address this, we introduce Humanoid-LLA, a Large Language Action Model that maps expressive language commands to physically executable whole-body actions for humanoid robots. Our approach integrates three core components: a unified motion vocabulary that aligns human and humanoid motion primitives into a shared discrete space; a vocabulary-directed controller distilled from a privileged policy to ensure physical feasibility; and a physics-informed fine-tuning stage using reinforcement learning with dynamics-aware rewards to enhance robustness and stability. Extensive evaluations in simulation and on real-world Unitree G1 and Booster T1 humanoids show that Humanoid-LLA delivers strong language generalization while maintaining high physical fidelity, outperforming existing language-conditioned controllers in motion naturalness, stability, and execution success rate.
Authors: Anik De, Abhirama Subramanyam Penamakuri, Rajeev Yadav, Aditya Rathore, Harshiv Shah, Devesh Sharma, Sagar Agarwal, Pravin Kumar, Anand Mishra
Abstract: Reading scene text, that is, text appearing in images, has numerous application areas, including assistive technology, search, and e-commerce. Although scene text recognition in English has advanced significantly and is often considered nearly a solved problem, Indian language scene text recognition remains an open challenge. This is due to script diversity, non-standard fonts, and varying writing styles, and, more importantly, the lack of high-quality datasets and open-source models. To address these gaps, we introduce the Bharat Scene Text Dataset (BSTD) - a large-scale and comprehensive benchmark for studying Indian Language Scene Text Recognition. It comprises more than 100K words that span 11 Indian languages and English, sourced from over 6,500 scene images captured across various linguistic regions of India. The dataset is meticulously annotated and supports multiple scene text tasks, including: (i) Scene Text Detection, (ii) Script Identification, (iii) Cropped Word Recognition, and (iv) End-to-End Scene Text Recognition. We evaluated state-of-the-art models originally developed for English by adapting (fine-tuning) them for Indian languages. Our results highlight the challenges and opportunities in Indian language scene text recognition. We believe that this dataset represents a significant step toward advancing research in this domain. All our models and data are open source.
Authors: Le Thien Phuc Nguyen, Zhuoran Yu, Samuel Low Yu Hang, Subin An, Jeongik Lee, Yohan Ban, SeungEun Chung, Thanh-Huy Nguyen, JuWan Maeng, Soochahn Lee, Yong Jae Lee
Abstract: Multimodal large language models (MLLMs) are expected to jointly interpret vision, audio, and language, yet existing video benchmarks rarely assess fine-grained reasoning about human speech. Many tasks remain visually solvable or only coarsely evaluate speech, offering limited insight into whether models can align who speaks, what is said, and when it occurs. We introduce AV-SpeakerBench, a curated benchmark of 3,212 multiple-choice questions focused on speaker-centric audiovisual reasoning in real-world videos. It features: (1) a speaker-centered formulation that treats speakers-not scenes-as the core reasoning unit; (2) fusion-grounded question design embedding audiovisual dependencies into question semantics; and (3) expert-curated annotations ensuring temporal precision and cross-modal validity. Comprehensive evaluations show that the Gemini family consistently outperforms open-source systems, with Gemini 2.5 Pro achieving the best results. Among open models, Qwen3-Omni-30B approaches Gemini 2.0 Flash but remains far behind Gemini 2.5 Pro, primarily due to weaker audiovisual fusion rather than visual perception. We believe AV-SpeakerBench establishes a rigorous foundation for advancing fine-grained audiovisual reasoning in future multimodal systems.
Authors: Haoming Liu, Jinnuo Liu, Yanhao Li, Liuyang Bai, Yunkai Ji, Yuanhe Guo, Shenji Wan, Hongyi Wen
Abstract: Flow-based diffusion models have emerged as a leading paradigm for training generative models across images and videos. However, their memorization-generalization behavior remains poorly understood. In this work, we revisit the flow matching (FM) objective and study its marginal velocity field, which admits a closed-form expression, allowing exact computation of the oracle FM target. Analyzing this oracle velocity field reveals that flow-based diffusion models inherently formulate a two-stage training target: an early stage guided by a mixture of data modes, and a later stage dominated by the nearest data sample. The two-stage objective leads to distinct learning behaviors: the early navigation stage generalizes across data modes to form global layouts, whereas the later refinement stage increasingly memorizes fine-grained details. Leveraging these insights, we explain the effectiveness of practical techniques such as timestep-shifted schedules, classifier-free guidance intervals, and latent space design choices. Our study deepens the understanding of diffusion model training dynamics and offers principles for guiding future architectural and algorithmic improvements. Our project page is available at: https://maps-research.github.io/from-navigation-to-refinement/.
URLs: https://maps-research.github.io/from-navigation-to-refinement/.
Authors: Melane Navaratnarajah, David A. Kelly, Hana Chockler
Abstract: Adversarial perturbations are a useful way to expose vulnerabilities in object detectors. Existing perturbation methods are frequently white-box, architecture specific and use a loss function. More importantly, while they are often successful, it is rarely clear why they work. Insights into the mechanism of this success would allow developers to understand and analyze these attacks, as well as fine-tune the model to prevent them. This paper presents BlackCAtt, a black-box algorithm and tool, which uses minimal, causally sufficient pixel sets to construct explainable, imperceptible, reproducible, architecture-agnostic attacks on object detectors. We evaluate BlackCAtt on standard benchmarks and compare it to other black-box adversarial attacks methods. When BlackCAtt has access only to the position and label of a bounding box, it produces attacks that are comparable or better to those produced by other black-box methods. When BlackCAtt has access to the model confidence as well, it can work as a meta-algorithm, improving the ability of standard black-box techniques to construct smaller, less perceptible attacks. As BlackCAtt attacks manipulate causes only, the attacks become fully explainable. We compare the performance of BlackCAtt with other black-box attack methods and show that targeting causal pixels leads to smaller and less perceptible attacks. For example, when using BlackCAtt with SquareAttack, it reduces the average distance ($L_0$ norm) of the attack from the original input from $0.987$ to $0.072$, while maintaining a similar success rate. We perform ablation studies on the BlackCAtt algorithm and analyze the effect of different components on its performance.
Authors: Zayne Sprague, Jack Lu, Manya Wadhwa, Sedrick Keh, Mengye Ren, Greg Durrett
Abstract: Reasoning models leveraging long chains of thought employ various cognitive skills, such as verification of their answers, backtracking, retrying by an alternate method, and more. Previous work has shown that when a base language model exhibits these skills, training that model further with reinforcement learning (RL) can learn to leverage them. How can we get models to leverage skills that aren't exhibited by base models? Our work, SkillFactory, is a method for fine-tuning models to roughly learn these skills during a supervised fine-tuning (SFT) stage prior to RL. Our approach does not rely on distillation from a stronger model, but instead uses samples from the model itself, rearranged to provide training data in the format of those skills. These "silver" SFT traces may be imperfect, but are nevertheless effective for priming a model to acquire skills during RL. Our evaluation shows that (1) starting from SkillFactory SFT initialization helps a model to generalize to harder variants of a task post-RL, despite lower performance pre-RL;(2) cognitive skills are indeed used by the model; (3) RLed SkillFactory models are more robust to regression on out-of-domain tasks than RLed base models. Our work suggests that inductive biases learned prior to RL help models learn robust cognitive skill use.
Authors: Thao Nguyen, Sicheng Mo, Krishna Kumar Singh, Yilin Wang, Jing Shi, Nicholas Kolkin, Eli Shechtman, Yong Jae Lee, Yuheng Li
Abstract: Humans do not just see attribute similarity -- we also see relational similarity. An apple is like a peach because both are reddish fruit, but the Earth is also like a peach: its crust, mantle, and core correspond to the peach's skin, flesh, and pit. This ability to perceive and recognize relational similarity, is arguable by cognitive scientist to be what distinguishes humans from other species. Yet, all widely used visual similarity metrics today (e.g., LPIPS, CLIP, DINO) focus solely on perceptual attribute similarity and fail to capture the rich, often surprising relational similarities that humans perceive. How can we go beyond the visible content of an image to capture its relational properties? How can we bring images with the same relational logic closer together in representation space? To answer these questions, we first formulate relational image similarity as a measurable problem: two images are relationally similar when their internal relations or functions among visual elements correspond, even if their visual attributes differ. We then curate 114k image-caption dataset in which the captions are anonymized -- describing the underlying relational logic of the scene rather than its surface content. Using this dataset, we finetune a Vision-Language model to measure the relational similarity between images. This model serves as the first step toward connecting images by their underlying relational structure rather than their visible appearance. Our study shows that while relational similarity has a lot of real-world applications, existing image similarity models fail to capture it -- revealing a critical gap in visual computing.
Authors: Qiushi Han, David Simchi-Levi, Renfei Tan, Zishuo Zhao
Abstract: We study a sequential mechanism design problem in which a principal seeks to elicit truthful reports from multiple rational agents while starting with no prior knowledge of agents' beliefs. We introduce Distributionally Robust Adaptive Mechanism (DRAM), a general framework combining insights from both mechanism design and online learning to jointly address truthfulness and cost-optimality. Throughout the sequential game, the mechanism estimates agents' beliefs and iteratively updates a distributionally robust linear program with shrinking ambiguity sets to reduce payments while preserving truthfulness. Our mechanism guarantees truthful reporting with high probability while achieving $\tilde{O}(\sqrt{T})$ cumulative regret, and we establish a matching lower bound showing that no feasible adaptive mechanism can asymptotically do better. The framework generalizes to plug-in estimators, supporting structured priors and delayed feedback. To our knowledge, this is the first adaptive mechanism under general settings that maintains truthfulness and achieves optimal regret when incentive constraints are unknown and must be learned.
Authors: Zibo Zhao (Arizona State University), Yuanting Zha (ShanghaiTech University), Haipeng Zhang (ShanghaiTech University), Xingcheng Xu (Shanghai Artificial Intelligence Laboratory)
Abstract: Self-reflection capabilities emerge in Large Language Models after RL post-training, with multi-turn RL achieving substantial gains over SFT counterparts. Yet the mechanism of how a unified optimization objective gives rise to functionally distinct capabilities of generating solutions and evaluating when to revise them remains opaque. To address this question, we introduce the Gradient Attribution Property to characterize how reward gradients distribute across policy components, formalized through the Two-Stage Decision-Sampling (DS) Hypothesis, which decomposes the policy into sampling ($\pi_{sample}$) for generation and decision ($\pi_{d}$) for verification. We prove that surrogate rewards exhibit Balanced Gradient Attribution, while SFT and KL penalties exhibit Unbalanced Gradient Attribution, with length-weighting creating asymmetric regularization that constrains $\pi_{sample}$ while leaving $\pi_{d}$ under-optimized, providing an theoretical explanation of why RL succeeds where SFT fails. We also empirically validate our theoretical predictions on arithmetic reasoning demonstrates that RL's superior generalization stems primarily from improved decision-making ($\pi_{d}$) rather than sampling capabilities, providing a first-principles mechanistic explanation for self-correction in thinking models.
Authors: Frank Mollard, Marcus Becker, Florian Roehrbein
Abstract: The paper introduces a white-box attack on computer vision models using SHAP values. It demonstrates how adversarial evasion attacks can compromise the performance of deep learning models by reducing output confidence or inducing misclassifications. Such attacks are particularly insidious as they can deceive the perception of an algorithm while eluding human perception due to their imperceptibility to the human eye. The proposed attack leverages SHAP values to quantify the significance of individual inputs to the output at the inference stage. A comparison is drawn between the SHAP attack and the well-known Fast Gradient Sign Method. We find evidence that SHAP attacks are more robust in generating misclassifications particularly in gradient hiding scenarios.
Authors: Jianan Wang, Nailei Hei, Li He, Huanzhen Wang, Aoxing Li, Yingkai Zhao, Yuxuan Lin, Haofen Wang, Chunyang Wang, Yan Wang, Wenqiang Zhang
Abstract: Human animation aims to generate temporally coherent and visually consistent videos over long sequences, yet modeling long-range dependencies while preserving frame quality remains challenging. Inspired by the human ability to leverage past observations for interpreting ongoing actions, we propose FrameCache, a training-free, causality-consistent reference frame framework. FrameCache explicitly converts historical generation results into causal guidance through two complementary mechanisms. First, at the reference level, a novel Screen-Cache-Match (SCM) strategy constructs a dynamic, high-quality reference memory, ensuring motion-consistent appearance guidance to reduce identity drift. Second, at the generative level, a Trajectory-Aware Autoregressive Generation (TAAG) mechanism aligns denoising trajectories across adjacent video chunks. This is achieved through an overlap-aware latent propagation and a dual-domain fusion strategy that seamlessly blends low-frequency structural layouts with high-frequency textural details. Extensive experiments on standard benchmarks demonstrate that FrameCache consistently improves temporal coherence and visual stability while integrating seamlessly with diverse diffusion baselines. Code will be made publicly available.
Authors: Yinsong Wang, Thomas Fletcher, Xinzhe Luo, Aine Travers Dineen, Rhodri Cusack, Chen Qin
Abstract: Reconstructing 3D fetal MR volumes from motion-corrupted stacks of 2D slices is a crucial and challenging task. Conventional slice-to-volume reconstruction (SVR) methods are time-consuming and require multiple orthogonal stacks for reconstruction. While learning-based SVR approaches have significantly reduced the time required at the inference stage, they heavily rely on ground truth information for training, which is inaccessible in practice. To address these challenges, we propose GaussianSVR, a self-supervised framework for slice-to-volume reconstruction. GaussianSVR represents the target volume using 3D Gaussian representations to achieve high-fidelity reconstruction. It leverages a simulated forward slice acquisition model to enable self-supervised training, alleviating the need for ground-truth volumes. Furthermore, to enhance both accuracy and efficiency, we introduce a multi-resolution training strategy that jointly optimizes Gaussian parameters and spatial transformations across different resolution levels. Experiments show that GaussianSVR outperforms the baseline methods on fetal MR volumetric reconstruction. Code is available at https://github.com/Yinsong0510/GaussianSVR-Self-Supervised-Slice-to-Volume-Reconstruction-with-Gaussian-Representations.
Authors: Safal Shrestha, Anubhav Shrestha, Aadim Nepal, Minwu Kim, Keith Ross
Abstract: Recent work has shown that layer pruning can effectively compress large language models (LLMs) while retaining strong performance on classification benchmarks, often with little or no finetuning. In contrast, generative reasoning tasks, such as GSM8K and HumanEval\textsuperscript{+}, exhibit substantially weaker recovery. We show that beyond surface-level text degradation, pruning leads to a loss of key algorithmic capabilities, including arithmetic computation and balanced parenthesis generation. Under realistic post-training constraints, without access to pretraining-scale data or compute, we evaluate a minimal recovery strategy based on supervised finetuning with self-generated responses. This approach recovers up to 90\% of baseline performance on classification tasks, but recovery for generative reasoning remains fundamentally limited. Notably, even models finetuned on $\sim$400B tokens after pruning fail to recover their original reasoning performance, suggesting that such capabilities are not as easily restored. This limitation persists even on simple tasks such as arithmetic, which do not require multi-step generation. Overall, we characterize the practical limits of layer pruning for generative reasoning and provide guidance on when depth reduction is effective under constrained post-training regimes.
Authors: Bryan Sangwoo Kim, Jonghyun Park, Jong Chul Ye
Abstract: Text-conditioned diffusion models have advanced image and video super-resolution by using prompts as semantic priors, and modern super-resolution pipelines typically rely on latent tiling to scale to high resolutions. In practice, a single global caption is used with the latent tiling, often causing prompt misguidance. Specifically, a coarse global prompt often misses localized details (errors of omission) and provides locally irrelevant guidance (errors of commission) which leads to substandard results at the tile level. To solve this, we propose Tiled Prompts, a unified framework for image and video super-resolution that generates a tile-specific prompt for each latent tile and performs super-resolution under locally text-conditioned posteriors to resolve prompt misguidance with minimal overhead. Our experiments on high resolution real-world images and videos show that tiled prompts bring consistent gains in perceptual quality and fidelity, while reducing hallucinations and tile-level artifacts that can be found in global-prompt baselines. Project Page: https://bryanswkim.github.io/tiled-prompts/.
Authors: Indraveni Chebolu, Arnab Mallick, Harmesh Rana
Abstract: We present SPEAR, a multi-agent coordination framework for smart contract auditing that applies established MAS patterns in a realistic security analysis workflow. SPEAR models auditing as a coordinated mission carried out by specialized agents: a Planning Agent prioritizes contracts using risk-aware heuristics, an Execution Agent allocates tasks via the Contract Net protocol, and a Repair Agent autonomously recovers from brittle generated artifacts using a programmatic-first repair policy. Agents maintain local beliefs updated through AGM-compliant revision, coordinate via negotiation and auction protocols, and revise plans as new information becomes available. An empirical study compares the multi-agent design with centralized and pipeline-based alternatives under controlled failure scenarios, focusing on coordination, recovery behavior, and resource use.
Authors: Eun Cheol Choi, Lindsay E. Young, Emilio Ferrara
Abstract: Large language models (LLMs) are increasingly used as proxies for human judgment in computational social science, yet their ability to reproduce patterns of susceptibility to misinformation remains unclear. We test whether LLM-simulated survey respondents, prompted with participant profiles drawn from social survey data measuring network, demographic, attitudinal and behavioral features, can reproduce human patterns of misinformation belief and sharing. Using three online surveys as baselines, we evaluate whether LLM outputs match observed response distributions and recover feature-outcome associations present in the original survey data. LLM-generated responses capture broad distributional tendencies and show modest correlation with human responses, but consistently overstate the association between belief and sharing. Linear models fit to simulated responses exhibit substantially higher explained variance and place disproportionate weight on attitudinal and behavioral features, while largely ignoring personal network characteristics, relative to models fit to human responses. Analyses of model-generated reasoning and LLM training data suggest that these distortions reflect systematic biases in how misinformation-related concepts are represented. Our findings suggest that LLM-based survey simulations are better suited for diagnosing systematic divergences from human judgment than for substituting it.
Authors: Prerna Ravi, Car\'umey Stevens, Beatriz Flamia Azevedo, Jasmine David, Brandon Hanks, Hal Abelson, Grace Lin, Emma Anderson
Abstract: Collaboration is a cornerstone of 21st-century learning, yet teachers continue to face challenges in supporting productive peer interaction. Emerging generative AI tools offer new possibilities for scaffolding collaboration, but their role in mediating in-person group work remains underexplored, especially from the perspective of educators. This paper presents findings from an exploratory qualitative study with 33 K12 teachers who interacted with Phoenix, a voice-based conversational agent designed to function as a near-peer in face-to-face group collaboration. Drawing on playtesting sessions, surveys, and focus groups, we examine how teachers perceived the agent's behavior, its influence on group dynamics, and its classroom potential. While many appreciated Phoenix's capacity to stimulate engagement, they also expressed concerns around autonomy, trust, anthropomorphism, and pedagogical alignment. We contribute empirical insights into teachers' mental models of AI, reveal core design tensions, and outline considerations for group-facing AI agents that support meaningful, collaborative learning.
Authors: Adolfo Gonz\'alez, V\'ictor Parada
Abstract: Business environments characterized by intermittent demand, high variability, and multi-step planning horizons require forecasting policies that support consistent operational decisions across heterogeneous SKU portfolios. Because no forecasting model is universally dominant, and model rankings vary across error metrics, demand regimes, and forecast horizons, forecasting model assignment is a nontrivial decision problem in inventory planning, procurement, and supply management. This study proposes the Adaptive Hybrid Selector for Intermittency and Variability (AHSIV), a horizon-aware and regime-conditioned model selection framework designed to address horizon-induced ranking instability. The framework integrates the Metric Degradation by Forecast Horizon (MDFH), structural demand classification, multi-objective Pareto dominance, and hierarchical bias refinement within a unified decision architecture. The empirical evaluation is conducted on the Walmart, M3, M4, and M5 datasets under multiple train-test partition schemes and twelve-step forecasting horizons. The results show that AHSIV achieves statistical equivalence with the strongest single-metric baseline in aggregated performance while improving the consistency of model assignment across forecast horizons in heterogeneous demand settings. These findings indicate that forecasting model selection in multi-SKU environments should not be treated as a static ranking problem. Instead, horizon-consistent and structurally adaptive selection mechanisms can provide more robust support for inventory planning, procurement alignment, and operational decision-making in business forecasting systems.
Authors: Rong Fu, Zijian Zhang, Kun Liu, Jiekai Wu, Xianda Li, Simon Fong
Abstract: Comparative analysis of adaptive immune repertoires at population scale is hampered by two practical bottlenecks: the near-quadratic cost of pairwise affinity evaluations and dataset imbalances that obscure clinically important minority clonotypes. We introduce SubQuad, an end-to-end pipeline that addresses these challenges by combining antigen-aware, near-subquadratic retrieval with GPU-accelerated affinity kernels, learned multimodal fusion, and fairness-constrained clustering. The system employs compact MinHash prefiltering to sharply reduce candidate comparisons, a differentiable gating module that adaptively weights complementary alignment and embedding channels on a per-pair basis, and an automated calibration routine that enforces proportional representation of rare antigen-specific subgroups. On large viral and tumor repertoires SubQuad achieves measured gains in throughput and peak memory usage while preserving or improving recall@k, cluster purity, and subgroup equity. By co-designing indexing, similarity fusion, and equity-aware objectives, SubQuad offers a scalable, bias-aware platform for repertoire mining and downstream translational tasks such as vaccine target prioritization and biomarker discovery.
Authors: Joao Manoel Herrera Pinheiro, Gabriela Do Nascimento Herrera, Luciana Bueno Dos Reis Fernandes, Alvaro Doria Dos Santos, Ricardo V. Godoy, Eduardo A. B. Almeida, Helena Carolina Onody, Marcelo Andrade Da Costa Vieira, Angelica Maria Penteado-Dias, Marcelo Becker
Abstract: Accurate taxonomic identification is the cornerstone of biodiversity monitoring and agricultural management, particularly for the hyper-diverse superfamily Ichneumonoidea. Comprising the families Ichneumonidae and Braconidae, these parasitoid wasps are ecologically critical for regulating insect populations, yet they remain one of the most taxonomically challenging groups due to their cryptic morphology and vast number of undescribed species. To address the scarcity of robust digital resources for these key groups, we present a curated image dataset designed to advance automated identification systems. The dataset contains 3,556 high-resolution images, primarily focused on Neotropical Ichneumonidae and Braconidae, while also including supplementary families such as Andrenidae, Apidae, Bethylidae, Chrysididae, Colletidae, Halictidae, Megachilidae, Pompilidae, and Vespidae to improve model robustness. Crucially, a subset of 1,739 images is annotated in COCO format, featuring multi-class bounding boxes for the full insect body, wing venation, and scale bars. This resource provides a foundation for developing computer vision models capable of identifying these families.
Authors: Zhaoyang Zhang, Shuli Jiang, Yantao Shen, Yuting Zhang, Dhananjay Ram, Shuo Yang, Zhuowen Tu, Wei Xia, Stefano Soatto
Abstract: Reinforcement learning (RL) post-training has recently driven major gains in long chain-of-thought reasoning large language models (LLMs), but the high inference cost of such models motivates distillation into smaller students. Most existing knowledge distillation (KD) methods are designed for supervised fine-tuning (SFT), relying on fixed teacher traces or teacher-student Kullback-Leibler (KL) divergence-based regularization. When combined with RL, these approaches often suffer from distribution mismatch and objective interference: teacher supervision may not align with the student's evolving rollout distribution, and the KL regularizer can compete with reward maximization and require careful loss balancing. To address these issues, we propose RL-aware distillation (RLAD), which performs selective imitation during RL -- guiding the student toward the teacher only when it improves the current policy update. Our core component, Trust Region Ratio Distillation (TRRD), replaces the teacher-student KL regularizer with a PPO/GRPO-style likelihood-ratio objective anchored to a teacher--old-policy mixture, yielding advantage-aware, trust-region-bounded distillation on student rollouts and naturally balancing exploration, exploitation, and imitation. Across diverse logic reasoning and math benchmarks, RLAD consistently outperforms offline distillation, standard GRPO, and KL-based on-policy teacher-student knowledge distillation.
Authors: Ruinan Jin, Yingbin Liang, Shaofeng Zou
Abstract: Despite Adam demonstrating faster empirical convergence than SGD in many applications, much of the existing theory yields guarantees essentially comparable to those of SGD, leaving the empirical performance gap insufficiently explained. In this paper, we uncover a key second-moment normalization in Adam and develop a stopping-time/martingale analysis that provably distinguishes Adam from SGD under the classical bounded variance model (a second moment assumption). In particular, we establish the first theoretical separation between the high-probability convergence behaviors of the two methods: Adam achieves a $\delta^{-1/2}$ dependence on the confidence parameter $\delta$, whereas corresponding high-probability guarantee for SGD necessarily incurs at least a $\delta^{-1}$ dependence.
Authors: Yuan Wu, Zongxian Yang, Jiayu Qian, Songpan Gao, Guanxing Chen, Qiankun Li, Yu-An Huang, Zhi-An Huang
Abstract: Large vision-language models (VLMs) often benefit from chain-of-thought (CoT) prompting in general domains, yet its efficacy in medical vision-language tasks remains underexplored. We report a counter-intuitive trend: on medical visual question answering, CoT frequently underperforms direct answering (DirA) across general-purpose and medical-specific models. We attribute this to a \emph{medical perception bottleneck}: subtle, domain-specific cues can weaken visual grounding, and CoT may compound early perceptual uncertainty rather than correct it. To probe this hypothesis, we introduce two training-free, inference-time grounding interventions: (i) \emph{perception anchoring} via region-of-interest cues and (ii) \emph{description grounding} via high-quality textual guidance. Across multiple benchmarks and model families, these interventions improve accuracy, mitigate CoT degradation, and in several settings reverse the CoT--DirA inversion. Our findings suggest that reliable clinical VLMs require robust visual grounding and cross-modal alignment, beyond extending text-driven reasoning chains. Code is available \href{https://github.com/TianYin123/Better_Eyes_Better_Thoughts}{here}.
URLs: https://github.com/TianYin123/Better_Eyes_Better_Thoughts
Authors: Minh-Duong Nguyen, Thien-Thanh Dao, Le-Tuan Nguyen, Dung D. Le, Kok-Seng Wong
Abstract: Rehearsal-based continual learning (CL) mitigates catastrophic forgetting by maintaining a subset of samples from previous tasks for replay. Existing studies primarily focus on optimizing memory storage through coreset selection strategies. While these methods are effective, they typically require storing a substantial number of samples per class (SPC), often exceeding 20, to maintain satisfactory performance. In this work, we propose to further compress the memory footprint by synthesizing and storing prototypical exemplars, which can form representative prototypes when passed through the feature extractor. Owing to their representative nature, these exemplars enable the model to retain previous knowledge using only a small number of samples while preserving privacy. Moreover, we introduce a perturbation-based augmentation mechanism that generates synthetic variants of previous data during training, thereby enhancing CL performance. Extensive evaluations on widely used benchmark datasets and settings demonstrate that the proposed algorithm achieves superior performance compared to existing baselines, particularly in scenarios involving large-scale datasets and a high number of tasks.
Authors: Zhexi Lian, Haoran Wang, Xuerun Yan, Weimeng Lin, Xianhong Zhang, Yongyu Chen, Jia Hu
Abstract: End-to-end autonomous driving is typically built upon imitation learning (IL), yet its performance is constrained by the quality of human demonstrations. To overcome this limitation, recent methods incorporate reinforcement learning (RL) through sequential fine-tuning. However, such a paradigm remains suboptimal: sequential RL fine-tuning can introduce policy drift and often leads to a performance ceiling due to its dependence on the pretrained IL policy. To address these issues, we propose PaIR-Drive, a general Parallel framework for collaborative Imitation and Reinforcement learning in end-to-end autonomous driving. During training, PaIR-Drive separates IL and RL into two parallel branches with conflict-free training objectives, enabling fully collaborative optimization. This design eliminates the need to retrain RL when applying a new IL policy. During inference, RL leverages the IL policy to further optimize the final plan, allowing performance beyond prior knowledge of IL. Furthermore, we introduce a tree-structured trajectory neural sampler to group relative policy optimization (GRPO) in the RL branch, which enhances exploration capability. Extensive analysis on NAVSIMv1 and v2 benchmark demonstrates that PaIR-Drive achieves Competitive performance of 91.2 PDMS and 87.9 EPDMS, building upon Transfuser and DiffusionDrive IL baselines. PaIR-Drive consistently outperforms existing RL fine-tuning methods, and could even correct human experts' suboptimal behaviors. Qualitative results further confirm that PaIR-Drive can effectively explore and generate high-quality trajectories.
Authors: Omkar Patil, Ondrej Biza, Thomas Weng, Karl Schmeckpeper, Wil Thomason, Xiaohan Zhang, Robin Walters, Nakul Gopalan, Sebastian Castro, Eric Rosen
Abstract: What happens when a pretrained generative robot policy is provided a constant initial noise as input, rather than repeatedly sampling it from a Gaussian? We demonstrate that the performance of a pretrained, frozen diffusion or flow matching policy can be improved with respect to a downstream reward by swapping the sampling of initial noise from the prior distribution (typically isotropic Gaussian) with a well-chosen, constant initial noise input -- a golden ticket. We propose a search method to find golden tickets using Monte-Carlo policy evaluation that keeps the pretrained policy frozen, does not train any new networks, and is applicable to all diffusion/flow matching policies (and therefore many VLAs). Our approach to policy improvement makes no assumptions beyond being able to inject initial noise into the policy and calculate (sparse) task rewards of episode rollouts, making it deployable with no additional infrastructure or models. Our method improves the performance of policies in 38 out of 43 tasks across simulated and real-world robot manipulation benchmarks, with relative improvements in success rate by up to 58% for some simulated tasks, and 60% within 50 search episodes for real-world tasks. We also show unique benefits of golden tickets for multi-task settings: the diversity of behaviors from different tickets naturally defines a Pareto frontier for balancing different objectives (e.g., speed, success rates); in VLAs, we find that a golden ticket optimized for one task can also boost performance in other related tasks. We release a codebase with pretrained policies and golden tickets for simulation benchmarks using VLAs, diffusion policies, and flow matching policies.
Authors: Mengxian Lyu, Cheng Peng, Ziyi Chen, Mengyuan Zhang, Jieting Li Lu, Yonghui Wu
Abstract: Automatic summarization of radiology reports is an essential application to reduce the burden on physicians. Previous studies have widely used the "pre-training, fine-tuning" strategy to adapt large language models (LLMs) for summarization. This study proposed a subdomain adaptation through a mid-training method to improve summarization. We explored three adaptation strategies: (1) general-domain pre-training, (2) clinical-domain pre-training, and (3) clinical-domain pre-training followed by subdomain mid-training. We developed models using large-scale clinical text from the University of Florida (UF) Health and conducted mid-training and fine-tuning experiments using widely used benchmark datasets including OpenI and MIMIC-CXR. The experimental results show that the mid-trained model, GatorTronT5-Radio, achieved the best performance, outperforming models without mid-training in both text-based measures (ROUGE-L) and factuality measures (RadGraph-F1). Our mid-training methods also demonstrate better few-shot learning and could alleviate the "cold start" problem reported in previous studies as a learning barrier. Our findings support the use of "pre-training, mid-training, fine-tuning," instead of the widely used direct fine-tuning strategy.
Authors: Sen Jia, Ning Zhu, Jinqin Zhong, Jiale Zhou, Huaping Zhang, Jenq-Neng Hwang, Lei Li
Abstract: RAM incorporates a motion-aware semantic tracker with adaptive Kalman filtering to achieve robust identity association under severe occlusions and dynamic interactions. A memory-augmented Temporal HMR module further enhances human motion reconstruction by injecting spatio-temporal priors for consistent and smooth motion estimation. Moreover, a lightweight Predictor module forecasts future poses to maintain reconstruction continuity, while a gated combiner adaptively fuses reconstructed and predicted features to ensure coherence and robustness. Experiments on in-the-wild multi-person benchmarks such as PoseTrack and 3DPW, demonstrate that RAM substantially outperforms previous state-of-the-art in both Zero-shot tracking stability and 3D accuracy, offering a generalizable paradigm for markerless 3D human motion capture in-the-wild.
Authors: Clemens Watzenb\"ock, Daniel Aletaha, Micha\"el Deman, Thomas Deimel, Jana Eder, Ivana Janickova, Robert Janiczek, Peter Mandl, Philipp Seeb\"ock, Gabriela Supp, Paul Weiser, Georg Langs
Abstract: Quantitative disease severity scoring in medical imaging is costly, time-consuming, and subject to inter-reader variability. At the same time, clinical archives contain far more longitudinal imaging data than expert-annotated severity scores. Existing self-supervised methods typically ignore this chronological structure. We introduce ChronoCon, a contrastive learning approach that replaces label-based ranking losses with rankings derived solely from the visitation order of a patient's longitudinal scans. Under the clinically plausible assumption of monotonic progression in irreversible diseases, the method learns disease-relevant representations without using any expert labels. This generalizes the idea of Rank-N-Contrast from label distances to temporal ordering. Evaluated on rheumatoid arthritis radiographs for severity assessment, the learned representations substantially improve label efficiency. In low-label settings, ChronoCon significantly outperforms a fully supervised baseline initialized from ImageNet weights. In a few-shot learning experiment, fine-tuning ChronoCon on expert scores from only five patients yields an intraclass correlation coefficient of 86% for severity score prediction. These results demonstrate the potential of chronological contrastive learning to exploit routinely available imaging metadata to reduce annotation requirements in the irreversible disease domain. Code is available at https://github.com/cirmuw/ChronoCon.
Authors: Robert Aufschl\"ager, Jakob Folz, Gautam Savaliya, Manjitha D Vidanalage, Michael Heigl, Martin Schramm
Abstract: Street-level imagery contains personally identifiable information (PII), some of which is context-dependent. Existing anonymization methods either over-process images or miss subtle identifiers, while API-based solutions compromise data sovereignty. We present an agentic framework CAIAMAR (\underline{C}ontext-\underline{A}ware \underline{I}mage \underline{A}nonymization with \underline{M}ulti-\underline{A}gent \underline{R}easoning) for context-aware PII segmentation with diffusion-based anonymization, combining pre-defined processing for high-confidence cases with multi-agent reasoning for indirect identifiers. Three specialized agents coordinate via round-robin speaker selection in a Plan-Do-Check-Act (PDCA) cycle, enabling large vision-language models to classify PII based on spatial context (private vs. public property) rather than rigid category rules. The agents implement spatially-filtered coarse-to-fine detection where a scout-and-zoom strategy identifies candidates, open-vocabulary segmentation processes localized crops, and $IoU$-based deduplication ($30\%$ threshold) prevents redundant processing. Modal-specific diffusion guidance with appearance decorrelation substantially reduces re-identification (Re-ID) risks. On CUHK03-NP, our method reduces person Re-ID risk by $73\%$ ($R1$: $16.9\%$ vs. $62.4\%$ baseline). For image quality preservation on CityScapes, we achieve KID: $0.001$, and FID: $9.1$, significantly outperforming existing anonymization. The agentic workflow detects non-direct PII instances across object categories, and downstream semantic segmentation is preserved. Operating entirely on-premise with open-source models, the framework generates human-interpretable audit trails supporting EU's GDPR transparency requirements while flagging failed cases for human review.
Authors: Haochuan Kevin Wang, Zechen Zhang
Abstract: Multi-agent LLM systems are entering production -- processing documents, managing workflows, acting on behalf of users -- yet their resilience to prompt injection is still evaluated with a single binary: did the attack succeed? This leaves architects without the diagnostic information needed to harden real pipelines. We introduce a kill-chain canary methodology that tracks a cryptographic token through four stages (EXPOSED -> PERSISTED -> RELAYED -> EXECUTED) across 950 runs, five frontier LLMs, six attack surfaces, and five defense conditions. The results reframe prompt injection as a pipeline-architecture problem: every model is fully exposed, yet outcomes diverge downstream -- Claude blocks all injections at memory-write (0/164 ASR), GPT-4o-mini propagates at 53%, and DeepSeek exhibits 0%/100% across surfaces from the same model. Three findings matter for deployment: (1) write-node placement is the highest-leverage safety decision -- routing writes through a verified model eliminates propagation; (2) all four defenses fail on at least one surface due to channel mismatch alone, no adversarial adaptation required; (3) invisible whitefont PDF payloads match or exceed visible-text ASR, meaning rendered-layer screening is insufficient. These dynamics apply directly to production: institutional investors and financial firms already run NLP pipelines over earnings calls, SEC filings, and analyst reports -- the document-ingestion workflows now migrating to LLM agents. Code, run logs, and tooling are publicly released.
Authors: Hita Kambhamettu, Will Crichton, Sean Welleck, Harrison Goldstein, Andrew Head
Abstract: LLM-generated explanations can make technical content more accessible, but there is a ceiling on what they can support interactively. Because LLM outputs are static text, they cannot be executed or stepped through. We argue that grounding explanations in a formalized representation enables interactive affordances beyond what static text supports. We instantiate this idea for mathematical proof comprehension with explorable theorems, a system that uses LLMs to translate a theorem and its written proof into Lean, a programming language for machine-checked proofs, and links the written proof with the Lean code. Readers can work through the proof at a step-level granularity, test custom examples or counterexamples, and trace the logical dependencies bridging each step. Each worked-out step is produced by executing the Lean proof on that example and extracting its intermediate state. A user study ($n = 16$) shows potential advantages of this approach: in a proof-reading task, participants who had access to the provided explorability features gave better, more correct, and more detailed answers to comprehension questions, demonstrating a stronger overall understanding of the underlying mathematics.
Authors: Myra Cheng, Isabel Sieh, Humishka Zope, Sunny Yu, Lujain Ibrahim, Aryaman Arora, Jared Moore, Desmond Ong, Dan Jurafsky, Diyi Yang
Abstract: LLMs can be socially sycophantic, affirming users when they ask questions like "am I in the wrong?" rather than providing genuine assessment. We hypothesize that this behavior arises from incorrect assumptions about the user, like underestimating how often users are seeking information over reassurance. We present Verbalized Assumptions, a framework for eliciting these assumptions from LLMs. Verbalized Assumptions provide insight into LLM sycophancy, delusion, and other safety issues, e.g., the top bigram in LLMs' assumptions on social sycophancy datasets is ``seeking validation.'' We provide evidence for a causal link between Verbalized Assumptions and sycophantic model behavior: our assumption probes (linear probes trained on internal representations of these assumptions) enable interpretable fine-grained steering of social sycophancy. We explore why LLMs default to sycophantic assumptions: on identical queries, people expect more objective and informative responses from AI than from other humans, but LLMs trained on human-human conversation do not account for this difference in expectations. Our work contributes a new understanding of assumptions as a mechanism for sycophancy.
Authors: Yi Zhou
Abstract: Large language models (LLMs) can generate code rapidly but remain unreliable for scientific algorithms whose correctness depends on structural assumptions rarely explicit in the source literature. We introduce a multi-stage LLM-assisted workflow that separates theory extraction, formal specification, and code implementation. The key step is an intermediate technical specification -- produced by a dedicated LLM agent and reviewed by the human researcher -- that externalizes implementation-critical computational knowledge absent from the source literature, including explicit index conventions, contraction orderings, and matrix-free operational constraints that avoid explicit storage of large operator matrices. A controlled comparison shows that it is this externalized content, rather than the formal document structure, that enables reliable code generation. As a stringent benchmark, we apply this workflow to the Density-Matrix Renormalization Group (DMRG), a canonical quantum many-body algorithm requiring exact tensor-index logic, gauge consistency, and memory-aware contractions. The resulting code reproduces the critical entanglement scaling of the spin-$1/2$ Heisenberg chain and the symmetry-protected topological order of the spin-$1$ Affleck--Kennedy--Lieb--Tasaki model. Across 16 tested combinations of leading foundation models, all workflows satisfied the same physics-validation criteria, compared to a 46\% success rate for direct, unmediated implementation. The workflow reduced a development cycle typically requiring weeks of graduate-level effort to under 24 hours.
Authors: Cheol Woo Kim, Jai Moondra, Roozbeh Nahavandi, Andrew Perrault, Milind Tambe, Swati Gupta
Abstract: The holy grail of LLM personalization is a single LLM for each user, perfectly aligned with that user's preferences. However, maintaining a separate LLM per user is impractical due to constraints on compute, memory, and system complexity. We address this challenge by developing a principled method for selecting a small portfolio of LLMs that captures representative behaviors across heterogeneous users. We model user preferences across multiple traits (e.g., safety, humor, brevity) through a multi-dimensional weight vector. Given reward functions across these dimensions, our algorithm PALM (Portfolio of Aligned LLMs) generates a small portfolio of LLMs such that, for any weight vector, the portfolio contains a near-optimal LLM for the corresponding scalarized objective. To the best of our knowledge, this is the first result that provides theoretical guarantees on both the size and approximation quality of LLM portfolios for personalization. It characterizes the trade-off between system cost and personalization, as well as the diversity of LLMs required to cover the landscape of user preferences. We provide empirical results that validate these guarantees and demonstrate greater output diversity over common baselines.
Authors: Zequn Chen, Wesley J. Marrero
Abstract: Researchers and practitioners are increasingly considering reinforcement learning to optimize decisions in complex domains like robotics and healthcare. To date, these efforts have largely utilized expectation-based learning. However, relying on expectation-focused objectives may be insufficient for making consistent decisions in highly uncertain situations involving multiple heterogeneous groups. While distributional reinforcement learning algorithms have been introduced to model the full distributions of outcomes, they can yield large discrepancies in realized benefits among comparable agents. This challenge is particularly acute in healthcare settings, where physicians (controllers) must manage multiple patients (subordinate agents) with uncertain disease progression and heterogeneous treatment responses. We propose a Boosted Distributional Reinforcement Learning (BDRL) algorithm that optimizes agent-specific outcome distributions while enforcing comparability among similar agents and analyze its convergence. To further stabilize learning, we incorporate a post-update projection step formulated as a constrained convex optimization problem, which efficiently aligns individual outcomes with a high-performing reference within a specified tolerance. We apply our algorithm to manage hypertension in a large subset of the US adult population by categorizing individuals into cardiovascular disease risk groups. Our approach modifies treatment plans for median and vulnerable patients by mimicking the behavior of high-performing references in each risk group. Furthermore, we find that BDRL improves the number and consistency of quality-adjusted life years compared with reinforcement learning baselines.
Authors: Jingwei Zuo, Xinze Feng, Zien Liu, Kaijian Wang, Fanjiang Ye, Ye Cao, Zhuang Wang, Yuke Wang
Abstract: Low-Rank Adaptation (LoRA) is now the dominant method for parameter-efficient fine-tuning of large language models, but achieving a high-quality adapter often requires systematic hyperparameter tuning because LoRA performance is highly sensitive to configuration choices. In practice, this leads to many concurrent LoRA jobs, often spanning heterogeneous tasks in multi-tenant environments. Existing systems largely handle these jobs independently, which both wastes computation on weak candidates and leaves GPUs underutilized. We present ALTO (Adaptive LoRA Tuning and Orchestration), a co-designed training system that accelerates LoRA hyperparameter tuning while enabling efficient cluster sharing across heterogeneous tasks. The central insight behind ALTO is that when multiple tuning jobs run concurrently over a shared frozen backbone, they expose optimization opportunities that single-job designs cannot exploit. Building on this, ALTO monitors loss trajectories to terminate unpromising configurations early, uses fused grouped GEMM together with a new rank-local adapter parallelism to co-locate surviving adapters and reclaim freed GPU capacity, and combines intra-task and inter-task scheduling to improve multi-task placement by leveraging the predictable duration of LoRA jobs. Extensive evaluation shows that ALTO achieves up to $13.8\times$ speedup over state-of-the-art without sacrificing adapter quality.
Authors: Zhengming Yu, Li Ma, Mingming He, Leo Isikdogan, Yuancheng Xu, Dmitriy Smirnov, Pablo Salamanca, Dao Mi, Pablo Delgado, Ning Yu, Julien Philip, Xin Li, Wenping Wang, Paul Debevec
Abstract: Most digital videos are stored in 8-bit low dynamic range (LDR) formats, where much of the original high dynamic range (HDR) scene radiance is lost due to saturation and quantization. This loss of highlight and shadow detail precludes mapping accurate luminance to HDR displays and limits meaningful re-exposure in post-production workflows. Although techniques have been proposed to convert LDR images to HDR through dynamic range expansion, they struggle to restore realistic detail in the over- and underexposed regions. To address this, we present DiffHDR, a framework that formulates LDR-to-HDR conversion as a generative radiance inpainting task within the latent space of a video diffusion model. By operating in Log-Gamma color space, DiffHDR leverages spatio-temporal generative priors from a pretrained video diffusion model to synthesize plausible HDR radiance in over- and underexposed regions while recovering the continuous scene radiance of the quantized pixels. Our framework further enables controllable LDR-to-HDR video conversion guided by text prompts or reference images. To address the scarcity of paired HDR video data, we develop a pipeline that synthesizes high-quality HDR video training data from static HDRI maps. Extensive experiments demonstrate that DiffHDR significantly outperforms state-of-the-art approaches in radiance fidelity and temporal stability, producing realistic HDR videos with considerable latitude for re-exposure.
Authors: Yiquan Wu, Yuhang Liu, Yifei Liu, Ang Li, Siying Zhou, Kun Kuang, Fei Wu
Abstract: Large language models have demonstrated remarkable capabilities across a wide range of natural language processing tasks, yet their application in the legal domain remains challenging due to the specialized terminology, complex reasoning requirements, and rapidly evolving legal knowledge involved. In this paper, we present WisdomInterrogatory (LuWen), an open-source Chinese legal language model built upon the Baichuan foundation model through three key techniques: continual pre-training on a large-scale legal corpus, supervised fine-tuning with carefully curated legal instruction data, and retrieval-augmented generation integrated with a comprehensive legal knowledge base. We evaluate LuWen on five representative legal tasks spanning both prediction and generation settings, including legal judgment prediction, judicial examination, legal text summarization, law article question answering, and judicial decision reasoning. Experimental results show that LuWen outperforms several strong baselines, demonstrating the effectiveness of our approach in adapting general-purpose language models to the legal domain.
Authors: Xue Qin, Simin Luan, John See, Cong Yang, Zhijun Li
Abstract: Embodied agents are increasingly expected to improve over time by updating their executable capabilities rather than rewriting the agent itself. Prior work has separately studied modular capability packaging, capability evolution, and runtime governance. However, a key systems problem remains underexplored: once an embodied capability module evolves into a new version, how can the hosting system deploy it safely without breaking policy constraints, execution assumptions, or recovery guarantees? We formulate governed capability evolution as a first-class systems problem for embodied agents. We propose a lifecycle-aware upgrade framework in which every new capability version is treated as a governed deployment candidate rather than an immediately executable replacement. The framework introduces four upgrade compatibility checks -- interface, policy, behavioral, and recovery -- and organizes them into a staged runtime pipeline comprising candidate validation, sandbox evaluation, shadow deployment, gated activation, online monitoring, and rollback. We evaluate over 6 rounds of capability upgrade with 15 random seeds. Naive upgrade achieves 72.9% task success but drives unsafe activation to 60% by the final round; governed upgrade retains comparable success (67.4%) while maintaining zero unsafe activations across all rounds (Wilcoxon p=0.003). Shadow deployment reveals 40% of regressions invisible to sandbox evaluation alone, and rollback succeeds in 79.8% of post-activation drift scenarios.
Authors: Seungjae Moon, Seunghyun Oh, Youngmin Ro
Abstract: Training-free open-vocabulary semantic segmentation(TF-OVSS) has recently attracted attention for its ability to perform dense prediction by leveraging the pretrained knowledge of large vision and vision-language models, without requiring additional training. However, due to the limited input resolution of these pretrained encoders, existing TF-OVSS methods commonly adopt a sliding-window strategy that processes cropped sub-images independently. While effective for managing high-resolution inputs, this approach prevents global attention over the full image, leading to fragmented feature representations and limited contextual reasoning. We propose OV-Stitcher, a training-free framework that addresses this limitation by stitching fragmented sub-image features directly within the final encoder block. By reconstructing attention representations from fragmented sub-image features, OV-Stitcher enables global attention within the final encoder block, producing coherent context aggregation and spatially consistent, semantically aligned segmentation maps. Extensive evaluations across eight benchmarks demonstrate that OV-Stitcher establishes a scalable and effective solution for open-vocabulary segmentation, achieving a notable improvement in mean Intersection over Union(mIoU) from 48.7 to 50.7 compared with prior training-free baselines.
Authors: Juwei Yue, Chuanrui Hu, Jiawei Sheng, Zuyi Zhou, Wenyuan Zhang, Tingwen Liu, Li Guo, Yafeng Deng
Abstract: Long-term memory is essential for conversational agents to maintain coherence, track persistent tasks, and provide personalized interactions across extended dialogues. However, existing approaches as Retrieval-Augmented Generation (RAG) and graph-based memory mostly rely on pairwise relations, which can hardly capture high-order associations, i.e., joint dependencies among multiple elements, causing fragmented retrieval. To this end, we propose HyperMem, a hypergraph-based hierarchical memory architecture that explicitly models such associations using hyperedges. Particularly, HyperMem structures memory into three levels: topics, episodes, and facts, and groups related episodes and their facts via hyperedges, unifying scattered content into coherent units. Leveraging this structure, we design a hybrid lexical-semantic index and a coarse-to-fine retrieval strategy, supporting accurate and efficient retrieval of high-order associations. Experiments on the LoCoMo benchmark show that HyperMem achieves state-of-the-art performance with 92.73% LLM-as-a-judge accuracy, demonstrating the effectiveness of HyperMem for long-term conversations.
Authors: Nishikanta Mohanty, Bikash K. Behera, Badshah Mukherjee, Pravat Dash
Abstract: We present a quantum-inspired ARIMA methodology that integrates quantum-assisted lag discovery with fixed-configuration variational quantum circuits (VQCs) for parameter estimation and weak-lag refinement. Differencing and candidate lags are identified via swap-test-driven quantum autocorrelation (QACF) and quantum partial autocorrelation (QPACF), with a delayed-matrix construction that aligns quantum projections to time-domain regressors, followed by standard information-criterion parsimony. Given the screened orders (p,d,q), we retain a fixed VQC ansatz, optimizer, and training budget, preventing hyperparameter leakage, and deploy the circuit in two estimation roles: VQC-AR for autoregressive coefficients and VQC-MA for moving-average coefficients. Between screening and estimation, a lightweight VQC weak-lag refinement re-weights or prunes screened AR lags without altering (p,d,q). Across environmental and industrial datasets, we perform rolling-origin evaluations against automated classical ARIMA, reporting out-of-sample mean squared error (MSE), mean absolute percentage error (MAPE), and Diebold-Mariano tests on MSE and MAE. Empirically, the seven quantum contributions (1) differencing selection, (2) QACF, (3) QPACF, (4) swap-test primitives with delayed-matrix construction, (5) VQC-AR, (6) VQC weak-lag refinement, and (7) VQC-MA collectively reduce meta-optimization overhead and make explicit where quantum effects enter order discovery, lag refinement, and AR/MA parameter estimation.
Authors: Rui Gan, Junyi Ma, Pei Li, Xingyou Yang, Kai Chen, Sikai Chen, Bin Ran
Abstract: Cooperative autonomous driving requires traffic scene understanding from both vehicle and infrastructure perspectives. While vision-language models (VLMs) show strong general reasoning capabilities, their performance in safety-critical traffic scenarios remains insufficiently evaluated due to the ego-vehicle focus of existing benchmarks. To bridge this gap, we present \textbf{CrashSight}, a large-scale vision-language benchmark for roadway crash understanding using real-world roadside camera data. The dataset comprises 250 crash videos, annotated with 13K multiple-choice question-answer pairs organized under a two-tier taxonomy. Tier 1 evaluates the visual grounding of scene context and involved parties, while Tier 2 probes higher-level reasoning, including crash mechanics, causal attribution, temporal progression, and post-crash outcomes. We benchmark 8 state-of-the-art VLMs and show that, despite strong scene description capabilities, current models struggle with temporal and causal reasoning in safety-critical scenarios. We provide a detailed analysis of failure scenarios and discuss directions for improving VLM crash understanding. The benchmark provides a standardized evaluation framework for infrastructure-assisted perception in cooperative autonomous driving. The CrashSight benchmark, including the full dataset and code, is accessible at https://mcgrche.github.io/crashsight.
Authors: Yunsong Zhou, Hangxu Liu, Xuekun Jiang, Xing Shen, Yuanzhen Zhou, Hui Wang, Baole Fang, Yang Tian, Mulin Yu, Qiaojun Yu, Li Ma, Hengjie Li, Hanqing Wang, Jia Zeng, Jiangmiao Pang
Abstract: Robotic manipulation with deformable objects represents a data-intensive regime in embodied learning, where shape, contact, and topology co-evolve in ways that far exceed the variability of rigids. Although simulation promises relief from the cost of real-world data acquisition, prevailing sim-to-real pipelines remain rooted in rigid-body abstractions, producing mismatched geometry, fragile soft dynamics, and motion primitives poorly suited for cloth interaction. We posit that simulation fails not for being synthetic, but for being ungrounded. To address this, we introduce SIM1, a physics-aligned real-to-sim-to-real data engine that grounds simulation in the physical world. Given limited demonstrations, the system digitizes scenes into metric-consistent twins, calibrates deformable dynamics through elastic modeling, and expands behaviors via diffusion-based trajectory generation with quality filtering. This pipeline transforms sparse observations into scaled synthetic supervision with near-demonstration fidelity. Experiments show that policies trained on purely synthetic data achieve parity with real-data baselines at a 1:15 equivalence ratio, while delivering 90% zero-shot success and 50% generalization gains in real-world deployment. These results validate physics-aligned simulation as scalable supervision for deformable manipulation and a practical pathway for data-efficient policy learning.