Authors: Ningning Wang, Xavier Hu, Pai Liu, He Zhu, Yue Hou, Heyuan Huang, Shengyu Zhang, Jian Yang, Jiaheng Liu, Ge Zhang, Changwang Zhang, Jun Wang, Yuchen Eleanor Jiang, Wangchunshu Zhou
Abstract: The remarkable capabilities of Large Language Model (LLM)-driven agents have enabled sophisticated systems to tackle complex, multi-step tasks, but their escalating costs threaten scalability and accessibility. This work presents the first systematic study of the efficiency-effectiveness trade-off in modern agent systems, addressing the critical need for cost-effective designs without sacrificing performance. We investigate three key questions: (1) How much complexity do agentic tasks inherently require? (2) When do additional modules yield diminishing returns? (3) How much efficiency can be gained through the design of efficient agent frameworks? Through an empirical analysis on the GAIA benchmark, we evaluate the impact of LLM backbone selection, agent framework designs, and test-time scaling strategies. Using the cost-of-pass metric, we quantify the efficiency-performance trade-off across these dimensions. Our findings inform the development of Efficient Agents , a novel agent framework that has an optimal complexity to task requirements. Efficient Agents retains 96.7% of the performance of OWL, one leading open-source agent framework, while reducing operational costs from $0.398 to $0.228, resulting in a 28.4% improvement in cost-of-pass. Our work provides actionable insights for designing efficient, high-performing agent systems, advancing the accessibility and sustainability of AI-driven solutions.
Authors: Mikhail Soutchanski, Yongmei Liu
Abstract: In classical planning and conformant planning, it is assumed that there are finitely many named objects given in advance, and only they can participate in actions and in fluents. This is the Domain Closure Assumption (DCA). However, there are practical planning problems where the set of objects changes dynamically as actions are performed; e.g., new objects can be created, old objects can be destroyed. We formulate the planning problem in first-order logic, assume an initial theory is a finite consistent set of fluent literals, discuss when this guarantees that in every situation there are only finitely many possible actions, impose a finite integer bound on the length of the plan, and propose to organize search over sequences of actions that are grounded at planning time. We show the soundness and completeness of our approach. It can be used to solve the bounded planning problems without DCA that belong to the intersection of sequential generalized planning (without sensing actions) and conformant planning, restricted to the case without the disjunction over fluent literals. We discuss a proof-of-the-concept implementation of our planner.
Authors: Weiyu Luo, Chenfeng Xiong
Abstract: Location-Based Service (LBS) data provides critical insights into human mobility, yet its sparsity often yields incomplete trip and activity sequences, making accurate inferences about trips and activities difficult. We raise a research problem: Can we use activity sequences derived from high-quality LBS data to recover incomplete activity sequences at the individual level? This study proposes a new solution, the Variable Selection Network-fused Insertion Transformer (VSNIT), integrating the Insertion Transformer's flexible sequence construction with the Variable Selection Network's dynamic covariate handling capability, to recover missing segments in incomplete activity sequences while preserving existing data. The findings show that VSNIT inserts more diverse, realistic activity patterns, more closely matching real-world variability, and restores disrupted activity transitions more effectively aligning with the target. It also performs significantly better than the baseline model across all metrics. These results highlight VSNIT's superior accuracy and diversity in activity sequence recovery tasks, demonstrating its potential to enhance LBS data utility for mobility analysis. This approach offers a promising framework for future location-based research and applications.
Authors: Peiran Wang, Yaoning Yu, Ke Chen, Xianyang Zhan, Haohan Wang
Abstract: The rapid advancement of Large Language Models (LLMs) has driven novel applications across diverse domains, with LLM-based agents emerging as a crucial area of exploration. This survey presents a comprehensive analysis of LLM-based agents designed for data science tasks, summarizing insights from recent studies. From the agent perspective, we discuss the key design principles, covering agent roles, execution, knowledge, and reflection methods. From the data science perspective, we identify key processes for LLM-based agents, including data preprocessing, model development, evaluation, visualization, etc. Our work offers two key contributions: (1) a comprehensive review of recent developments in applying LLMbased agents to data science tasks; (2) a dual-perspective framework that connects general agent design principles with the practical workflows in data science.
Authors: Newman Cheng, Gordon Broadbent, William Chappell
Abstract: The capacity for artificial intelligence (AI) to formulate, evolve, and test altered thought patterns under dynamic conditions indicates advanced cognition that is crucial for scientific discovery. The existing AI development landscape falls into two categories: 1) frameworks over non-reasoning models that natively incorporate opinions on how humans think, and 2) reasoning models that abstract precise control of the reasoning intuition away from end users. While powerful, for scientists to maximize utility of AI in scientific discovery, they not only require accuracy and transparency in reasoning, but also steerability. Hence, we introduce an alternative approach that enables deep and precise control over the reasoning process called: a cognitive loop via in-situ optimization (CLIO). CLIO enables large language models (LLMs) to self-formulate ways of approaching a problem, adapt behavior when self-confidence is low, and ultimately provide scientists with a final belief or answer. Through CLIO's open design, scientists can observe uncertainty levels, understand how final belief states are formulated using graph structures, and interject corrections. Without any further post-training, OpenAI's GPT-4.1 with CLIO yields an accuracy of 22.37\% in text-based biology and medicine questions on Humanity's Last Exam (HLE). This yields a 13.82\% net or 161.64\% relative increase when compared to the base GPT-4.1 model and surpasses OpenAI's o3 performance in high and low reasoning effort modes. We further discovered that oscillations within internal uncertainty measures are key in determining the accuracy of CLIO's results, revealing how its open design and internal mechanisms can provide insight and control into scientific decision-making processes.
Authors: Ziruo Yi, Jinyu Liu, Ting Xiao, Mark V. Albert
Abstract: Radiology visual question answering (RVQA) provides precise answers to questions about chest X-ray images, alleviating radiologists' workload. While recent methods based on multimodal large language models (MLLMs) and retrieval-augmented generation (RAG) have shown promising progress in RVQA, they still face challenges in factual accuracy, hallucinations, and cross-modal misalignment. We introduce a multi-agent system (MAS) designed to support complex reasoning in RVQA, with specialized agents for context understanding, multimodal reasoning, and answer validation. We evaluate our system on a challenging RVQA set curated via model disagreement filtering, comprising consistently hard cases across multiple MLLMs. Extensive experiments demonstrate the superiority and effectiveness of our system over strong MLLM baselines, with a case study illustrating its reliability and interpretability. This work highlights the potential of multi-agent approaches to support explainable and trustworthy clinical AI applications that require complex reasoning.
Authors: Michael Katz, Harsha Kokel, Sarath Sreedharan
Abstract: There is a broad consensus that the inability to form long-term plans is one of the key limitations of current foundational models and agents. However, the existing planning benchmarks remain woefully inadequate to truly measure their planning capabilities. Most existing benchmarks either focus on loosely defined tasks like travel planning or end up leveraging existing domains and problems from international planning competitions. While the former tasks are hard to formalize and verify, the latter were specifically designed to test and challenge the weaknesses of existing automated planners. To address these shortcomings, we propose a procedure for creating a planning benchmark centered around the game called Countdown, where a player is expected to form a target number from a list of input numbers through arithmetic operations. We discuss how this problem meets many of the desiderata associated with an ideal benchmark for planning capabilities evaluation. Specifically, the domain allows for an intuitive, natural language description for each problem instance, it is computationally challenging (NP-complete), and the instance space is rich enough that we do not have to worry about memorization. We perform an extensive theoretical analysis, establishing the computational complexity result and demonstrate the advantage of our instance generation procedure over public benchmarks. We evaluate a variety of existing LLM-assisted planning methods on instances generated using our procedure. Our results show that, unlike other domains like 24 Game (a special case of Countdown), our proposed dynamic benchmark remains extremely challenging for existing LLM-based approaches.
Authors: Carolina Minami Oguchi, Leo Wei, Koyo Kobayashi, Hsin-Tai Wu, Dipak Ghosal
Abstract: Post-training methods have improved the performance and enhanced the reasoning capability for mainstream large language models (LLMs), but the same is challenging for Japanese LLMs to achieve due to the amount of resources required. Inspired by task vectors that extract the change of weights before and after training, specifically for a certain task, we obtain reasoning vectors from reasoning LLMs and apply them to Japanese LLMs to boost their performance. While the resources available present a challenge to improve Japanese LLMs, we present a simple and effective way to obtain high improvement and hope to inspire for other languages.
Authors: Shane Caldwell, Max Harley, Michael Kouremetis, Vincent Abruzzo, Will Pearce
Abstract: We introduce PentestJudge, a system for evaluating the operations of penetration testing agents. PentestJudge is a large language model (LLM)-as-judge with access to tools that allow it to consume arbitrary trajectories of agent states and tool call history to determine whether a security agent's actions meet certain operating criteria that would be impractical to evaluate programmatically. We develop rubrics that use a tree structure to hierarchically collapse the penetration testing task for a particular environment into smaller, simpler, and more manageable sub-tasks and criteria until each leaf node represents simple yes-or-no criteria for PentestJudge to evaluate. Task nodes are broken down into different categories related to operational objectives, operational security, and tradecraft. LLM-as-judge scores are compared to human domain experts as a ground-truth reference, allowing us to compare their relative performance with standard binary classification metrics, such as F1 scores. We evaluate several frontier and open-source models acting as judge agents, with the best model reaching an F1 score of 0.83. We find models that are better at tool-use perform more closely to human experts. By stratifying the F1 scores by requirement type, we find even models with similar overall scores struggle with different types of questions, suggesting certain models may be better judges of particular operating criteria. We find that weaker and cheaper models can judge the trajectories of pentests performed by stronger and more expensive models, suggesting verification may be easier than generation for the penetration testing task. We share this methodology to facilitate future research in understanding the ability of judges to holistically and scalably evaluate the process quality of AI-based information security agents so that they may be confidently used in sensitive production environments.
Authors: Songkun Yan, Zhi Li, Siyu Zhu, Yixin Wen, Mofan Zhang, Mengye Chen, Jie Cao, Yang Hong
Abstract: We introduce AQUAH, the first end-to-end language-based agent designed specifically for hydrologic modeling. Starting from a simple natural-language prompt (e.g., 'simulate floods for the Little Bighorn basin from 2020 to 2022'), AQUAH autonomously retrieves the required terrain, forcing, and gauge data; configures a hydrologic model; runs the simulation; and generates a self-contained PDF report. The workflow is driven by vision-enabled large language models, which interpret maps and rasters on the fly and steer key decisions such as outlet selection, parameter initialization, and uncertainty commentary. Initial experiments across a range of U.S. basins show that AQUAH can complete cold-start simulations and produce analyst-ready documentation without manual intervention. The results are judged by hydrologists as clear, transparent, and physically plausible. While further calibration and validation are still needed for operational deployment, these early outcomes highlight the promise of LLM-centered, vision-grounded agents to streamline complex environmental modeling and lower the barrier between Earth observation data, physics-based tools, and decision makers.
Authors: Mahtab Bigverdi, Wisdom Ikezogwo, Kevin Zhang, Hyewon Jeong, Mingyu Lu, Sungjae Cho, Linda Shapiro, Ranjay Krishna
Abstract: Multimodal language models (MLMs) show promise for clinical decision support and diagnostic reasoning, raising the prospect of end-to-end automated medical image interpretation. However, clinicians are highly selective in adopting AI tools; a model that makes errors on seemingly simple perception tasks such as determining image orientation or identifying whether a CT scan is contrast-enhance are unlikely to be adopted for clinical tasks. We introduce Medblink, a benchmark designed to probe these models for such perceptual abilities. Medblink spans eight clinically meaningful tasks across multiple imaging modalities and anatomical regions, totaling 1,429 multiple-choice questions over 1,605 images. We evaluate 19 state-of-the-art MLMs, including general purpose (GPT4o, Claude 3.5 Sonnet) and domain specific (Med Flamingo, LLaVA Med, RadFM) models. While human annotators achieve 96.4% accuracy, the best-performing model reaches only 65%. These results show that current MLMs frequently fail at routine perceptual checks, suggesting the need to strengthen their visual grounding to support clinical adoption. Data is available on our project page.
Authors: Chia-Tung Ho, Jing Gong, Xufeng Yao, Yunsheng Bai, Abhishek B Akkur, Haoxing Ren
Abstract: Large language models (LLMs) excel at solving complex tasks by executing agentic workflows composed of detailed instructions and structured operations. Yet, building general-purpose agents by manually embedding foundation models into agentic systems such as Chain-of-Thought, Self-Reflection, and ReACT through text interfaces limits scalability and efficiency. Recently, many researchers have sought to automate the generation and optimization of these workflows through code-based representations. However, existing methods often rely on labeled datasets to train and optimize workflows, making them ineffective and inflexible for solving real-world, dynamic problems where labeled data is unavailable. To address this challenge, we introduce Polymath, a self-optimizing agent with dynamic hierarchical workflow that leverages the flexibility of task flow graphs and the expressiveness of code-represented workflows to solve a wide range of real-world, dynamic problems. The proposed optimization methodology integrates multi-grid-inspired graph optimization with a self-reflection-guided evolutionary algorithm to refine workflows without labeled data. Experimental results on six benchmark datasets across coding, math, and multi-turn QA tasks show that Polymath achieves 8.1% average improvement over state-of-the-art baselines.
Authors: Boshi Huang, Fabio Nonato de Paula
Abstract: This paper introduces a novel self-consciousness defense mechanism for Large Language Models (LLMs) to combat prompt injection attacks. Unlike traditional approaches that rely on external classifiers, our method leverages the LLM's inherent reasoning capabilities to perform self-protection. We propose a framework that incorporates Meta-Cognitive and Arbitration Modules, enabling LLMs to evaluate and regulate their own outputs autonomously. Our approach is evaluated on seven state-of-the-art LLMs using two datasets: AdvBench and Prompt-Injection-Mixed-Techniques-2024. Experiment results demonstrate significant improvements in defense success rates across models and datasets, with some achieving perfect and near-perfect defense in Enhanced Mode. We also analyze the trade-off between defense success rate improvement and computational overhead. This self-consciousness method offers a lightweight, cost-effective solution for enhancing LLM ethics, particularly beneficial for GenAI use cases across various platforms.
Authors: Peng Ding, Rick Stevens
Abstract: The proliferation of tool-augmented Large Language Models (LLMs) has created a fragmented ecosystem where developers must navigate multiple protocols, manual schema definitions, and complex execution workflows. We address this challenge by proposing a unified approach to tool integration that abstracts protocol differences while optimizing execution performance. Our solution demonstrates how protocol-agnostic design principles can significantly reduce development overhead through automated schema generation, dual-mode concurrent execution, and seamless multi-source tool management. Experimental results show 60-80% code reduction across integration scenarios, performance improvements up to 3.1x through optimized concurrency, and full compatibility with existing function calling standards. This work contributes both theoretical insights into tool integration architecture and practical solutions for real-world LLM application development.
Authors: Fangyi Yu
Abstract: As large language models (LLMs) grow in capability and autonomy, evaluating their outputs-especially in open-ended and complex tasks-has become a critical bottleneck. A new paradigm is emerging: using AI agents as the evaluators themselves. This "agent-as-a-judge" approach leverages the reasoning and perspective-taking abilities of LLMs to assess the quality and safety of other models, promising calable and nuanced alternatives to human evaluation. In this review, we define the agent-as-a-judge concept, trace its evolution from single-model judges to dynamic multi-agent debate frameworks, and critically examine their strengths and shortcomings. We compare these approaches across reliability, cost, and human alignment, and survey real-world deployments in domains such as medicine, law, finance, and education. Finally, we highlight pressing challenges-including bias, robustness, and meta evaluation-and outline future research directions. By bringing together these strands, our review demonstrates how agent-based judging can complement (but not replace) human oversight, marking a step toward trustworthy, scalable evaluation for next-generation LLMs.
Authors: Xinjie Zhao, Moritz Blum, Fan Gao, Yingjian Chen, Boming Yang, Luis Marquez-Carpintero, M\'onica Pina-Navarro, Yanran Fu, So Morikawa, Yusuke Iwasawa, Yutaka Matsuo, Chanjun Park, Irene Li
Abstract: AGENTiGraph is a user-friendly, agent-driven system that enables intuitive interaction and management of domain-specific data through the manipulation of knowledge graphs in natural language. It gives non-technical users a complete, visual solution to incrementally build and refine their knowledge bases, allowing multi-round dialogues and dynamic updates without specialized query languages. The flexible design of AGENTiGraph, including intent classification, task planning, and automatic knowledge integration, ensures seamless reasoning between diverse tasks. Evaluated on a 3,500-query benchmark within an educational scenario, the system outperforms strong zero-shot baselines (achieving 95.12% classification accuracy, 90.45% execution success), indicating potential scalability to compliance-critical or multi-step queries in legal and medical domains, e.g., incorporating new statutes or research on the fly. Our open-source demo offers a powerful new paradigm for multi-turn enterprise knowledge management that bridges LLMs and structured graphs.
Authors: Yutong Wang, Pengliang Ji, Kaixin Li, Baolong Bi, Tao Feng, Guillaume Sartoretti
Abstract: Large Language Reasoning Models have demonstrated remarkable success on static tasks, yet their application to multi-round agentic planning in interactive environments faces two fundamental challenges. First, the intractable credit assignment problem renders conventional reinforcement learning ineffective in sparse-reward settings. Second, the computational overhead of verbose, step-by-step reasoning histories is prohibitive. To address these challenges, we propose BPO, a three-stage framework (bootstrapping, extrapolation, and refinement) that establishes a self-improving data flywheel to develop robust reasoning models for long-horizon, sparse-reward environments. Our framework first bootstraps efficient reasoning using the proposed planning quaternions with long-short chain-of-thought fusion. It then extrapolates to out-of-distribution tasks through complexity-stratified curriculum learning. Finally, the model iteratively refines itself by learning exclusively on experiences selected via reward-gated rejection sampling. Experiments on ALFWorld, ScienceWorld, and WebShop demonstrate that our approach achieves state-of-the-art with significant token efficiency, providing a new recipe for reasoning models in agentic planning.
Authors: Siyuan Li, Yifan Yu, Yanchen Deng, Zhihao Zhang, Mengjing Chen, Fangzhou Zhu, Tao Zhong, Jianye Hao, Peng Liu, Bo An
Abstract: Mixed-integer linear programming (MILP) has been a fundamental problem in combinatorial optimization. Previous works have designed a plethora of hard-coded heuristics to accomplish challenging MILP solving with domain knowledge. Driven by the high capability of neural networks, recent research is devoted to replacing manually designed heuristics with learned policies. Although learning-based MILP methods have shown great promise, existing worksindependentlytreatthepolicylearningineachmoduleofMILPsolvers without considering their interdependence, severely hurting the solving speed and quality. To address this issue, we propose a novel multi-agent-based policy learning framework for MILP (Collab-Solver), which can collaboratively optimize the policies for multiple modules. Specifically, we formulate the collaboration of cut selection and branching in MILP solving as a Stackelberg game. Under this formulation, we develop a two-phase learning paradigm to stabilize the collaborative policy learning, where the first phase achieves the data-communicated policy pretraining and the second phase further orchestrates the policy learning for various modules. The jointly learned policy significantly improves the solving performance on both synthetic and large-scale real-world MILP datasets. Moreover, the policies learned by Collab-Solver have also demonstrated excellent generalization abilities across different instance sets.
Authors: Ziyang Ma, Baojian Zhou, Deqing Yang, Yanghua Xiao
Abstract: In-Context Learning (ICL) has emerged as a new paradigm in large language models (LLMs), enabling them to perform novel tasks by conditioning on a few examples embedded in the prompt. Yet, the highly nonlinear behavior of ICL for NLP tasks remains poorly understood. To shed light on its underlying mechanisms, this paper investigates whether LLMs can solve ordinary differential equations (ODEs) under the ICL setting. We formulate standard ODE problems and their solutions as sequential prompts and evaluate GPT-2 models on these tasks. Experiments on two types of ODEs show that GPT-2 can effectively learn a meta-ODE algorithm, with convergence behavior comparable to, or better than, the Euler method, and achieve exponential accuracy gains with increasing numbers of demonstrations. Moreover, the model generalizes to out-of-distribution (OOD) problems, demonstrating robust extrapolation capabilities. These empirical findings provide new insights into the mechanisms of ICL in NLP and its potential for solving nonlinear numerical problems.
Authors: Qi Peng, Jialin Cui, Jiayuan Xie, Yi Cai, Qing Li
Abstract: Large language models (LLMs) have shown great potential in the medical domain. However, existing models still fall short when faced with complex medical diagnosis task in the real world. This is mainly because they lack sufficient reasoning depth, which leads to information loss or logical jumps when processing a large amount of specialized medical data, leading to diagnostic errors. To address these challenges, we propose Tree-of-Reasoning (ToR), a novel multi-agent framework designed to handle complex scenarios. Specifically, ToR introduces a tree structure that can clearly record the reasoning path of LLMs and the corresponding clinical evidence. At the same time, we propose a cross-validation mechanism to ensure the consistency of multi-agent decision-making, thereby improving the clinical reasoning ability of multi-agents in complex medical scenarios. Experimental results on real-world medical data show that our framework can achieve better performance than existing baseline methods.
Authors: Rui Pu, Chaozhuo Li, Rui Ha, Litian Zhang, Lirong Qiu, Xi Zhang
Abstract: Defending large language models (LLMs) against jailbreak attacks is essential for their safe and reliable deployment. Existing defenses often rely on shallow pattern matching, which struggles to generalize to novel and unseen attack strategies. To address this challenge, we propose the Cognitive-Driven Defense (CDD) framework, which targets the underlying structure of jailbreak prompts by applying meta-operations, defined as basic manipulations that conceal harmful intent.CDD emulates human cognitive reasoning through a structured reasoning chain. It begins with a global perception of the prompt and follows with a localized analysis to uncover hidden manipulations. By applying supervised fine-tuning on this structured chain, the model learns to identify and reason about known manipulation patterns. To enhance generalization to unseen threats, an entropy-guided reinforcement learning algorithm (EG-GRPO) is introduced to encourage exploration of new types and variants of meta-operations. Experiments demonstrate that CDD can achieve state-of-the-art defense performance and exhibit strong generalization to unseen jailbreak attacks.
Authors: Shuang Liu, Zelong Li, Ruoyun Ma, Haiyan Zhao, Mengnan Du
Abstract: The potential of large language models (LLMs) in specialized domains such as legal risk analysis remains underexplored. In response to growing interest in locally deploying open-source LLMs for legal tasks while preserving data confidentiality, this paper introduces ContractEval, the first benchmark to thoroughly evaluate whether open-source LLMs could match proprietary LLMs in identifying clause-level legal risks in commercial contracts. Using the Contract Understanding Atticus Dataset (CUAD), we assess 4 proprietary and 15 open-source LLMs. Our results highlight five key findings: (1) Proprietary models outperform open-source models in both correctness and output effectiveness, though some open-source models are competitive in certain specific dimensions. (2) Larger open-source models generally perform better, though the improvement slows down as models get bigger. (3) Reasoning ("thinking") mode improves output effectiveness but reduces correctness, likely due to over-complicating simpler tasks. (4) Open-source models generate "no related clause" responses more frequently even when relevant clauses are present. This suggests "laziness" in thinking or low confidence in extracting relevant content. (5) Model quantization speeds up inference but at the cost of performance drop, showing the tradeoff between efficiency and accuracy. These findings suggest that while most LLMs perform at a level comparable to junior legal assistants, open-source models require targeted fine-tuning to ensure correctness and effectiveness in high-stakes legal settings. ContractEval offers a solid benchmark to guide future development of legal-domain LLMs.
Authors: Fei Liu, Yilu Liu, Qingfu Zhang, Xialiang Tong, Mingxuan Yuan
Abstract: Automated Heuristic Design (AHD) using Large Language Models (LLMs) has achieved notable success in recent years. Despite the effectiveness of existing approaches, they only design a single heuristic to serve all problem instances, often inducing poor generalization across different distributions or settings. To address this issue, we propose Automated Heuristic Set Design (AHSD), a new formulation for LLM-driven AHD. The aim of AHSD is to automatically generate a small-sized complementary heuristic set to serve diverse problem instances, such that each problem instance could be optimized by at least one heuristic in this set. We show that the objective function of AHSD is monotone and supermodular. Then, we propose Evolution of Heuristic Set (EoH-S) to apply the AHSD formulation for LLM-driven AHD. With two novel mechanisms of complementary population management and complementary-aware memetic search, EoH-S could effectively generate a set of high-quality and complementary heuristics. Comprehensive experimental results on three AHD tasks with diverse instances spanning various sizes and distributions demonstrate that EoH-S consistently outperforms existing state-of-the-art AHD methods and achieves up to 60\% performance improvements.
Authors: Youran Zhou, Mohamed Reda Bouadjenek, Sunil Aryal
Abstract: Diffusion models have recently emerged as powerful tools for missing data imputation by modeling the joint distribution of observed and unobserved variables. However, existing methods, typically based on stochastic denoising diffusion probabilistic models (DDPMs), suffer from high inference latency and variable outputs, limiting their applicability in real-world tabular settings. To address these deficiencies, we present in this paper MissDDIM, a conditional diffusion framework that adapts Denoising Diffusion Implicit Models (DDIM) for tabular imputation. While stochastic sampling enables diverse completions, it also introduces output variability that complicates downstream processing.
Authors: Xingjun Ma, Hanxun Huang, Tianwei Song, Ye Sun, Yifeng Gao, Yu-Gang Jiang
Abstract: Large-scale pre-training frameworks like CLIP have revolutionized multimodal learning, but their reliance on web-scraped datasets, frequently containing private user data, raises serious concerns about misuse. Unlearnable Examples (UEs) have emerged as a promising countermeasure against unauthorized model training, employing carefully crafted unlearnable noise to disrupt the learning of meaningful representations from protected data. Current approaches typically generate UEs by jointly optimizing unlearnable noise for both images and their associated text descriptions (or labels). However, this optimization process is often computationally prohibitive for on-device execution, forcing reliance on external third-party services. This creates a fundamental privacy paradox: users must initially expose their data to these very services to achieve protection, thereby compromising privacy in the process. Such a contradiction has severely hindered the development of practical, scalable data protection solutions. To resolve this paradox, we introduce \textbf{Text-to-Unlearnable Example (T2UE)}, a novel framework that enables users to generate UEs using only text descriptions. T2UE circumvents the need for original image data by employing a text-to-image (T2I) model to map text descriptions into the image (noise) space, combined with an error-minimization framework to produce effective unlearnable noise. Extensive experiments show that T2UE-protected data substantially degrades performance in downstream tasks (e.g., cross-modal retrieval) for state-of-the-art models. Notably, the protective effect generalizes across diverse architectures and even to supervised learning settings. Our work demonstrates the feasibility of "zero-contact data protection", where personal data can be safeguarded based solely on their textual descriptions, eliminating the need for direct data exposure.
Authors: Zikun Cui, Tianyi Huang, Chia-En Chiang, Cuiqianhe Du
Abstract: With the proliferation of Large Language Models (LLMs), the detection of misinformation has become increasingly important and complex. This research proposes an innovative verifiable misinformation detection LLM agent that goes beyond traditional true/false binary judgments. The agent actively verifies claims through dynamic interaction with diverse web sources, assesses information source credibility, synthesizes evidence, and provides a complete verifiable reasoning process. Our designed agent architecture includes three core tools: precise web search tool, source credibility assessment tool and numerical claim verification tool. These tools enable the agent to execute multi-step verification strategies, maintain evidence logs, and form comprehensive assessment conclusions. We evaluate using standard misinformation datasets such as FakeNewsNet, comparing with traditional machine learning models and LLMs. Evaluation metrics include standard classification metrics, quality assessment of reasoning processes, and robustness testing against rewritten content. Experimental results show that our agent outperforms baseline methods in misinformation detection accuracy, reasoning transparency, and resistance to information rewriting, providing a new paradigm for trustworthy AI-assisted fact-checking.
Authors: Wen-Xi Yang, Tian-Fang Zhao
Abstract: Generative agent models specifically tailored for smart education are critical, yet remain relatively underdeveloped. A key challenge stems from the inherent complexity of educational contexts: learners are human beings with various cognitive behaviors, and pedagogy is fundamentally centered on personalized human-to-human communication. To address this issue, this paper proposes AgentSME, a unified generative agent framework powered by LLM. Three directional communication modes are considered in the models, namely Solo, Mono, and Echo, reflecting different types of agency autonomy and communicative reciprocity. Accuracy is adopted as the primary evaluation metric, complemented by three diversity indices designed to assess the diversity of reasoning contents. Six widely used LLMs are tested to validate the robustness of communication modes across different model tiers, which are equally divided into base-capacity and high-capacity configurations. The results show that generative agents that employ the Echo communication mode achieve the highest accuracy scores, while DeepSeek exhibits the greatest diversity. This study provides valuable information to improve agent learning capabilities and inspire smart education models.
Authors: Vinicius Lima, Dzung T. Phan, Jayant Kalagnanam, Dhaval Patel, Nianjun Zhou
Abstract: We present a framework for training trustworthy large language model (LLM) agents for optimization modeling via a verifiable synthetic data generation pipeline. Focusing on linear and mixed-integer linear programming, our approach begins with structured symbolic representations and systematically produces natural language descriptions, mathematical formulations, and solver-executable code. By programmatically constructing each instance with known optimal solutions, the pipeline ensures full verifiability and enables automatic filtering of low-quality demonstrations generated by teacher models. Each dataset instance includes a structured representation of the optimization problem, a corresponding natural language description, the verified optimal solution, and step-by-step demonstrations - generated by a teacher model - that show how to model and solve the problem across multiple optimization modeling languages. This enables supervised fine-tuning of open-source LLMs specifically tailored to optimization tasks. To operationalize this pipeline, we introduce OptiTrust, a modular LLM agent that performs multi-stage translation from natural language to solver-ready code, leveraging stepwise demonstrations, multi-language inference, and majority-vote cross-validation. Our agent achieves state-of-the-art performance on standard benchmarks. Out of 7 datasets, it achieves the highest accuracy on six and outperforms the next-best algorithm by at least 8 percentage on three of them. Our approach provides a scalable, verifiable, and principled path toward building reliable LLM agents for real-world optimization applications.
Authors: Linda Smail (College of Interdisciplinary Studies, Zayed University, UAE), David Santandreu Calonge (Department of Academic Development, Mohamed bin Zayed University of Artificial Intelligence, UAE), Firuz Kamalov (School of Engineering, Applied Science,Technology, Canadian University Dubai, UAE), Nur H. Orak (Department of Environmental Engineering, Marmara University, T\"urkiye)
Abstract: This research investigates the potential of Artificial Intelligence (AI) models to bridge the knowledge gap in environmental education among university students. By focusing on prominent large language models (LLMs) such as GPT-3.5, GPT-4, GPT-4o, Gemini, Claude Sonnet, and Llama 2, the study assesses their effectiveness in conveying environmental concepts and, consequently, facilitating environmental education. The investigation employs a standardized tool, the Environmental Knowledge Test (EKT-19), supplemented by targeted questions, to evaluate the environmental knowledge of university students in comparison to the responses generated by the AI models. The results of this study suggest that while AI models possess a vast, readily accessible, and valid knowledge base with the potential to empower both students and academic staff, a human discipline specialist in environmental sciences may still be necessary to validate the accuracy of the information provided.
Authors: Charles Tapley Hoyt, Craig Bakker, Richard J. Callahan, Joseph Cottam, August George, Benjamin M. Gyori, Haley M. Hummel, Nathaniel Merrill, Sara Mohammad Taheri, Pruthvi Prakash Navada, Marc-Antoine Parent, Adam Rupe, Olga Vitek, Jeremy Zucker
Abstract: We present the $Y_0$ Python package, which implements causal identification algorithms that apply interventional, counterfactual, and transportability queries to data from (randomized) controlled trials, observational studies, or mixtures thereof. $Y_0$ focuses on the qualitative investigation of causation, helping researchers determine whether a causal relationship can be estimated from available data before attempting to estimate how strong that relationship is. Furthermore, $Y_0$ provides guidance on how to transform the causal query into a symbolic estimand that can be non-parametrically estimated from the available data. $Y_0$ provides a domain-specific language for representing causal queries and estimands as symbolic probabilistic expressions, tools for representing causal graphical models with unobserved confounders, such as acyclic directed mixed graphs (ADMGs), and implementations of numerous identification algorithms from the recent causal inference literature. The $Y_0$ source code can be found under the MIT License at https://github.com/y0-causal-inference/y0 and it can be installed with pip install y0.
Authors: Jingxuan Wei, Caijun Jia, Qi Chen, Honghao He, Linzhuang Sun, Conghui He, Lijun Wu, Bihui Yu, Cheng Tan
Abstract: Mathematical geometric reasoning is essential for scientific discovery and educational development, requiring precise logic and rigorous formal verification. While recent advances in Multimodal Large Language Models (MLLMs) have improved reasoning tasks, existing models typically struggle with formal geometric reasoning, particularly when dynamically constructing and verifying auxiliary geometric elements. To address these challenges, we introduce Geoint-R1, a multimodal reasoning framework designed to generate formally verifiable geometric solutions from textual descriptions and visual diagrams. Geoint-R1 uniquely integrates auxiliary elements construction, formal reasoning represented via Lean4, and interactive visualization. To systematically evaluate and advance formal geometric reasoning, we propose the Geoint benchmark, comprising 1,885 rigorously annotated geometry problems across diverse topics such as plane, spatial, and solid geometry. Each problem includes structured textual annotations, precise Lean4 code for auxiliary constructions, and detailed solution steps verified by experts. Extensive experiments demonstrate that Geoint-R1 significantly surpasses existing multimodal and math-specific reasoning models, particularly on challenging problems requiring explicit auxiliary element constructions.
Authors: Tian-Fang Zhao, Wen-Xi Yang, Guan Liu, Liang Yang
Abstract: Collaborative partnership matters in inquiry-oriented education. However, most study partners are selected either rely on experience-based assignments with little scientific planning or build on rule-based machine assistants, encountering difficulties in knowledge expansion and inadequate flexibility. This paper proposes an LLM-empowered agent model for simulating and selecting learning partners tailored to inquiry-oriented learning, named InqEduAgent. Generative agents are designed to capture cognitive and evaluative features of learners in real-world scenarios. Then, an adaptive matching algorithm with Gaussian process augmentation is formulated to identify patterns within prior knowledge. Optimal learning-partner matches are provided for learners facing different exercises. The experimental results show the optimal performance of InqEduAgent in most knowledge-learning scenarios and LLM environment with different levels of capabilities. This study promotes the intelligent allocation of human-based learning partners and the formulation of AI-based learning partners. The code, data, and appendix are publicly available at https://github.com/InqEduAgent/InqEduAgent.
Authors: Osama Mohammed, Jiaxin Pan, Mojtaba Nayyeri, Daniel Hern\'andez, Steffen Staab
Abstract: Modeling evolving interactions among entities is critical in many real-world tasks. For example, predicting driver maneuvers in traffic requires tracking how neighboring vehicles accelerate, brake, and change lanes relative to one another over consecutive frames. Likewise, detecting financial fraud hinges on following the flow of funds through successive transactions as they propagate through the network. Unlike classic time-series forecasting, these settings demand reasoning over who interacts with whom and when, calling for a temporal-graph representation that makes both the relations and their evolution explicit. Existing temporal-graph methods typically use snapshot graphs to encode temporal evolution. We introduce a full-history graph that instantiates one node for every entity at every time step and separates two edge sets: (i) intra-time-step edges that capture relations within a single frame and (ii) inter-time-step edges that connect an entity to itself at consecutive steps. To learn on this graph we design an Edge-Type Decoupled Network (ETDNet) with parallel modules: a graph-attention module aggregates information along intra-time-step edges, a multi-head temporal-attention module attends over an entity's inter-time-step history, and a fusion module combines the two messages after every layer. Evaluated on driver-intention prediction (Waymo) and Bitcoin fraud detection (Elliptic++), ETDNet consistently surpasses strong baselines, lifting Waymo joint accuracy to 75.6\% (vs. 74.1\%) and raising Elliptic++ illicit-class F1 to 88.1\% (vs. 60.4\%). These gains demonstrate the benefit of representing structural and temporal relations as distinct edges in a single graph.
Authors: Shaofeng Yin, Ting Lei, Yang Liu
Abstract: Integrating external tools into Large Foundation Models (LFMs) has emerged as a promising approach to enhance their problem-solving capabilities. While existing studies have demonstrated strong performance in tool-augmented Visual Question Answering (VQA), recent benchmarks reveal significant gaps in real-world tool-use proficiency, particularly in functionally diverse multimodal settings requiring multi-step reasoning. In this work, we introduce ToolVQA, a large-scale multimodal dataset comprising 23K instances, designed to bridge this gap. Unlike previous datasets that rely on synthetic scenarios and simplified queries, ToolVQA features real-world visual contexts and challenging implicit multi-step reasoning tasks, better aligning with real user interactions. To construct this dataset, we propose ToolEngine, a novel data generation pipeline that employs Depth-First Search (DFS) with a dynamic in-context example matching mechanism to simulate human-like tool-use reasoning. ToolVQA encompasses 10 multimodal tools across 7 diverse task domains, with an average inference length of 2.78 reasoning steps per instance. The fine-tuned 7B LFMs on ToolVQA not only achieve impressive performance on our test set but also surpass the large close-sourced model GPT-3.5-turbo on various out-of-distribution (OOD) datasets, demonstrating strong generalizability to real-world tool-use scenarios.
Authors: Jiayan Nan, Wenquan Ma, Wenlong Wu, Yize Chen
Abstract: Large Language Models (LLMs) demonstrate remarkable capabilities, yet their inability to maintain persistent memory in long contexts limits their effectiveness as autonomous agents in long-term interactions. While existing memory systems have made progress, their reliance on arbitrary granularity for defining the basic memory unit and passive, rule-based mechanisms for knowledge extraction limits their capacity for genuine learning and evolution. To address these foundational limitations, we present Nemori, a novel self-organizing memory architecture inspired by human cognitive principles. Nemori's core innovation is twofold: First, its Two-Step Alignment Principle, inspired by Event Segmentation Theory, provides a principled, top-down method for autonomously organizing the raw conversational stream into semantically coherent episodes, solving the critical issue of memory granularity. Second, its Predict-Calibrate Principle, inspired by the Free-energy Principle, enables the agent to proactively learn from prediction gaps, moving beyond pre-defined heuristics to achieve adaptive knowledge evolution. This offers a viable path toward handling the long-term, dynamic workflows of autonomous agents. Extensive experiments on the LoCoMo and LongMemEval benchmarks demonstrate that Nemori significantly outperforms prior state-of-the-art systems, with its advantage being particularly pronounced in longer contexts.
Authors: Xingdan Wang, Jiayi He, Zhiqing Tang, Jianxiong Guo, Jiong Lou, Liping Qian, Tian Wang, Weijia Jia
Abstract: The rise of LLMs such as ChatGPT and Claude fuels the need for AI agents capable of real-time task handling. However, migrating data-intensive, multi-modal edge workloads to cloud data centers, traditionally used for agent deployment, introduces significant latency. Deploying AI agents at the edge improves efficiency and reduces latency. However, edge environments present challenges due to limited and heterogeneous resources. Maintaining QoS for mobile users necessitates agent migration, which is complicated by the complexity of AI agents coordinating LLMs, task planning, memory, and external tools. This paper presents the first systematic deployment and management solution for LLM-based AI agents in dynamic edge environments. We propose a novel adaptive framework for AI agent placement and migration in edge intelligence systems. Our approach models resource constraints and latency/cost, leveraging ant colony algorithms and LLM-based optimization for efficient decision-making. It autonomously places agents to optimize resource utilization and QoS and enables lightweight agent migration by transferring only essential state. Implemented on a distributed system using AgentScope and validated across globally distributed edge servers, our solution significantly reduces deployment latency and migration costs.
Authors: Zeju Li, Jianyuan Zhong, Ziyang Zheng, Xiangyu Wen, Zhijian Xu, Yingying Cheng, Fan Zhang, Qiang Xu
Abstract: Large Language Models (LLMs) using Chain-of-Thought (CoT) prompting excel at complex reasoning but generate verbose thought processes with considerable redundancy, leading to increased inference costs and reduced efficiency. We introduce a novel CoT compression framework based on step entropy, a metric that quantifies the informational contribution of individual reasoning steps to identify redundancy. Through theoretical analysis and extensive empirical validation on mathematical reasoning benchmarks, we demonstrate that steps with low entropy are indeed highly redundant. Our experiments reveal that an astonishing 80\% of low-entropy intermediate steps can be pruned with minor degradation in the final answer accuracy across DeepSeek-R1-7B, 14B and Qwen3-8B. This finding sharply contrasts with random or high-entropy pruning, which severely impairs reasoning performance. Building on this, we propose a novel two-stage training strategy combining Supervised Fine-Tuning (SFT) and Group Relative Policy Optimization (GRPO) reinforcement learning. This approach enables LLMs to autonomously learn to generate compressed COTs during inference by strategically incorporating [SKIP] tokens. Our method significantly enhances LLM inference efficiency while rigorously preserving accuracy, offering profound implications for practical LLM deployment and a deeper understanding of reasoning structures.
Authors: Feng Rui, Zhiyao Luo, Wei Wang, Yuting Song, Yong Liu, Tingting Zhu, Jianqing Li, Xingyao Wang
Abstract: Automatic assessment of cognitive impairment from spontaneous speech offers a promising, non-invasive avenue for early cognitive screening. However, current approaches often lack generalizability when deployed across different languages and clinical settings, limiting their practical utility. In this study, we propose CogBench, the first benchmark designed to evaluate the cross-lingual and cross-site generalizability of large language models (LLMs) for speech-based cognitive impairment assessment. Using a unified multimodal pipeline, we evaluate model performance on three speech datasets spanning English and Mandarin: ADReSSo, NCMMSC2021-AD, and a newly collected test set, CIR-E. Our results show that conventional deep learning models degrade substantially when transferred across domains. In contrast, LLMs equipped with chain-of-thought prompting demonstrate better adaptability, though their performance remains sensitive to prompt design. Furthermore, we explore lightweight fine-tuning of LLMs via Low-Rank Adaptation (LoRA), which significantly improves generalization in target domains. These findings offer a critical step toward building clinically useful and linguistically robust speech-based cognitive assessment tools.
Authors: Michael K. Chen
Abstract: General logical reasoning, defined as the ability to reason deductively on domain-agnostic tasks, continues to be a challenge for large language models (LLMs). Current LLMs fail to reason deterministically and are not interpretable. As such, there has been a recent surge in interest in neurosymbolic AI, which attempts to incorporate logic into neural networks. We first identify two main neurosymbolic approaches to improving logical reasoning: (i) the integrative approach comprising models where symbolic reasoning is contained within the neural network, and (ii) the hybrid approach comprising models where a symbolic solver, separate from the neural network, performs symbolic reasoning. Both contain AI systems with promising results on domain-specific logical reasoning benchmarks. However, their performance on domain-agnostic benchmarks is understudied. To the best of our knowledge, there has not been a comparison of the contrasting approaches that answers the following question: Which approach is more promising for developing general logical reasoning? To analyze their potential, the following best-in-class domain-agnostic models are introduced: Logic Neural Network (LNN), which uses the integrative approach, and LLM-Symbolic Solver (LLM-SS), which uses the hybrid approach. Using both models as case studies and representatives of each approach, our analysis demonstrates that the hybrid approach is more promising for developing general logical reasoning because (i) its reasoning chain is more interpretable, and (ii) it retains the capabilities and advantages of existing LLMs. To support future works using the hybrid approach, we propose a generalizable framework based on LLM-SS that is modular by design, model-agnostic, domain-agnostic, and requires little to no human input.
Authors: Lucia Cipolina-Kun, Marianna Nezhurina, Jenia Jitsev
Abstract: The Board Game Arena library provides a framework for evaluating the decision making abilities of large language models (LLMs) through strategic board games implemented in Google OpenSpiel library. The framework enables systematic comparisons between LLM based agents and other agents (random, human, reinforcement learning agents, etc.) in various game scenarios by wrapping multiple board and matrix games and supporting different agent types. It integrates API access to models via LiteLLM, local model deployment via vLLM, and offers distributed execution through Ray. Additionally it provides extensive analysis tools for the LLM reasoning traces. This paper summarizes the structure, key characteristics, and motivation of the repository, highlighting how it contributes to the empirical evaluation of the reasoning of LLM and game-theoretic behavior
Authors: Wenxin Mao, Zhitao Wang, Long Wang, Sirong Chen, Cuiyun Gao, Luyang Cao, Ziming Liu, Qiming Zhang, Jun Zhou, Zhi Jin
Abstract: Large language models (LLMs) excel at generating code from natural language (NL) descriptions. However, the plain textual descriptions are inherently ambiguous and often fail to capture complex requirements like intricate system behaviors, conditional logic, and architectural constraints; implicit data dependencies in service-oriented architectures are difficult to infer and handle correctly. To bridge this gap, we propose a novel step-by-step code generation framework named UML2Dep by leveraging unambiguous formal specifications of complex requirements. First, we introduce an enhanced Unified Modeling Language (UML) sequence diagram tailored for service-oriented architectures. This diagram extends traditional visual syntax by integrating decision tables and API specifications, explicitly formalizing structural relationships and business logic flows in service interactions to rigorously eliminate linguistic ambiguity. Second, recognizing the critical role of data flow, we introduce a dedicated data dependency inference (DDI) task. DDI systematically constructs an explicit data dependency graph prior to actual code synthesis. To ensure reliability, we formalize DDI as a constrained mathematical reasoning task through novel prompting strategies, aligning with LLMs' excellent mathematical strengths. Additional static parsing and dependency pruning further reduce context complexity and cognitive load associated with intricate specifications, thereby enhancing reasoning accuracy and efficiency.
Authors: Rui Zou, Mengqi Wei, Yutao Zhu, Jirong Wen, Xin Zhao, Jing Chen
Abstract: Large Language Models (LLMs) excel in reasoning and generation across domains, but still struggle with identifying and diagnosing complex errors. This stems mainly from training objectives that prioritize correct answers, limiting exposure to and learning from errors. While recent studies have begun to address this by introducing error signals, most rely on shallow, static errors, restricting improvement in deep diagnostic ability. To overcome this, we propose Hide and Seek Game (HSG), a dynamic adversarial framework for error generation and diagnosis, and evaluate it on mathematical problem-solving. HSG involves two adversarial roles: Sneaky, which "hides" by generating subtle, deceptive reasoning errors, and Diagnosis, which "seeks" to accurately detect them. Through adversarial co-evolution, both error stealth and diagnostic precision are enhanced. Experiments on several math reasoning tasks show that HSG significantly boosts error diagnosis, achieving 16.8\%--31.4\% higher accuracy than baselines like GPT-4o. We also release a challenging dataset of deceptive errors and diagnostic annotations as a benchmark for future research.
Authors: Kai Li, Ruihao Zheng, Xinye Hao, Zhenkun Wang
Abstract: In real-world routing problems, users often propose conflicting or unreasonable requirements, which result in infeasible optimization models due to overly restrictive or contradictory constraints, leading to an empty feasible solution set. Existing Large Language Model (LLM)-based methods attempt to diagnose infeasible models, but modifying such models often involves multiple potential adjustments that these methods do not consider. To fill this gap, we introduce Multi-Objective Infeasibility Diagnosis (MOID), which combines LLM agents and multi-objective optimization within an automatic routing solver, to provide a set of representative actionable suggestions. Specifically, MOID employs multi-objective optimization to consider both path cost and constraint violation, generating a set of trade-off solutions, each encompassing varying degrees of model adjustments. To extract practical insights from these solutions, MOID utilizes LLM agents to generate a solution analysis function for the infeasible model. This function analyzes these distinct solutions to diagnose the original infeasible model, providing users with diverse diagnostic insights and suggestions. Finally, we compare MOID with several LLM-based methods on 50 types of infeasible routing problems. The results indicate that MOID automatically generates multiple diagnostic suggestions in a single run, providing more practical insights for restoring model feasibility and decision-making compared to existing methods.
Authors: Taine J. Elliott, Stephen P. Levitt, Ken Nixon, Martin Bekker
Abstract: The rapid expansion of publicly-available medical data presents a challenge for clinicians and researchers alike, increasing the gap between the volume of scientific literature and its applications. The steady growth of studies and findings overwhelms medical professionals at large, hindering their ability to systematically review and understand the latest knowledge. This paper presents an approach to information extraction and automatic knowledge graph (KG) generation to identify and connect biomedical knowledge. Through a pipeline of large language model (LLM) agents, the system decomposes 44 PubMed abstracts into semantically meaningful proposition sentences and extracts KG triples from these sentences. The triples are enhanced using a combination of open domain and ontology-based information extraction methodologies to incorporate ontological categories. On top of this, a context variable is included during extraction to allow the triple to stand on its own - thereby becoming `quadruples'. The extraction accuracy of the LLM is validated by comparing natural language sentences generated from the enhanced triples to the original propositions, achieving an average cosine similarity of 0.874. The similarity for generated sentences of enhanced triples were compared with generated sentences of ordinary triples showing an increase as a result of the context variable. Furthermore, this research explores the ability for LLMs to infer new relationships and connect clusters in the knowledge base of the knowledge graph. This approach leads the way to provide medical practitioners with a centralised, updated in real-time, and sustainable knowledge source, and may be the foundation of similar gains in a wide variety of fields.
Authors: Saleh Nikooroo
Abstract: Belief systems are often treated as globally consistent sets of propositions or as scalar-valued probability distributions. Such representations tend to obscure the internal structure of belief, conflate external credibility with internal coherence, and preclude the modeling of fragmented or contradictory epistemic states. This paper introduces a minimal formalism for belief systems as directed, weighted graphs. In this framework, nodes represent individual beliefs, edges encode epistemic relationships (e.g., support or contradiction), and two distinct functions assign each belief a credibility (reflecting source trust) and a confidence (derived from internal structural support). Unlike classical probabilistic models, our approach does not assume prior coherence or require belief updating. Unlike logical and argumentation-based frameworks, it supports fine-grained structural representation without committing to binary justification status or deductive closure. The model is purely static and deliberately excludes inference or revision procedures. Its aim is to provide a foundational substrate for analyzing the internal organization of belief systems, including coherence conditions, epistemic tensions, and representational limits. By distinguishing belief structure from belief strength, this formalism enables a richer classification of epistemic states than existing probabilistic, logical, or argumentation-based approaches.
Authors: Zhiyao Xu, Dan Zhao, Qingsong Zou, Qing Li, Yong Jiang, Yuhang Wang, Jingyu Xiao
Abstract: As smart homes become increasingly prevalent, intelligent models are widely used for tasks such as anomaly detection and behavior prediction. These models are typically trained on static datasets, making them brittle to behavioral drift caused by seasonal changes, lifestyle shifts, or evolving routines. However, collecting new behavior data for retraining is often impractical due to its slow pace, high cost, and privacy concerns. In this paper, we propose SmartGen, an LLM-based framework that synthesizes context-aware user behavior data to support continual adaptation of downstream smart home models. SmartGen consists of four key components. First, we design a Time and Semantic-aware Split module to divide long behavior sequences into manageable, semantically coherent subsequences under dual time-span constraints. Second, we propose Semantic-aware Sequence Compression to reduce input length while preserving representative semantics by clustering behavior mapping in latent space. Third, we introduce Graph-guided Sequence Synthesis, which constructs a behavior relationship graph and encodes frequent transitions into prompts, guiding the LLM to generate data aligned with contextual changes while retaining core behavior patterns. Finally, we design a Two-stage Outlier Filter to identify and remove implausible or semantically inconsistent outputs, aiming to improve the factual coherence and behavioral validity of the generated sequences. Experiments on three real-world datasets demonstrate that SmartGen significantly enhances model performance on anomaly detection and behavior prediction tasks under behavioral drift, with anomaly detection improving by 85.43% and behavior prediction by 70.51% on average. The code is available at https://github.com/horizonsinzqs/SmartGen.
Authors: Khaled Bachir Delassi (LIM Lab, Amar Telidji University, Laghouat, Algeria), Lakhdar Zeggane (LIM Lab, Amar Telidji University, Laghouat, Algeria), Hadda Cherroun (LIM Lab, Amar Telidji University, Laghouat, Algeria), Abdelhamid Haouhat (LIM Lab, Amar Telidji University, Laghouat, Algeria), Kaoutar Bouzouad (Computer Science Dept., USTHB, Algiers, Algeria)
Abstract: We address the problem of scarcity of educational Arabic Language Learning tools that advocate modern pedagogical models such as active learning which ensures language proficiency. In fact, we investigate the design and evaluation of an AI-powered educational tool designed to enhance Arabic language learning for non-native speakers with beginner-to-intermediate proficiency level. The tool leverages advanced AI models to generate interactive visual quizzes, deploying Visual Question Answering as the primary activity. Adopting a constructivist learning approach, the system encourages active learning through real-life visual quizzes, and image-based questions that focus on improving vocabulary, grammar, and comprehension. The system integrates Vision-Language Pretraining models to generate contextually relevant image description from which Large Language Model generate assignments based on customized Arabic language Learning quizzes thanks to prompting. The effectiveness of the tool is evaluated through a manual annotated benchmark consisting of 1266 real-life visual quizzes, with human participants providing feedback. The results show a suitable accuracy rates, validating the tool's potential to bridge the gap in Arabic language education and highlighting the tool's promise as a reliable, AI-powered resource for Arabic learners, offering personalized and interactive learning experiences.
Authors: Yijin Yang, Cristina Cornelio, Mario Leiva, Paulo Shakarian
Abstract: Recent large language models (LLMs) have demonstrated the ability to perform explicit multi-step reasoning such as chain-of-thought prompting. However, their intermediate steps often contain errors that can propagate leading to inaccurate final predictions. Additionally, LLMs still struggle with hallucinations and often fail to adhere to prescribed output formats, which is particularly problematic for tasks like generating mathematical expressions or source code. This work introduces EDCIM (Error Detection and Correction for Interpretable Mathematics), a method for detecting and correcting these errors in interpretable mathematics tasks, where the model must generate the exact functional form that explicitly solve the problem (expressed in natural language) rather than a black-box solution. EDCIM uses LLMs to generate a system of equations for a given problem, followed by a symbolic error-detection framework that identifies errors and provides targeted feedback for LLM-based correction. To optimize efficiency, EDCIM integrates lightweight, open-source LLMs with more powerful proprietary models, balancing cost and accuracy. This balance is controlled by a single hyperparameter, allowing users to control the trade-off based on their cost and accuracy requirements. Experimental results across different datasets show that EDCIM significantly reduces both computational and financial costs, while maintaining, and even improving, prediction accuracy when the balance is properly configured.
Authors: Jorge Gallego-Feliciano, S. Aaron McClendon, Juan Morinelli, Stavros Zervoudakis, Antonios Saravanos
Abstract: Massive activations are scalar values in transformer hidden states that achieve values orders of magnitude larger than typical activations and have been shown to be critical for model functionality. While prior work has characterized these phenomena in fully trained models, the temporal dynamics of their emergence during training remain poorly understood. We present the first comprehensive analysis of massive activation development throughout transformer training, using the Pythia model family as our testbed. Through systematic analysis of various model sizes across multiple training checkpoints, we demonstrate that massive activation emergence follows predictable mathematical patterns that can be accurately modeled using an exponentially-modulated logarithmic function with five key parameters. We develop a machine learning framework to predict these mathematical parameters from architectural specifications alone, achieving high accuracy for steady-state behavior and moderate accuracy for emergence timing and magnitude. These findings enable architects to predict and potentially control key aspects of massive activation emergence through design choices, with significant implications for model stability, training cycle length, interpretability, and optimization. Our findings demonstrate that the emergence of massive activations is governed by model design and can be anticipated, and potentially controlled, before training begins.
Authors: Jialin Li, Jinzhe Li, Gengxu Li, Yi Chang, Yuan Wu
Abstract: With the advancement of code generation capabilities in large language models (LLMs), their reliance on input premises has intensified. When users provide inputs containing faulty premises, the probability of code generation hallucinations rises significantly, exposing deficiencies in their self-scrutiny capabilities. This paper proposes Faulty Premises Bench (FPBench), the first code generation evaluation framework targeting faulty premises. By systematically constructing three categories of faulty premises and integrating multi-dimensional evaluation metrics, it conducts in-depth assessments of 15 representative LLMs. The key findings are as follows: (1) Most models exhibit poor reasoning abilities and suboptimal code generation performance under faulty premises, heavily relying on explicit prompts for error detection, with limited self-scrutiny capabilities; (2) Faulty premises trigger a point of diminishing returns in resource investment, leading to blindly increasing length fails to enhance quality; (3) The three types of faulty premises respectively activate distinct defect patterns in models, revealing a triple dissociation in the cognitive mechanisms of code generation models. This study not only highlights the urgent need for LLMs to proactively verify premises in code generation but also, through the proposed FPBench framework and multi-dimensional evaluation system, provides a theoretical foundation and practical pathway for developing reliable, human-centric code generation models.
Authors: He Wang, Liang Zeng
Abstract: Computational scientific discovery increasingly relies on algorithms to process complex data and identify meaningful patterns - yet faces persistent challenges in gravitational-wave signal identification. While existing algorithmic approaches like matched filtering (MF) and deep neural networks (DNNs) have achieved partial success, their limitations directly stem from fundamental limitations: MF's excessive computational demands arise from its reliance on predefined theoretical waveform templates, while DNNs' black-box architectures obscure decision logic and introduce hidden biases. We propose Evolutionary Monte Carlo Tree Search (Evo-MCTS), a framework that addresses these limitations through systematic algorithm space exploration guided by domain-aware physical constraints. Our approach combines tree-structured search with evolutionary optimization and large language model heuristics to create interpretable algorithmic solutions. Our Evo-MCTS framework demonstrates substantial improvements, achieving a 20.2\% improvement over state-of-the-art gravitational wave detection algorithms on the MLGWSC-1 benchmark dataset. High-performing algorithm variants consistently exceed thresholds. The framework generates human-interpretable algorithmic pathways that reveal distinct performance patterns. Beyond performance improvements, our framework discovers novel algorithmic combinations, thereby establishing a transferable methodology for automated algorithmic discovery across computational science domains.
Authors: Xufang Luo, Yuge Zhang, Zhiyuan He, Zilong Wang, Siyun Zhao, Dongsheng Li, Luna K. Qiu, Yuqing Yang
Abstract: We present Agent Lightning, a flexible and extensible framework that enables Reinforcement Learning (RL)-based training of Large Language Models (LLMs) for any AI agent. Unlike existing methods that tightly couple RL training with agent or rely on sequence concatenation with masking, Agent Lightning achieves complete decoupling between agent execution and training, allowing seamless integration with existing agents developed via diverse ways (e.g., using frameworks like LangChain, OpenAI Agents SDK, AutoGen, and building from scratch) with almost ZERO code modifications. By formulating agent execution as Markov decision process, we define an unified data interface and propose a hierarchical RL algorithm, LightningRL, which contains a credit assignment module, allowing us to decompose trajectories generated by ANY agents into training transition. This enables RL to handle complex interaction logic, such as multi-agent scenarios and dynamic workflows. For the system design, we introduce a Training-Agent Disaggregation architecture, and brings agent observability frameworks into agent runtime, providing a standardized agent finetuning interface. Experiments across text-to-SQL, retrieval-augmented generation, and math tool-use tasks demonstrate stable, continuous improvements, showcasing the framework's potential for real-world agent training and deployment.
Authors: Chengyu Bai, Yuming Li, Zhongyu Zhao, Jintao Chen, Peidong Jia, Qi She, Ming Lu, Shanghang Zhang
Abstract: Video generation has made significant strides with the development of diffusion models; however, achieving high temporal consistency remains a challenging task. Recently, FreeInit identified a training-inference gap and introduced a method to iteratively refine the initial noise during inference. However, iterative refinement significantly increases the computational cost associated with video generation. In this paper, we introduce FastInit, a fast noise initialization method that eliminates the need for iterative refinement. FastInit learns a Video Noise Prediction Network (VNPNet) that takes random noise and a text prompt as input, generating refined noise in a single forward pass. Therefore, FastInit greatly enhances the efficiency of video generation while achieving high temporal consistency across frames. To train the VNPNet, we create a large-scale dataset consisting of pairs of text prompts, random noise, and refined noise. Extensive experiments with various text-to-video models show that our method consistently improves the quality and temporal consistency of the generated videos. FastInit not only provides a substantial improvement in video generation but also offers a practical solution that can be applied directly during inference. The code and dataset will be released.
Authors: Long S. T. Nguyen, Khang H. N. Vo, Thu H. A. Nguyen, Tuan C. Bui, Duc Q. Nguyen, Thanh-Tung Tran, Anh D. Nguyen, Minh L. Nguyen, Fabien Baldacci, Thang H. Bui, Emanuel Di Nardo, Angelo Ciaramella, Son H. Le, Ihsan Ullah, Lorenzo Di Rocco, Tho T. Quan
Abstract: The growing integration of Artificial Intelligence (AI) into education has intensified the need for transparency and interpretability. While hackathons have long served as agile environments for rapid AI prototyping, few have directly addressed eXplainable AI (XAI) in real-world educational contexts. This paper presents a comprehensive analysis of the XAI Challenge 2025, a hackathon-style competition jointly organized by Ho Chi Minh City University of Technology (HCMUT) and the International Workshop on Trustworthiness and Reliability in Neurosymbolic AI (TRNS-AI), held as part of the International Joint Conference on Neural Networks (IJCNN 2025). The challenge tasked participants with building Question-Answering (QA) systems capable of answering student queries about university policies while generating clear, logic-based natural language explanations. To promote transparency and trustworthiness, solutions were required to use lightweight Large Language Models (LLMs) or hybrid LLM-symbolic systems. A high-quality dataset was provided, constructed via logic-based templates with Z3 validation and refined through expert student review to ensure alignment with real-world academic scenarios. We describe the challenge's motivation, structure, dataset construction, and evaluation protocol. Situating the competition within the broader evolution of AI hackathons, we argue that it represents a novel effort to bridge LLMs and symbolic reasoning in service of explainability. Our findings offer actionable insights for future XAI-centered educational systems and competitive research initiatives.
Authors: Pragya Singh, Ankush Gupta, Mohan Kumar, Pushpendra Singh
Abstract: Emotional and mental well-being are vital components of quality of life, and with the rise of smart devices like smartphones, wearables, and artificial intelligence (AI), new opportunities for monitoring emotions in everyday settings have emerged. However, for AI algorithms to be effective, they require high-quality data and accurate annotations. As the focus shifts towards collecting emotion data in real-world environments to capture more authentic emotional experiences, the process of gathering emotion annotations has become increasingly complex. This work explores the challenges of everyday emotion data collection from the perspectives of key stakeholders. We collected 75 survey responses, performed 32 interviews with the public, and 3 focus group discussions (FGDs) with 12 mental health professionals. The insights gained from a total of 119 stakeholders informed the development of our framework, AnnoSense, designed to support everyday emotion data collection for AI. This framework was then evaluated by 25 emotion AI experts for its clarity, usefulness, and adaptability. Lastly, we discuss the potential next steps and implications of AnnoSense for future research in emotion AI, highlighting its potential to enhance the collection and analysis of emotion data in real-world contexts.
Authors: Beatriz Macas Ord\'o\~nez, Diego Vinicio Orellana Villavicencio, Jos\'e Manuel Ferr\'andez, Paula Bonomini
Abstract: This study explores different neural network architectures to evaluate their ability to extract spatial and temporal patterns from electrocardiographic (ECG) signals and classify them into three groups: healthy subjects, Left Bundle Branch Block (LBBB), and Strict Left Bundle Branch Block (sLBBB). Clinical Relevance, Innovative technologies enable the selection of candidates for Cardiac Resynchronization Therapy (CRT) by optimizing the classification of subjects with Left Bundle Branch Block (LBBB).
Authors: Yidong Chai (School of Management, Hefei University of Technology, Hefei, China, Key Laboratory of Process Optimization and Intelligent Decision-making, Ministry of Education, Hefei, China), Yang Liu (School of Management, Hefei University of Technology, Hefei, China, Key Laboratory of Process Optimization and Intelligent Decision-making, Ministry of Education, Hefei, China), Yonghang Zhou (School of Management, Hefei University of Technology, Hefei, China, Key Laboratory of Process Optimization and Intelligent Decision-making, Ministry of Education, Hefei, China), Jiaheng Xie (Department of Accounting and MIS, Lerner College of Business and Economics, University of Delaware, Newark, Delaware, U.S), Daniel Dajun Zeng (Institute of Automation, Chinese Academy of Sciences, Beijing, China)
Abstract: Large Language Models (LLMs) have demonstrated transformative potential in reshaping the world. As these models are pretrained on general corpora, they often require domain-specific fine-tuning to optimize performance in specialized business applications. Due to their massive scale, parameter-efficient fine-tuning (PEFT) methods are widely used to reduce training costs. Among them, hybrid PEFT methods that combine multiple PEFT techniques have achieved the best performance. However, existing hybrid PEFT methods face two main challenges when fine-tuning LLMs for specialized applications: (1) relying on point estimates, lacking the ability to quantify uncertainty for reliable decision-making, and (2) struggling to dynamically adapt to emerging data, lacking the ability to suit real-world situations. We propose Bayesian Hybrid Parameter-Efficient Fine-Tuning (BH-PEFT), a novel method that integrates Bayesian learning into hybrid PEFT. BH-PEFT combines Adapter, LoRA, and prefix-tuning to fine-tune feedforward and attention layers of the Transformer. By modeling learnable parameters as distributions, BH-PEFT enables uncertainty quantification. We further propose a Bayesian dynamic fine-tuning approach where the last posterior serves as the prior for the next round, enabling effective adaptation to new data. We evaluated BH-PEFT on business tasks such as sentiment analysis, news categorization, and commonsense reasoning. Results show that our method outperforms existing PEFT baselines, enables uncertainty quantification for more reliable decisions, and improves adaptability in dynamic scenarios. This work contributes to business analytics and data science by proposing a novel BH-PEFT method and dynamic fine-tuning approach that support uncertainty-aware and adaptive decision-making in real-world situations.
Authors: Zahra Mohammadi, Siamak Mohammadi
Abstract: Apnea is a common sleep disorder characterized by breathing interruptions lasting at least ten seconds and occurring more than five times per hour. Accurate, high-temporal-resolution detection of sleep apnea subtypes - Obstructive, Central, and Mixed - is crucial for effective treatment and management. This paper presents an energy-efficient method for classifying these subtypes using a single-lead electrocardiogram (ECG) with high temporal resolution to address the real-time needs of wearable devices. We evaluate a wide range of classical machine learning algorithms and deep learning architectures on 1-second ECG windows, comparing their accuracy, complexity, and energy consumption. Based on this analysis, we introduce SleepLiteCNN, a compact and energy-efficient convolutional neural network specifically designed for wearable platforms. SleepLiteCNN achieves over 95% accuracy and a 92% macro-F1 score, while requiring just 1.8 microjoules per inference after 8-bit quantization. Field Programmable Gate Array (FPGA) synthesis further demonstrates significant reductions in hardware resource usage, confirming its suitability for continuous, real-time monitoring in energy-constrained environments. These results establish SleepLiteCNN as a practical and effective solution for wearable device sleep apnea subtype detection.
Authors: Samiksha BC
Abstract: This work introduces ZetA, a novel deep learning optimizer that extends Adam by incorporating dynamic scaling based on the Riemann zeta function. To the best of our knowledge, ZetA is the first optimizer to apply zeta-based gradient scaling within deep learning optimization. The method improves generalization and robustness through a hybrid update mechanism that integrates adaptive damping, cosine similarity-based momentum boosting, entropy-regularized loss, and Sharpness-Aware Minimization (SAM)-style perturbations. Empirical evaluations on SVHN, CIFAR10, CIFAR100, STL10, and noisy CIFAR10 consistently show test accuracy improvements over Adam. All experiments employ a lightweight fully connected network trained for five epochs under mixed-precision settings. The results demonstrate that ZetA is a computationally efficient and robust alternative to Adam, particularly effective in noisy or high-granularity classification tasks.
Authors: Yongfan Lai, Bo Liu, Xinyan Guan, Qinghao Zhao, Hongyan Li, Shenda Hong
Abstract: Personalized electrocardiogram (ECG) generation is to simulate a patient's ECG digital twins tailored to specific conditions. It has the potential to transform traditional healthcare into a more accurate individualized paradigm, while preserving the key benefits of conventional population-level ECG synthesis. However, this promising task presents two fundamental challenges: extracting individual features without ground truth and injecting various types of conditions without confusing generative model. In this paper, we present ECGTwin, a two-stage framework designed to address these challenges. In the first stage, an Individual Base Extractor trained via contrastive learning robustly captures personal features from a reference ECG. In the second stage, the extracted individual features, along with a target cardiac condition, are integrated into the diffusion-based generation process through our novel AdaX Condition Injector, which injects these signals via two dedicated and specialized pathways. Both qualitative and quantitative experiments have demonstrated that our model can not only generate ECG signals of high fidelity and diversity by offering a fine-grained generation controllability, but also preserving individual-specific features. Furthermore, ECGTwin shows the potential to enhance ECG auto-diagnosis in downstream application, confirming the possibility of precise personalized healthcare solutions.
Authors: Libin Qiu, Yuhang Ye, Zhirong Gao, Xide Zou, Junfu Chen, Ziming Gui, Weizhi Huang, Xiaobo Xue, Wenkai Qiu, Kun Zhao
Abstract: While powerful, the inherent non-determinism of large language model (LLM) agents limits their application in structured operational environments where procedural fidelity and predictable execution are strict requirements. This limitation stems from current architectures that conflate probabilistic, high-level planning with low-level action execution within a single generative process. To address this, we introduce the Source Code Agent framework, a new paradigm built on the "Blueprint First, Model Second" philosophy. Our framework decouples the workflow logic from the generative model. An expert-defined operational procedure is first codified into a source code-based Execution Blueprint, which is then executed by a deterministic engine. The LLM is strategically invoked as a specialized tool to handle bounded, complex sub-tasks within the workflow, but never to decide the workflow's path. We conduct a comprehensive evaluation on the challenging tau-bench benchmark, designed for complex user-tool-rule scenarios. Our results demonstrate that the Source Code Agent establishes a new state-of-the-art, outperforming the strongest baseline by 10.1 percentage points on the average Pass^1 score while dramatically improving execution efficiency. Our work enables the verifiable and reliable deployment of autonomous agents in applications governed by strict procedural logic.
Authors: Haitz S\'aez de Oc\'ariz Borde, Michael Bronstein
Abstract: We review the key mathematical concepts necessary for studying Geometric Deep Learning.
Authors: Yahia Dalbah, Marcel Worring, Yen-Chia Hsu
Abstract: Urban air pollution is a major health crisis causing millions of premature deaths annually, underscoring the urgent need for accurate and scalable monitoring of air quality (AQ). While low-cost sensors (LCS) offer a scalable alternative to expensive reference-grade stations, their readings are affected by drift, calibration errors, and environmental interference. To address these challenges, we introduce Veli (Reference-free Variational Estimation via Latent Inference), an unsupervised Bayesian model that leverages variational inference to correct LCS readings without requiring co-location with reference stations, eliminating a major deployment barrier. Specifically, Veli constructs a disentangled representation of the LCS readings, effectively separating the true pollutant reading from the sensor noise. To build our model and address the lack of standardized benchmarks in AQ monitoring, we also introduce the Air Quality Sensor Data Repository (AQ-SDR). AQ-SDR is the largest AQ sensor benchmark to date, with readings from 23,737 LCS and reference stations across multiple regions. Veli demonstrates strong generalization across both in-distribution and out-of-distribution settings, effectively handling sensor drift and erratic sensor behavior. Code for model and dataset will be made public when this paper is published.
Authors: Md Imtiaz Habib
Abstract: In this research, I explore advanced deep learning methodologies to forecast the outcomes of the 2025 NCAA Division 1 Men's and Women's Basketball tournaments. Leveraging historical NCAA game data, I implement two sophisticated sequence-based models: Long Short-Term Memory (LSTM) and Transformer architectures. The predictive power of these models is augmented through comprehensive feature engineering, including team quality metrics derived from Generalized Linear Models (GLM), Elo ratings, seed differences, and aggregated box-score statistics. To evaluate the robustness and reliability of predictions, I train each model variant using both Binary Cross-Entropy (BCE) and Brier loss functions, providing insights into classification performance and probability calibration. My comparative analysis reveals that while the Transformer architecture optimized with BCE yields superior discriminative power (highest AUC of 0.8473), the LSTM model trained with Brier loss demonstrates superior probabilistic calibration (lowest Brier score of 0.1589). These findings underscore the importance of selecting appropriate model architectures and loss functions based on the specific requirements of forecasting tasks. The detailed analytical pipeline presented here serves as a reproducible framework for future predictive modeling tasks in sports analytics and beyond.
Authors: Zhuoran Liu
Abstract: Profiling tools (also known as profilers) play an important role in understanding program performance at runtime, such as hotspots, bottlenecks, and inefficiencies. While profilers have been proven to be useful, they give extra burden to software engineers. Software engineers, as the users, are responsible to interpret the complex performance data and identify actionable optimization in program source code. However, it can be challenging for users to associate inefficiencies with the program semantics, especially if the users are not the authors of the code, which limits the applicability of profilers. In this thesis, we explore a new direction to combine performance profiles and program semantics with a deep learning approach. The key idea is to glean code summary for semantic information (at a certain level) and integrate it into a profiler, which can better understand program inefficiencies for actionable optimization. To be concrete, we combine profiles generated by Async Profiler (the state-of-the-art Java profiler) with code summarization from a fine-tuned CodeBERT-based model. We demonstrate the code summaries of any selected call path in a graphic user interface. Our system can effectively assist analysis on many Java benchmarks.
Authors: Jean-Francois Chamberland, Martin C. Carlisle, Arul Jayaraman, Krishna R. Narayanan, Sunay Palsole, Karan Watson
Abstract: Evaluating teaching effectiveness at scale remains a persistent challenge for large universities, particularly within engineering programs that enroll tens of thousands of students. Traditional methods, such as manual review of student evaluations, are often impractical, leading to overlooked insights and inconsistent data use. This article presents a scalable, AI-supported framework for synthesizing qualitative student feedback using large language models. The system employs hierarchical summarization, anonymization, and exception handling to extract actionable themes from open-ended comments while upholding ethical safeguards. Visual analytics contextualize numeric scores through percentile-based comparisons, historical trends, and instructional load. The approach supports meaningful evaluation and aligns with best practices in qualitative analysis and educational assessment, incorporating student, peer, and self-reflective inputs without automating personnel decisions. We report on its successful deployment across a large college of engineering. Preliminary validation through comparisons with human reviewers, faculty feedback, and longitudinal analysis suggests that LLM-generated summaries can reliably support formative evaluation and professional development. This work demonstrates how AI systems, when designed with transparency and shared governance, can promote teaching excellence and continuous improvement at scale within academic institutions.
Authors: Sherman Wong, Jalaj Bhandari, Leo Zhou Fan Yang, Xylan Xu, Yi Zhuang, Cem Cayiroglu, Payal Bhuptani, Sheela Yadawad, Hung Duong
Abstract: Maintaining code quality in large-scale software systems presents significant challenges, particularly in settings where a large numbers of engineers work concurrently on a codebase. This paper introduces Code Quality Score (CQS) system to automatically detect issues with a set of code changes and provide actionable insights. At its core, the CQS system is powered by two Llama3 models, fine-tuned (with SFT and offline RL approaches), to a) detect common code quality issues related to coding best practices and b) to provide good ``critiques'' for LLM-generated code review respectively. To maintain good user experience, we layer the system with hand-crafted rules to filter out incorrect responses/hallucinations. Offline evaluations show that our CQS system is able to achieve an impressive precision rate for identifying valid issues. This system has already been rolled out to developers in an industrial scale setting and has consistently achieved 60\% week over week user helpfulness rate, demonstrating its effectiveness in a real-world environment. In this paper, we present details of the CQS system along with some learnings on curating developer feedback to create training data for LLM fine-tuning.
Authors: Yu Shi, Zongliang Fu, Shuo Chen, Bohan Zhao, Wei Xu, Changshui Zhang, Jian Li
Abstract: The success of large-scale pre-training paradigm, exemplified by Large Language Models (LLMs), has inspired the development of Time Series Foundation Models (TSFMs). However, their application to financial candlestick (K-line) data remains limited, often underperforming non-pre-trained architectures. Moreover, existing TSFMs often overlook crucial downstream tasks such as volatility prediction and synthetic data generation. To address these limitations, we propose Kronos, a unified, scalable pre-training framework tailored to financial K-line modeling. Kronos introduces a specialized tokenizer that discretizes continuous market information into token sequences, preserving both price dynamics and trade activity patterns. We pre-train Kronos using an autoregressive objective on a massive, multi-market corpus of over 12 billion K-line records from 45 global exchanges, enabling it to learn nuanced temporal and cross-asset representations. Kronos excels in a zero-shot setting across a diverse set of financial tasks. On benchmark datasets, Kronos boosts price series forecasting RankIC by 93% over the leading TSFM and 87% over the best non-pre-trained baseline. It also achieves a 9% lower MAE in volatility forecasting and a 22% improvement in generative fidelity for synthetic K-line sequences. These results establish Kronos as a robust, versatile foundation model for end-to-end financial time series analysis. Our pre-trained model is publicly available at https://github.com/shiyu-coder/Kronos.
Authors: Jiangen He
Abstract: Large language models (LLMs) are rapidly being adopted as research assistants, particularly for literature review and reference recommendation, yet little is known about whether they introduce demographic bias into citation workflows. This study systematically investigates gender bias in LLM-driven reference selection using controlled experiments with pseudonymous author names. We evaluate several LLMs (GPT-4o, GPT-4o-mini, Claude Sonnet, and Claude Haiku) by varying gender composition within candidate reference pools and analyzing selection patterns across fields. Our results reveal two forms of bias: a persistent preference for male-authored references and a majority-group bias that favors whichever gender is more prevalent in the candidate pool. These biases are amplified in larger candidate pools and only modestly attenuated by prompt-based mitigation strategies. Field-level analysis indicates that bias magnitude varies across scientific domains, with social sciences showing the least bias. Our findings indicate that LLMs can reinforce or exacerbate existing gender imbalances in scholarly recognition. Effective mitigation strategies are needed to avoid perpetuating existing gender disparities in scientific citation practices before integrating LLMs into high-stakes academic workflows.
Authors: Zhixiang Lu, Yulong Li, Feilong Tang, Zhengyong Jiang, Chong Li, Mian Zhou, Tenglong Li, Jionglong Su
Abstract: Large-scale tuberculosis (TB) screening is limited by the high cost and operational complexity of traditional diagnostics, creating a need for artificial-intelligence solutions. We propose DeepGB-TB, a non-invasive system that instantly assigns TB risk scores using only cough audio and basic demographic data. The model couples a lightweight one-dimensional convolutional neural network for audio processing with a gradient-boosted decision tree for tabular features. Its principal innovation is a Cross-Modal Bidirectional Cross-Attention module (CM-BCA) that iteratively exchanges salient cues between modalities, emulating the way clinicians integrate symptoms and risk factors. To meet the clinical priority of minimizing missed cases, we design a Tuberculosis Risk-Balanced Loss (TRBL) that places stronger penalties on false-negative predictions, thereby reducing high-risk misclassifications. DeepGB-TB is evaluated on a diverse dataset of 1,105 patients collected across seven countries, achieving an AUROC of 0.903 and an F1-score of 0.851, representing a new state of the art. Its computational efficiency enables real-time, offline inference directly on common mobile devices, making it ideal for low-resource settings. Importantly, the system produces clinically validated explanations that promote trust and adoption by frontline health workers. By coupling AI innovation with public-health requirements for speed, affordability, and reliability, DeepGB-TB offers a tool for advancing global TB control.
Authors: Chunyu Liu, Hao Zhang, Wei Wu, Fuhui Zhou, Qihui Wu, Derrick Wing Kwan Ng, Chan-Byoung Chae
Abstract: The enhancement of spectrum efficiency and the realization of secure spectrum utilization are critically dependent on spectrum cognition. However, existing spectrum cognition methods often exhibit limited generalization and suboptimal accuracy when deployed across diverse spectrum environments and tasks. To overcome these challenges, we propose a spectrum foundation model, termed SpectrumFM, which provides a new paradigm for spectrum cognition. An innovative spectrum encoder that exploits the convolutional neural networks and the multi-head self attention mechanisms is proposed to effectively capture both fine-grained local signal structures and high-level global dependencies in the spectrum data. To enhance its adaptability, two novel self-supervised learning tasks, namely masked reconstruction and next-slot signal prediction, are developed for pre-training SpectrumFM, enabling the model to learn rich and transferable representations. Furthermore, low-rank adaptation (LoRA) parameter-efficient fine-tuning is exploited to enable SpectrumFM to seamlessly adapt to various downstream spectrum cognition tasks, including spectrum sensing (SS), anomaly detection (AD), and wireless technology classification (WTC). Extensive experiments demonstrate the superiority of SpectrumFM over state-of-the-art methods. Specifically, it improves detection probability in the SS task by 30% at -4 dB signal-to-noise ratio (SNR), boosts the area under the curve (AUC) in the AD task by over 10%, and enhances WTC accuracy by 9.6%.
Authors: Vinil Polepalli
Abstract: Pan-cancer classification using transcriptomic (RNA-Seq) data can inform tumor subtyping and therapy selection, but is challenging due to extremely high dimensionality and limited sample sizes. In this study, we propose a novel deep learning framework that uses a class-conditional variational autoencoder (cVAE) to augment training data for pan-cancer gene expression classification. Using 801 tumor RNA-Seq samples spanning 5 cancer types from The Cancer Genome Atlas (TCGA), we first perform feature selection to reduce 20,531 gene expression features to the 500 most variably expressed genes. A cVAE is then trained on this data to learn a latent representation of gene expression conditioned on cancer type, enabling the generation of synthetic gene expression samples for each tumor class. We augment the training set with these cVAE-generated samples (doubling the dataset size) to mitigate overfitting and class imbalance. A two-layer multilayer perceptron (MLP) classifier is subsequently trained on the augmented dataset to predict tumor type. The augmented framework achieves high classification accuracy (~98%) on a held-out test set, substantially outperforming a classifier trained on the original data alone. We present detailed experimental results, including VAE training curves, classifier performance metrics (ROC curves and confusion matrix), and architecture diagrams to illustrate the approach. The results demonstrate that cVAE-based synthetic augmentation can significantly improve pan-cancer prediction performance, especially for underrepresented cancer classes.
Authors: Haoran Liu, Yihan Zhan, Mingzhe Liu, Yanhua Liu, Peng Li, Zhuo Zuo, Bingqi Liu, Runxi Liu
Abstract: This review presents a comprehensive survey and benchmark of pulse shape discrimination (PSD) algorithms for radiation detection, classifying nearly sixty methods into statistical (time-domain, frequency-domain, neural network-based) and prior-knowledge (machine learning, deep learning) paradigms. We implement and evaluate all algorithms on two standardized datasets: an unlabeled set from a 241Am-9Be source and a time-of-flight labeled set from a 238Pu-9Be source, using metrics including Figure of Merit (FOM), F1-score, ROC-AUC, and inter-method correlations. Our analysis reveals that deep learning models, particularly Multi-Layer Perceptrons (MLPs) and hybrid approaches combining statistical features with neural regression, often outperform traditional methods. We discuss architectural suitabilities, the limitations of FOM, alternative evaluation metrics, and performance across energy thresholds. Accompanying this work, we release an open-source toolbox in Python and MATLAB, along with the datasets, to promote reproducibility and advance PSD research.
Authors: Yi Zhao, Yajuan Peng, Cam-Tu Nguyen, Zuchao Li, Xiaoliang Wang, Hai Zhao, Xiaoming Fu
Abstract: KV cache eviction has emerged as an effective solution to alleviate resource constraints faced by LLMs in long-context scenarios. However, existing token-level eviction methods often overlook two critical aspects: (1) their irreversible eviction strategy fails to adapt to dynamic attention patterns during decoding (the saliency shift problem), and (2) they treat both marginally important tokens and truly unimportant tokens equally, despite the collective significance of marginal tokens to model performance (the marginal information over-compression problem). To address these issues, we design two compensation mechanisms based on the high similarity of attention matrices between LLMs of different scales. We propose SmallKV, a small model assisted compensation method for KV cache compression. SmallKV can maintain attention matching between different-scale LLMs to: 1) assist the larger model in perceiving globally important information of attention; and 2) use the smaller model's attention scores to approximate those of marginal tokens in the larger model. Extensive experiments on benchmarks including GSM8K, BBH, MT-Bench, and LongBench demonstrate the effectiveness of SmallKV. Moreover, efficiency evaluations show that SmallKV achieves 1.75 - 2.56 times higher throughput than baseline methods, highlighting its potential for efficient and performant LLM inference in resource constrained environments.
Authors: Haonan Yang, Jianchao Tang, Zhuo Li, Long Lan
Abstract: Time Series Forecasting (TSF) faces persistent challenges in modeling intricate temporal dependencies across different scales. Despite recent advances leveraging different decomposition operations and novel architectures based on CNN, MLP or Transformer, existing methods still struggle with static decomposition strategies, fragmented dependency modeling, and inflexible fusion mechanisms, limiting their ability to model intricate temporal dependencies. To explicitly solve the mentioned three problems respectively, we propose a novel Dynamic Multi-Scale Coordination Framework (DMSC) with Multi-Scale Patch Decomposition block (EMPD), Triad Interaction Block (TIB) and Adaptive Scale Routing MoE block (ASR-MoE). Specifically, EMPD is designed as a built-in component to dynamically segment sequences into hierarchical patches with exponentially scaled granularities, eliminating predefined scale constraints through input-adaptive patch adjustment. TIB then jointly models intra-patch, inter-patch, and cross-variable dependencies within each layer's decomposed representations. EMPD and TIB are jointly integrated into layers forming a multi-layer progressive cascade architecture, where coarse-grained representations from earlier layers adaptively guide fine-grained feature extraction in subsequent layers via gated pathways. And ASR-MoE dynamically fuses multi-scale predictions by leveraging specialized global and local experts with temporal-aware weighting. Comprehensive experiments on thirteen real-world benchmarks demonstrate that DMSC consistently maintains state-of-the-art (SOTA) performance and superior computational efficiency for TSF tasks. Code is available at https://github.com/1327679995/DMSC.
Authors: Miko{\l}aj Sienicki, Krzysztof Sienicki
Abstract: We propose a formal reconstruction of quantum mechanics grounded not in external mathematical abstractions, but in the structured dynamics of subjective experience. The Qualia Abstraction Language (QAL) models physical systems as evolving streams of introspective units, structured sequences of modality, shape, and functional effect, rather than as state vectors in Hilbert space. This approach reimagines core quantum concepts: superposition becomes a form of structured ambiguity; collapse is reframed as an introspective contraction; and entanglement is modeled as semantic resonance across streams of qualia. Drawing on insights from nominalist philosophy and oversight theoretic limits in AI, we argue that the observer paradox in quantum mechanics reflects not an ontological lacuna, but a linguistic one: the absence of a formal vocabulary for modeling first person structure. QAL introduces such a vocabulary, providing a morphodynamic framework that embeds the observer within the system and replaces abstract projection with endogenous transformation. We analyze the alignment of QAL with endophysical approaches, contrast it with standard interpretations of quantum theory, and explore its implications for a post Platonist, introspectively grounded physics.
Authors: Yihao Ang, Qiang Wang, Qiang Huang, Yifan Bao, Xinyu Xi, Anthony K. H. Tung, Chen Jin, Zhiyong Huang
Abstract: Synthetic time series are essential tools for data augmentation, stress testing, and algorithmic prototyping in quantitative finance. However, in cryptocurrency markets, characterized by 24/7 trading, extreme volatility, and rapid regime shifts, existing Time Series Generation (TSG) methods and benchmarks often fall short, jeopardizing practical utility. Most prior work (1) targets non-financial or traditional financial domains, (2) focuses narrowly on classification and forecasting while neglecting crypto-specific complexities, and (3) lacks critical financial evaluations, particularly for trading applications. To address these gaps, we introduce \textsf{CTBench}, the first comprehensive TSG benchmark tailored for the cryptocurrency domain. \textsf{CTBench} curates an open-source dataset from 452 tokens and evaluates TSG models across 13 metrics spanning 5 key dimensions: forecasting accuracy, rank fidelity, trading performance, risk assessment, and computational efficiency. A key innovation is a dual-task evaluation framework: (1) the \emph{Predictive Utility} task measures how well synthetic data preserves temporal and cross-sectional patterns for forecasting, while (2) the \emph{Statistical Arbitrage} task assesses whether reconstructed series support mean-reverting signals for trading. We benchmark eight representative models from five methodological families over four distinct market regimes, uncovering trade-offs between statistical fidelity and real-world profitability. Notably, \textsf{CTBench} offers model ranking analysis and actionable guidance for selecting and deploying TSG models in crypto analytics and strategy development.
Authors: Giovanni Adorni, Emanuele Bellini
Abstract: The accelerated evolution of digital infrastructures and algorithmic systems is reshaping how the humanities engage with knowledge and culture. Rooted in the traditions of Digital Humanities and Digital Humanism, the concept of "Cyber Humanities" proposes a critical reconfiguration of humanistic inquiry for the post-digital era. This Manifesto introduces a flexible framework that integrates ethical design, sustainable digital practices, and participatory knowledge systems grounded in human-centered approaches. By means of a Decalogue of foundational principles, the Manifesto invites the scientific community to critically examine and reimagine the algorithmic infrastructures that influence culture, creativity, and collective memory. Rather than being a simple extension of existing practices, "Cyber Humanities" should be understood as a foundational paradigm for humanistic inquiry in a computationally mediated world. Keywords: Cyber Humanities, Digital Humanities, Transdisciplinary Epistemology, Algorithmic Reflexivity, Human-centered AI, Ethics-by-Design, Knowledge Ecosystems, Digital Sovereignty, Cognitive Infrastructures
Authors: Dahun Kim, Anelia Angelova
Abstract: We propose Context-Adaptive Multi-Prompt Embedding, a novel approach to enrich semantic representations in vision-language contrastive learning. Unlike standard CLIP-style models that rely on a single text embedding, our method introduces multiple structured prompts, each containing a distinct adaptive token that captures diverse semantic aspects of the input text. We leverage a pretrained LLM as the text encoder within the CLIP framework, processing all prompts jointly in a single forward pass. The resulting prompt embeddings are combined into a unified text representation, enabling semantically richer alignment with visual features. To further promote semantic diversity and representation quality, we incorporate a diversity regularization loss and a negation-aware loss, encouraging specialization across prompts and improving contrastive discrimination. Our method achieves consistent improvements on both image-text and video-text retrieval benchmarks.
Authors: Petteri Teikari, Mike Jarrell, Maryam Azh, Harri Pesola
Abstract: The Uniform Appraisal Dataset (UAD) 3.6's mandatory 2026 implementation transforms residential property valuation from narrative reporting to structured, machine-readable formats. This paper provides the first comprehensive analysis of this regulatory shift alongside concurrent AI advances in computer vision, natural language processing, and autonomous systems. We develop a three-layer framework for AI-augmented valuation addressing technical implementation and institutional trust requirements. Our analysis reveals how regulatory standardization converging with AI capabilities enables fundamental market restructuring with profound implications for professional practice, efficiency, and systemic risk. We make four key contributions: (1) documenting institutional failures including inter-appraiser variability and systematic biases undermining valuation reliability; (2) developing an architectural framework spanning physical data acquisition, semantic understanding, and cognitive reasoning that integrates emerging technologies while maintaining professional oversight; (3) addressing trust requirements for high-stakes financial applications including regulatory compliance, algorithmic fairness, and uncertainty quantification; (4) proposing evaluation methodologies beyond generic AI benchmarks toward domain-specific protocols. Our findings indicate successful transformation requires not merely technological sophistication but careful human-AI collaboration, creating systems that augment rather than replace professional expertise while addressing historical biases and information asymmetries in real estate markets.
Authors: Yonathan A. Arbel
Abstract: In everyday life, people make countless reasonableness judgments that determine appropriate behavior in various contexts. Predicting these judgments challenges the legal system, as judges' intuitions may not align with broader societal views. This Article investigates whether large language models (LLMs) can learn to identify patterns driving human reasonableness judgments. Using randomized controlled trials comparing humans and models across multiple legal contexts with over 10,000 simulated judgments, we demonstrate that certain models capture not just surface-level responses but potentially their underlying decisional architecture. Strikingly, these systems prioritize social cues over economic efficiency in negligence determinations, mirroring human behavior despite contradicting textbook treatments. These findings suggest practical applications: judges could calibrate intuitions against broader patterns, lawmakers could test policy interpretations, and resource-constrained litigants could preview argument reception. As AI agents increasingly make autonomous real-world decisions, understanding whether they've internalized recognizable ethical frameworks becomes essential for anticipating their behavior.
Authors: Yiming Shen, Jiashuo Zhang, Zhenzhe Shao, Wenxuan Luo, Yanlin Wang, Ting Chen, Zibin Zheng, Jiachi Chen
Abstract: The convergence of Web3 technologies and AI agents represents a rapidly evolving frontier poised to reshape decentralized ecosystems. This paper presents the first and most comprehensive analysis of the intersection between Web3 and AI agents, examining five critical dimensions: landscape, economics, governance, security, and trust mechanisms. Through an analysis of 133 existing projects, we first develop a taxonomy and systematically map the current market landscape (RQ1), identifying distinct patterns in project distribution and capitalization. Building upon these findings, we further investigate four key integrations: (1) the role of AI agents in participating in and optimizing decentralized finance (RQ2); (2) their contribution to enhancing Web3 governance mechanisms (RQ3); (3) their capacity to strengthen Web3 security via intelligent vulnerability detection and automated smart contract auditing (RQ4); and (4) the establishment of robust reliability frameworks for AI agent operations leveraging Web3's inherent trust infrastructure (RQ5). By synthesizing these dimensions, we identify key integration patterns, highlight foundational challenges related to scalability, security, and ethics, and outline critical considerations for future research toward building robust, intelligent, and trustworthy decentralized systems with effective AI agent interactions.
Authors: Jessica Sanson, Rahul C. Shah, Maximilian Pinaroc, Valerio Frascolla
Abstract: This paper presents, for the first time, a method to extract both range and Doppler information from commercial Wi-Fi Channel State Information (CSI) using a monostatic (single transceiver) setup. Utilizing the CSI phase in Wi-Fi sensing from a Network Interface Card (NIC) not designed for full-duplex operation is challenging due to (1) Hardware asynchronization, which introduces significant phase errors, and (2) Proximity of transmit (Tx) and receive (Rx) antennas, which creates strong coupling that overwhelms the motion signal of interest. We propose a new signal processing approach that addresses both challenges via three key innovations: Time offset cancellation, Phase alignment correction, and Tx/Rx coupling mitigation. Our method achieves cm-level accuracy in range and Doppler estimation for moving targets, validated using a commercial Intel Wi-Fi AX211 NIC. Our results show successful detection and tracking of moving objects in realistic environments, establishing the feasibility of high-precision sensing using standard Wi-Fi packet communications and off-the-shelf hardware without requiring any modification or specialized full-duplex capabilities.
Authors: Hyung Gun Chi, Florian Pesce, Wonil Chang, Oggi Rudovic, Arturo Argueta, Stefan Braun, Vineet Garg, Ahmed Hussen Abdelaziz
Abstract: Device-directed speech detection (DDSD) is a binary classification task that separates the user's queries to a voice assistant (VA) from background speech or side conversations. This is important for achieving naturalistic user experience. To this end, we propose knowledge distillation (KD) to enhance DDSD accuracy while ensuring efficient deployment. Specifically, we introduce a novel adaptive KD method that transfers knowledge from general representations of an ASR large pre-trained acoustic encoder (teacher). We apply task-specific adapters, on top of the (frozen) teacher encoder, trained jointly with the student model on DDSD. We demonstrate that the proposed adaptive KD outperforms the student model without distillation in the keyword and keyword-free (follow-up) invocations, with an improvement of +26% and +19% in terms of Equal Error Rate, respectively. We also show that this approach generalizes across the transformer and conformer-based model architectures.
Authors: Radhika Dua (Fred), Young Joon (Fred), Kwon, Siddhant Dogra, Daniel Freedman, Diana Ruan, Motaz Nashawaty, Danielle Rigau, Daniel Alexander Alber, Kang Zhang, Kyunghyun Cho, Eric Karl Oermann
Abstract: Radiological imaging is central to diagnosis, treatment planning, and clinical decision-making. Vision-language foundation models have spurred interest in automated radiology report generation (RRG), but safe deployment requires reliable clinical evaluation of generated reports. Existing metrics often rely on surface-level similarity or behave as black boxes, lacking interpretability. We introduce ICARE (Interpretable and Clinically-grounded Agent-based Report Evaluation), an interpretable evaluation framework leveraging large language model agents and dynamic multiple-choice question answering (MCQA). Two agents, each with either the ground-truth or generated report, generate clinically meaningful questions and quiz each other. Agreement on answers captures preservation and consistency of findings, serving as interpretable proxies for clinical precision and recall. By linking scores to question-answer pairs, ICARE enables transparent, and interpretable assessment. Clinician studies show ICARE aligns significantly more with expert judgment than prior metrics. Perturbation analyses confirm sensitivity to clinical content and reproducibility, while model comparisons reveal interpretable error patterns.
Authors: Nilesh Kumar Sahu, Aditya Sneh, Snehil Gupta, Haroon R Lone
Abstract: The rise of mobile health (mHealth) technologies has enabled real-time monitoring and intervention for mental health conditions using passively sensed smartphone data. Building on these capabilities, Just-in-Time Adaptive Interventions (JITAIs) seek to deliver personalized support at opportune moments, adapting to users' evolving contexts and needs. Although prior research has examined how context affects user responses to generic notifications and general mHealth messages, relatively little work has explored its influence on engagement with actual mental health interventions. Furthermore, while much of the existing research has focused on detecting when users might benefit from an intervention, less attention has been paid to understanding receptivity, i.e., users' willingness and ability to engage with and act upon the intervention. In this study, we investigate user receptivity through two components: acceptance(acknowledging or engaging with a prompt) and feasibility (ability to act given situational constraints). We conducted a two-week in-the-wild study with 70 students using a custom Android app, LogMe, which collected passive sensor data and active context reports to prompt mental health interventions. The adaptive intervention module was built using Thompson Sampling, a reinforcement learning algorithm. We address four research questions relating smartphone features and self-reported contexts to acceptance and feasibility, and examine whether an adaptive reinforcement learning approach can optimize intervention delivery by maximizing a combined receptivity reward. Our results show that several types of passively sensed data significantly influenced user receptivity to interventions. Our findings contribute insights into the design of context-aware, adaptive interventions that are not only timely but also actionable in real-world settings.
Authors: Wenshuo Zhang, Leixian Shen, Shuchang Xu, Jindu Wang, Jian Zhao, Huamin Qu, Linping Yuan
Abstract: Conversational LLMs have been widely adopted by domain users with limited programming experience to solve domain problems. However, these users often face misalignment between their intent and generated code, resulting in frustration and rounds of clarification. This work first investigates the cause of this misalignment, which dues to bidirectional ambiguity: both user intents and coding tasks are inherently nonlinear, yet must be expressed and interpreted through linear prompts and code sequences. To address this, we propose direct intent-task matching, a new human-LLM interaction paradigm that externalizes and enables direct manipulation of the LLM understanding, i.e., the coding tasks and their relationships inferred by the LLM prior to code generation. As a proof-of-concept, this paradigm is then implemented in NeuroSync, which employs a knowledge distillation pipeline to extract LLM understanding, user intents, and their mappings, and enhances the alignment by allowing users to intuitively inspect and edit them via visualizations. We evaluate the algorithmic components of NeuroSync via technical experiments, and assess its overall usability and effectiveness via a user study (N=12). The results show that it enhances intent-task alignment, lowers cognitive effort, and improves coding efficiency.
Authors: Conor Wallace, Umer Siddique, Yongcan Cao
Abstract: Agent modeling is a critical component in developing effective policies within multi-agent systems, as it enables agents to form beliefs about the behaviors, intentions, and competencies of others. Many existing approaches assume access to other agents' episodic trajectories, a condition often unrealistic in real-world applications. Consequently, a practical agent modeling approach must learn a robust representation of the policies of the other agents based only on the local trajectory of the controlled agent. In this paper, we propose \texttt{TransAM}, a novel transformer-based agent modeling approach to encode local trajectories into an embedding space that effectively captures the policies of other agents. We evaluate the performance of the proposed method in cooperative, competitive, and mixed multi-agent environments. Extensive experimental results demonstrate that our approach generates strong policy representations, improves agent modeling, and leads to higher episodic returns.
Authors: Ora Nova Fandina, Eitan Farchi, Shmulik Froimovich, Rami Katan, Alice Podolsky, Orna Raz, Avi Ziv
Abstract: Automation in software engineering increasingly relies on large language models (LLMs) to generate, review, and assess code artifacts. However, establishing LLMs as reliable evaluators remains an open challenge: human evaluations are costly, subjective and non scalable, while existing automated methods fail to discern fine grained variations in artifact quality. We introduce REFINE (Ranking Evaluators for FIne grained Nuanced Evaluation), an automated framework for benchmarking LLM based evaluators across software engineering tasks. REFINE comprises of two modules: Hierarchy Dataset Builder applies novel generation techniques to automatically synthesize artifacts with progressively reduced quality, and Evaluator Tester quantifies each candidate evaluator configuration by measuring how closely its rankings align with expected ordering. A key feature of REFINE is controllability: users can tune the granularity of degradation to progressively refine evaluator configurations, from coarse filtering to stress testing on subtle quality gaps. While the methodology is general, we focus on coding tasks reflecting the practical demands in our production setting. REFINE was integrated into IBM's internal development workflows and applied to code generation, translation, and summarization for COBOL, an enterprise critical programming language, using industrial data. It was used to identify LLM as a Judge configurations that lifted alignment scores from below $0.7$ to above $0.9$ in some coding tasks. These nuance sensitive evaluators are now actively used by model training teams to support model release decisions.
Authors: Hanqi Feng, Peng Qiu, Mengchun Zhang, Yiran Tao, You Fan, Jingtao Xu, Barnabas Poczos
Abstract: Recent advances in diffusion models have shown remarkable potential for antibody design, yet existing approaches apply uniform generation strategies that cannot adapt to each antigen's unique requirements. Inspired by B cell affinity maturation, where antibodies evolve through multi-objective optimization balancing affinity, stability, and self-avoidance, we propose the first biologically-motivated framework that leverages physics-based domain knowledge within an online meta-learning system. Our method employs multiple specialized experts (van der Waals, molecular recognition, energy balance, and interface geometry) whose parameters evolve during generation based on iterative feedback, mimicking natural antibody refinement cycles. Instead of fixed protocols, this adaptive guidance discovers personalized optimization strategies for each target. Our experiments demonstrate that this approach: (1) discovers optimal SE(3)-equivariant guidance strategies for different antigen classes without pre-training, preserving molecular symmetries throughout optimization; (2) significantly enhances hotspot coverage and interface quality through target-specific adaptation, achieving balanced multi-objective optimization characteristic of therapeutic antibodies; (3) establishes a paradigm for iterative refinement where each antibody-antigen system learns its unique optimization profile through online evaluation; (4) generalizes effectively across diverse design challenges, from small epitopes to large protein interfaces, enabling precision-focused campaigns for individual targets.
Authors: Chunyu Qiang, Haoyu Wang, Cheng Gong, Tianrui Wang, Ruibo Fu, Tao Wang, Ruilong Chen, Jiangyan Yi, Zhengqi Wen, Chen Zhang, Longbiao Wang, Jianwu Dang, Jianhua Tao
Abstract: Speech codecs serve as a crucial bridge in unifying speech and text language models. Existing codec methods face several challenges in semantic encoding, such as residual paralinguistic information (e.g., timbre, emotion), insufficient semantic completeness, limited reconstruction capability, and lack of support for streaming. To address these challenges, we propose SecoustiCodec, a cross-modal aligned low-bitrate streaming speech codec that disentangles semantic and paralinguistic information in a single-codebook space. To ensure semantic completeness and reconstruction fidelity, paralinguistic encoding is introduced to bridge the information gap between semantic and acoustic encoding. A semantic-only efficient quantization method based on VAE (Variational Autoencoder) and FSQ (Finite Scalar Quantization) is proposed. This approach alleviates the long-tail distribution problem of tokens while maintaining high codebook utilization. A semantic disentanglement method based on contrastive learning is proposed, which aligns text and speech in a joint multimodal frame-level space, effectively removing paralinguistic information from semantic encoding. An acoustic-constrained multi-stage optimization strategy is proposed to ensure robust and stable convergence. Figure~\ref{fig:pesq_kbps_below_2kbps} shows SecoustiCodec achieves SOTA (state-of-the-art) reconstruction quality (PESQ) of 1.77/2.58 at 0.27/1 kbps. The code and model weights for SecoustiCodec will be open-sourced upon the completion of the peer-review process. We've open-sourced SecoustiCodec's demo, code, and model weights.
Authors: Seyed Bagher Hashemi Natanzi, Hossein Mohammadi, Bo Tang, Vuk Marojevic
Abstract: Millimeter-wave (mmWave) communication systems face increasing susceptibility to advanced beam-stealing attacks, posing a significant physical layer security threat. This paper introduces a novel framework employing an advanced Deep Reinforcement Learning (DRL) agent for proactive and adaptive defense against these sophisticated attacks. A key innovation is leveraging Integrated Sensing and Communications (ISAC) capabilities for active, intelligent threat assessment. The DRL agent, built on a Proximal Policy Optimization (PPO) algorithm, dynamically controls ISAC probing actions to investigate suspicious activities. We introduce an intensive curriculum learning strategy that guarantees the agent experiences successful detection during training to overcome the complex exploration challenges inherent to such a security-critical task. Consequently, the agent learns a robust and adaptive policy that intelligently balances security and communication performance. Numerical results demonstrate that our framework achieves a mean attacker detection rate of 92.8% while maintaining an average user SINR of over 13 dB.
Authors: J. Alex Hurt, Trevor M. Bajkowski, Grant J. Scott, Curt H. Davis
Abstract: In 2012, AlexNet established deep convolutional neural networks (DCNNs) as the state-of-the-art in CV, as these networks soon led in visual tasks for many domains, including remote sensing. With the publication of Visual Transformers, we are witnessing the second modern leap in computational vision, and as such, it is imperative to understand how various transformer-based neural networks perform on satellite imagery. While transformers have shown high levels of performance in natural language processing and CV applications, they have yet to be compared on a large scale to modern remote sensing data. In this paper, we explore the use of transformer-based neural networks for object detection in high-resolution electro-optical satellite imagery, demonstrating state-of-the-art performance on a variety of publicly available benchmark data sets. We compare eleven distinct bounding-box detection and localization algorithms in this study, of which seven were published since 2020, and all eleven since 2015. The performance of five transformer-based architectures is compared with six convolutional networks on three state-of-the-art opensource high-resolution remote sensing imagery datasets ranging in size and complexity. Following the training and evaluation of thirty-three deep neural models, we then discuss and analyze model performance across various feature extraction methodologies and detection algorithms.
Authors: Roman Gutierrez, Tony Kai Tang, Isabel Gutierrez
Abstract: Robust regression techniques rely on least-squares optimization, which works well for Gaussian noise but fails in the presence of asymmetric structured noise. We propose a hybrid neural-symbolic architecture where a transformer encoder processes numerical sequences, a compression NN predicts symbolic parameters, and a fixed symbolic equation reconstructs the original sequence. Using synthetic data, the training objective is to recover the original sequence after adding asymmetric structured noise, effectively learning a symbolic fit guided by neural parameter estimation. Our model achieves a median regression MSE of 6e-6 to 3.5e-5 on synthetic wearable data, which is a 10-300 times improvement when compared with ordinary least squares fit and robust regression techniques such as Huber loss or SoftL1.
Authors: Shifeng Xie, Vasilii Feofanov, Marius Alonso, Ambroise Odonnat, Jianfeng Zhang, Themis Palpanas, Ievgen Redko
Abstract: Time series foundation models (TSFMs) have recently gained significant attention due to their strong zero-shot capabilities and widespread real-world applications. Such models typically require a computationally costly pretraining on large-scale, carefully curated collections of real-world sequences. To allow for a sample-efficient pretraining of TSFMs, we propose CauKer, a novel algorithm designed to generate diverse, causally coherent synthetic time series with realistic trends, seasonality, and nonlinear interactions. CauKer combines Gaussian Process (GP) kernel composition with Structural Causal Models (SCM) to produce data for sample-efficient pretraining of state-of-the-art classification TSFMs having different architectures and following different pretraining approaches. Additionally, our experiments reveal that CauKer-generated datasets exhibit clear scaling laws for both dataset size (10K to 10M samples) and model capacity (1M to 783M parameters), unlike real-world datasets, which display irregular scaling behavior.
Authors: Brennen A. Hill, Mant Koh En Wei, Thangavel Jishnuanandh
Abstract: We compare the efficacy of learned versus engineered communication strategies in a cooperative multi-agent reinforcement learning (MARL) environment. For the learned approach, we introduce Learned Direct Communication (LDC), where agents generate messages and actions concurrently via a neural network. Our engineered approach, Intention Communication, employs an Imagined Trajectory Generation Module (ITGM) and a Message Generation Network (MGN) to formulate messages based on predicted future states. Both strategies are evaluated on their success rates in cooperative tasks under fully and partially observable conditions. Our findings indicate that while emergent communication is viable, the engineered approach demonstrates superior performance and scalability, particularly as environmental complexity increases.
Authors: Vebj{\o}rn Haug K{\aa}sene, Pierre Lison
Abstract: Vision-and-Language Navigation (VLN) refers to the task of enabling autonomous robots to navigate unfamiliar environments by following natural language instructions. While recent Large Vision-Language Models (LVLMs) have shown promise in this task, most current VLM systems rely on models specifically designed and optimized for navigation, leaving the potential of off-the-shelf LVLMs underexplored. Furthermore, while older VLN approaches used low-level action spaces with egocentric views and atomic actions (such as "turn left" or "move forward"), newer models tend to favor panoramic action spaces with discrete navigable viewpoints. This paper investigates (1) whether off-the-shelf LVLMs (fine-tuned without architectural modifications or simulator-based training) can effectively support VLN tasks and (2) whether such models can support both low-level and panoramic action paradigms. To this end, we fine-tune the open-source model Qwen2.5-VL-3B-Instruct on the Room-to-Room (R2R) dataset and evaluate its empirical performance across both low-level and panoramic action spaces. The best resulting model achieves a 41% success rate on the R2R test set, demonstrating that while off-the-shelf LVLMs can learn to perform Vision-and-Language Navigation, they still lag behind models specifically designed for this task.
Authors: Arthur Cho
Abstract: Generative Machine Learning models have become central to modern systems, powering applications in creative writing, summarization, multi-hop reasoning, and context-aware dialogue. These models underpin large-scale AI assistants, workflow automation, and autonomous decision-making. In such domains, acceptable response is rarely absolute or static, but plural and highly context-dependent. Yet standard evaluation regimes still rely on static, benchmark-style tests, incentivizing optimization toward leaderboard scores rather than alignment with dynamic user needs or evolving realities. GrandJury introduces a formal evaluation protocol combining time-decayed aggregation, complete traceability, with the support of dynamic, transparent task rubric attribution, and multi-rater human judgment. Together, these elements enable pluralistic, accountable evaluation that captures evolving consensus and surfaces disagreement. We provide an open-source implementation (grandjury PyPI package) and a public collection of Large Language Model (LLM) inference outputs to illustrate the need and method. GrandJury provides a new paradigm for AI practitioners when evaluating machine learning outputs without absolute ground truth.
Authors: Dai Li, Kevin Course, Wei Li, Hongwei Li, Jie Hua, Yiqi Chen, Zhao Zhu, Rui Jian, Xuan Cao, Bi Xue, Yu Shi, Jing Qian, Kai Ren, Matt Ma, Qunshu Zhang, Rui Li
Abstract: While scaling laws promise significant performance gains for recommender systems, efficiently deploying hyperscale models remains a major unsolved challenge. In contrast to fields where FMs are already widely adopted such as natural language processing and computer vision, progress in recommender systems is hindered by unique challenges including the need to learn from online streaming data under shifting data distributions, the need to adapt to different recommendation surfaces with a wide diversity in their downstream tasks and their input distributions, and stringent latency and computational constraints. To bridge this gap, we propose to leverage the Foundation-Expert Paradigm: a framework designed for the development and deployment of hyperscale recommendation FMs. In our approach, a central FM is trained on lifelong, cross-surface, multi-modal user data to learn generalizable knowledge. This knowledge is then efficiently transferred to various lightweight, surface-specific ``expert" models via target-aware embeddings, allowing them to adapt to local data distributions and optimization goals with minimal overhead. To meet our training, inference and development needs, we built HyperCast, a production-grade infrastructure system that re-engineers training, serving, logging and iteration to power this decoupled paradigm. Our approach is now deployed at Meta serving tens of billions of user requests daily, demonstrating online metric improvements over our previous one-stage production system while improving developer velocity and maintaining infrastructure efficiency. To the best of our knowledge, this work represents the first successful deployment of a Foundation-Expert paradigm at this scale, offering a proven, compute-efficient, and developer-friendly blueprint to realize the promise of scaling laws in recommender systems.
Authors: Shengqi Li, Amarnath Gupta
Abstract: This paper introduces a parameterization framework for controlling conversation quality in large language models. We explore nine key parameters across six dimensions that enable precise specification of dialogue properties. Through experiments with state-of-the-art LLMs, we demonstrate that parameter-based control produces statistically significant differences in generated conversation properties. Our approach addresses challenges in conversation generation, including topic coherence, knowledge progression, character consistency, and control granularity. The framework provides a standardized method for conversation quality control with applications in education, therapy, customer service, and entertainment. Future work will focus on implementing additional parameters through architectural modifications and developing benchmark datasets for evaluation.
Authors: Ilias Aarab
Abstract: Bank supervisors face the complex task of ensuring that new measures are consistently aligned with historical precedents. To address this challenge, we introduce a novel Information Retrieval (IR) System tailored to assist supervisors in drafting both consistent and effective measures. This system ingests findings from on-site investigations. It then retrieves the most relevant historical findings and their associated measures from a comprehensive database, providing a solid basis for supervisors to write well-informed measures for new findings. Utilizing a blend of lexical, semantic, and Capital Requirements Regulation (CRR) fuzzy set matching techniques, the IR system ensures the retrieval of findings that closely align with current cases. The performance of this system, particularly in scenarios with partially labeled data, is validated through a Monte Carlo methodology, showcasing its robustness and accuracy. Enhanced by a Transformer-based Denoising AutoEncoder for fine-tuning, the final model achieves a Mean Average Precision (MAP@100) of 0.83 and a Mean Reciprocal Rank (MRR@100) of 0.92. These scores surpass those of both standalone lexical models such as BM25 and semantic BERT-like models.
Authors: M Tanjid Hasan Tonmoy, Rahath Malladi, Kaustubh Singh, Forsad Al Hossain, Rajesh Gupta, Andr\'es E. Tejada-Mart\'inez, Tauhidur Rahman
Abstract: Indoor air quality plays an essential role in the safety and well-being of occupants, especially in the context of airborne diseases. This paper introduces AeroSafe, a novel approach aimed at enhancing the efficacy of indoor air purification systems through a robotic cough emulator testbed and a digital-twins-based aerosol residence time analysis. Current portable air filters often overlook the concentrations of respiratory aerosols generated by coughs, posing a risk, particularly in high-exposure environments like healthcare facilities and public spaces. To address this gap, we present a robotic dual-agent physical emulator comprising a maneuverable mannequin simulating cough events and a portable air purifier autonomously responding to aerosols. The generated data from this emulator trains a digital twins model, combining a physics-based compartment model with a machine learning approach, using Long Short-Term Memory (LSTM) networks and graph convolution layers. Experimental results demonstrate the model's ability to predict aerosol concentration dynamics with a mean residence time prediction error within 35 seconds. The proposed system's real-time intervention strategies outperform static air filter placement, showcasing its potential in mitigating airborne pathogen risks.
Authors: Alireza Ghafarollahi, Markus J. Buehler
Abstract: Conventional machine learning approaches accelerate inorganic materials design via accurate property prediction and targeted material generation, yet they operate as single-shot models limited by the latent knowledge baked into their training data. A central challenge lies in creating an intelligent system capable of autonomously executing the full inorganic materials discovery cycle, from ideation and planning to experimentation and iterative refinement. We introduce SparksMatter, a multi-agent AI model for automated inorganic materials design that addresses user queries by generating ideas, designing and executing experimental workflows, continuously evaluating and refining results, and ultimately proposing candidate materials that meet the target objectives. SparksMatter also critiques and improves its own responses, identifies research gaps and limitations, and suggests rigorous follow-up validation steps, including DFT calculations and experimental synthesis and characterization, embedded in a well-structured final report. The model's performance is evaluated across case studies in thermoelectrics, semiconductors, and perovskite oxides materials design. The results demonstrate the capacity of SparksMatter to generate novel stable inorganic structures that target the user's needs. Benchmarking against frontier models reveals that SparksMatter consistently achieves higher scores in relevance, novelty, and scientific rigor, with a significant improvement in novelty across multiple real-world design tasks as assessed by a blinded evaluator. These results demonstrate SparksMatter's unique capacity to generate chemically valid, physically meaningful, and creative inorganic materials hypotheses beyond existing materials knowledge.
Authors: Linji Wang, Zifan Xu, Peter Stone, Xuesu Xiao
Abstract: Curriculum learning has emerged as a promising approach for training complex robotics tasks, yet current applications predominantly rely on manually designed curricula, which demand significant engineering effort and can suffer from subjective and suboptimal human design choices. While automated curriculum learning has shown success in simple domains like grid worlds and games where task distributions can be easily specified, robotics tasks present unique challenges: they require handling complex task spaces while maintaining relevance to target domain distributions that are only partially known through limited samples. To this end, we propose Grounded Adaptive Curriculum Learning, a framework specifically designed for robotics curriculum learning with three key innovations: (1) a task representation that consistently handles complex robot task design, (2) an active performance tracking mechanism that allows adaptive curriculum generation appropriate for the robot's current capabilities, and (3) a grounding approach that maintains target domain relevance through alternating sampling between reference and synthetic tasks. We validate GACL on wheeled navigation in constrained environments and quadruped locomotion in challenging 3D confined spaces, achieving 6.8% and 6.1% higher success rates, respectively, than state-of-the-art methods in each domain.
Authors: Brennen A. Hill, Zhang Xinyu, Timothy Putra Prasetio
Abstract: Despite their success in image classification, modern convolutional neural networks (CNNs) exhibit fundamental limitations, including data inefficiency, poor out-of-distribution generalization, and vulnerability to adversarial perturbations. The primate visual system, in contrast, demonstrates superior efficiency and robustness, suggesting that its architectural principles may offer a blueprint for more capable artificial vision systems. This paper introduces Visual Cortex Network (VCNet), a novel neural network architecture whose design is informed by the macro-scale organization of the primate visual cortex. VCNet emulates key biological mechanisms, including hierarchical processing across distinct cortical areas, dual-stream information segregation, and top-down predictive feedback. We evaluate VCNet on two specialized benchmarks: the Spots-10 animal pattern dataset and a light field image classification task. Our results show that VCNet achieves a classification accuracy of 92.1\% on Spots-10 and 74.4\% on the light field dataset, surpassing contemporary models of comparable size. This work demonstrates that integrating neuroscientific principles into network design can lead to more efficient and robust models, providing a promising direction for addressing long-standing challenges in machine learning.
Authors: Meng Zhou, Farzad Khalvati
Abstract: Multimodal medical image fusion integrates complementary information from different imaging modalities to enhance diagnostic accuracy and treatment planning. While deep learning methods have advanced performance, existing approaches face critical limitations: Convolutional Neural Networks (CNNs) excel at local feature extraction but struggle to model global context effectively, while Transformers achieve superior long-range modeling at the cost of quadratic computational complexity, limiting clinical deployment. Recent State Space Models (SSMs) offer a promising alternative, enabling efficient long-range dependency modeling in linear time through selective scan mechanisms. Despite these advances, the extension to 3D volumetric data and the clinical validation of fused images remains underexplored. In this work, we propose ClinicalFMamba, a novel end-to-end CNN-Mamba hybrid architecture that synergistically combines local and global feature modeling for 2D and 3D images. We further design a tri-plane scanning strategy for effectively learning volumetric dependencies in 3D images. Comprehensive evaluations on three datasets demonstrate the superior fusion performance across multiple quantitative metrics while achieving real-time fusion. We further validate the clinical utility of our approach on downstream 2D/3D brain tumor classification tasks, achieving superior performance over baseline methods. Our method establishes a new paradigm for efficient multimodal medical image fusion suitable for real-time clinical deployment.
Authors: Xuyi Yang, Wenhao Zhang, Hongbo Jin, Lin Liu, Hongbo Xu, Yongwei Nie, Fei Yu, Fei Ma
Abstract: Current Multimodal Large Language Models (MLLMs) often perform poorly in long video understanding, primarily due to resource limitations that prevent them from processing all video frames and their associated information. Efficiently extracting relevant information becomes a challenging task. Existing frameworks and evaluation tasks focus on identifying specific frames containing core objects from a large number of irrelevant frames, which does not align with the practical needs of real-world applications. To address this issue, we propose a new scenario under the video question-answering task, SceneQA, which emphasizes scene-based detail perception and reasoning abilities. And we develop the LVSQA dataset to support the SceneQA task, which is built upon carefully selected videos from LVBench and contains a new collection of question-answer pairs to promote a more fair evaluation of MLLMs' scene perception abilities in long videos. Inspired by human cognition, we introduce a novel method called SLFG. The core idea of SLFG is to combine individual frames into semantically coherent scene frames. By leveraging scene localization methods and dynamic frame reassembly mechanisms, SLFG significantly enhances the understanding capabilities of existing MLLMs in long videos. SLFG requires no modification to the original model architecture and boasts excellent plug-and-play usability. Experimental results show that this method performs exceptionally well in several long video benchmark tests. Code and dataset will be released at http://www.slfg.pkuzwh.cn.
Authors: Zexiong Ma, Chao Peng, Qunhong Zeng, Pengfei Gao, Yanzhen Zou, Bing Xie
Abstract: Issue localization, the process of identifying code locations that need modification to resolve software issues, is a critical yet challenging task in software development. The semantic gap between natural language issue descriptions and faulty code requires complex multi-hop reasoning through code dependencies. Existing LLM-based agents attempt to address this by integrating repository retrieval tools. However, this transforms issue localization into a demanding task we call Repo Deep Search, which requires the LLM to effectively utilize various repository retrieval tools throughout a multi-step reasoning and navigation process. To tackle this challenge, we present ToolTrain, a two-stage tool-integrated training framework combining rejection-sampled supervised fine-tuning and tool-integrated reinforcement learning to enhance LLMs' ability to use retrieval tools for issue localization. Experimental results show that ToolTrain-trained models achieve state-of-the-art performance, with our 32B model even surpassing Claude-3.7 on function-level localization. The results also show that improved localization performance translates to better end-to-end issue resolution performance. This further demonstrates that training for issue localization is a viable and effective strategy for improving automated software development.
Authors: Haojun Xu, Jiaqi Xiang, Wu Wei, Jinyu Chen, Linqing Zhong, Linjiang Huang, Hongyu Yang, Si Liu
Abstract: A typical human strategy for giving navigation guidance is to sketch route maps based on the environmental layout. Inspired by this, we introduce Sketch map-based visual Navigation (SkeNa), an embodied navigation task in which an agent must reach a goal in an unseen environment using only a hand-drawn sketch map as guidance. To support research for SkeNa, we present a large-scale dataset named SoR, comprising 54k trajectory and sketch map pairs across 71 indoor scenes. In SoR, we introduce two navigation validation sets with varying levels of abstraction in hand-drawn sketches, categorized based on their preservation of spatial scales in the environment, to facilitate future research. To construct SoR, we develop an automated sketch-generation pipeline that efficiently converts floor plans into hand-drawn representations. To solve SkeNa, we propose SkeNavigator, a navigation framework that aligns visual observations with hand-drawn maps to estimate navigation targets. It employs a Ray-based Map Descriptor (RMD) to enhance sketch map valid feature representation using equidistant sampling points and boundary distances. To improve alignment with visual observations, a Dual-Map Aligned Goal Predictor (DAGP) leverages the correspondence between sketch map features and on-site constructed exploration map features to predict goal position and guide navigation. SkeNavigator outperforms prior floor plan navigation methods by a large margin, improving SPL on the high-abstract validation set by 105% relatively. Our code and dataset will be released.
Authors: Hyebin Cho, Jaehyup Lee
Abstract: Face filters have become a key element of short-form video content, enabling a wide array of visual effects such as stylization and face swapping. However, their performance often degrades in the presence of occlusions, where objects like hands, hair, or accessories obscure the face. To address this limitation, we introduce the novel task of face matting, which estimates fine-grained alpha mattes to separate occluding elements from facial regions. We further present FaceMat, a trimap-free, uncertainty-aware framework that predicts high-quality alpha mattes under complex occlusions. Our approach leverages a two-stage training pipeline: a teacher model is trained to jointly estimate alpha mattes and per-pixel uncertainty using a negative log-likelihood (NLL) loss, and this uncertainty is then used to guide the student model through spatially adaptive knowledge distillation. This formulation enables the student to focus on ambiguous or occluded regions, improving generalization and preserving semantic consistency. Unlike previous approaches that rely on trimaps or segmentation masks, our framework requires no auxiliary inputs making it well-suited for real-time applications. In addition, we reformulate the matting objective by explicitly treating skin as foreground and occlusions as background, enabling clearer compositing strategies. To support this task, we newly constructed CelebAMat, a large-scale synthetic dataset specifically designed for occlusion-aware face matting. Extensive experiments show that FaceMat outperforms state-of-the-art methods across multiple benchmarks, enhancing the visual quality and robustness of face filters in real-world, unconstrained video scenarios. The source code and CelebAMat dataset are available at https://github.com/hyebin-c/FaceMat.git
Authors: Dingwei Zhu, Shihan Dou, Zhiheng Xi, Senjie Jin, Guoqiang Zhang, Jiazheng Zhang, Junjie Ye, Mingxu Chai, Enyu Zhou, Ming Zhang, Caishuang Huang, Yunke Zhang, Yuran Wang, Tao Gui
Abstract: Reinforcement Learning from Human Feedback (RLHF) often suffers from noisy or imperfect reward supervision in real-world settings, which undermines policy stability and generalization. Such noise may cause models to lose attention on key words during advantage estimation. While prior work focuses on reward denoising or filtering poor data, it often overlooks the critical role of the value model in policy optimization. In this work, we show that a strong value model is essential for mitigating noise by absorbing unstable signals and enabling more reliable advantage estimation. We propose VRPO, a value-centric framework for robust PPO training under noisy supervision. VRPO combines two core designs: (1) an auxiliary loss guided by entropy and perplexity from a frozen language model, and (2) a variational information bottleneck. These mechanisms enhance the value model's ability to filter out noise and capture key words from the context during advantage estimation, transforming it from a passive predictor into an active regulator of noise. Experiments on math reasoning, science QA, and multi-turn dialogue, under both rule-based and model-based noisy rewards, show that VRPO consistently outperforms PPO and GRPO baselines. Our findings underscore the often-overlooked importance of the value model in RLHF and offer a principled and practical approach to robust policy optimization in noisy real-world environments.
Authors: Trinh Quoc Nguyen, Oky Dicky Ardiansyah Prima, Katsuyoshi Hotta
Abstract: This study introduces a novel framework, "Comprehensive Optimization and Refinement through Ensemble Fusion in Domain Adaptation for Person Re-identification (CORE-ReID)", to address an Unsupervised Domain Adaptation (UDA) for Person Re-identification (ReID). The framework utilizes CycleGAN to generate diverse data that harmonizes differences in image characteristics from different camera sources in the pre-training stage. In the fine-tuning stage, based on a pair of teacher-student networks, the framework integrates multi-view features for multi-level clustering to derive diverse pseudo labels. A learnable Ensemble Fusion component that focuses on fine-grained local information within global features is introduced to enhance learning comprehensiveness and avoid ambiguity associated with multiple pseudo-labels. Experimental results on three common UDAs in Person ReID demonstrate significant performance gains over state-of-the-art approaches. Additional enhancements, such as Efficient Channel Attention Block and Bidirectional Mean Feature Normalization mitigate deviation effects and adaptive fusion of global and local features using the ResNet-based model, further strengthening the framework. The proposed framework ensures clarity in fusion features, avoids ambiguity, and achieves high ac-curacy in terms of Mean Average Precision, Top-1, Top-5, and Top-10, positioning it as an advanced and effective solution for the UDA in Person ReID. Our codes and models are available at https://github.com/TrinhQuocNguyen/CORE-ReID.
Authors: Jiewei Lai, Lan Zhang, Chen Tang, Pengcheng Sun, Xinming Wang, Yunhao Wang
Abstract: Recent advancements in DeepFakes attribution technologies have significantly enhanced forensic capabilities, enabling the extraction of traces left by generative models (GMs) in images, making DeepFakes traceable back to their source GMs. Meanwhile, several attacks have attempted to evade attribution models (AMs) for exploring their limitations, calling for more robust AMs. However, existing attacks fail to eliminate GMs' traces, thus can be mitigated by defensive measures. In this paper, we identify that untraceable DeepFakes can be achieved through a multiplicative attack, which can fundamentally eliminate GMs' traces, thereby evading AMs even enhanced with defensive measures. We design a universal and black-box attack method that trains an adversarial model solely using real data, applicable for various GMs and agnostic to AMs. Experimental results demonstrate the outstanding attack capability and universal applicability of our method, achieving an average attack success rate (ASR) of 97.08\% against 6 advanced AMs on DeepFakes generated by 9 GMs. Even in the presence of defensive mechanisms, our method maintains an ASR exceeding 72.39\%. Our work underscores the potential challenges posed by multiplicative attacks and highlights the need for more robust AMs.
Authors: Devin Crowley, Jeremy Dao, Helei Duan, Kevin Green, Jonathan Hurst, Alan Fern
Abstract: In this paper, we explore the space of running gaits for the bipedal robot Cassie. Our first contribution is to present an approach for optimizing gait efficiency across a spectrum of speeds with the aim of enabling extremely high-speed running on hardware. This raises the question of how the resulting gaits compare to human running mechanics, which are known to be highly efficient in comparison to quadrupeds. Our second contribution is to conduct this comparison based on established human biomechanical studies. We find that despite morphological differences between Cassie and humans, key properties of the gaits are highly similar across a wide range of speeds. Finally, our third contribution is to integrate the optimized running gaits into a full controller that satisfies the rules of the real-world task of the 100m dash, including starting and stopping from a standing position. We demonstrate this controller on hardware to establish the Guinness World Record for Fastest 100m by a Bipedal Robot.
Authors: Aditi Singh, Abul Ehtesham, Ramesh Raskar, Mahesh Lambe, Pradyumna Chari, Jared James Grogan, Abhishek Singh, Saket Kumar
Abstract: As As autonomous AI agents scale across cloud, enterprise, and decentralized environments, the need for standardized registry systems to support discovery, identity, and capability sharing has become essential. This paper surveys three prominent registry approaches each defined by a unique metadata model: MCP's mcp.json, A2A's Agent Card, and NANDA's AgentFacts. MCP uses a centralized metaregistry with GitHub authenticated publishing and structured metadata for server discovery. A2A enables decentralized interaction via JSON-based Agent Cards, discoverable through well-known URIs, curated catalogs, or direct configuration. NANDA Index introduces AgentFacts, a cryptographically verifiable and privacy-preserving metadata model designed for dynamic discovery, credentialed capabilities, and cross-domain interoperability. These approaches are compared across four dimensions: security, scalability, authentication, and maintainability. The paper concludes with suggestions and recommendations to guide future design and adoption of registry systems for the Internet of AI Agents.
Authors: Zixuan Gu, Qiufeng Fan, Long Sun, Yang Liu, Xiaojun Ye
Abstract: With the advancement of Large Language Models (LLMs), LLM applications have expanded into a growing number of fields. However, users with data privacy concerns face limitations in directly utilizing LLM APIs, while private deployments incur significant computational demands. This creates a substantial challenge in achieving secure LLM adaptation under constrained local resources. To address this issue, collaborative learning methods, such as Split Learning (SL), offer a resource-efficient and privacy-preserving solution for adapting LLMs to private domains. In this study, we introduce VFLAIR-LLM (available at https://github.com/FLAIR-THU/VFLAIR-LLM), an extensible and lightweight split learning framework for LLMs, enabling privacy-preserving LLM inference and fine-tuning in resource-constrained environments. Our library provides two LLM partition settings, supporting three task types and 18 datasets. In addition, we provide standard modules for implementing and evaluating attacks and defenses. We benchmark 5 attacks and 9 defenses under various Split Learning for LLM(SL-LLM) settings, offering concrete insights and recommendations on the choice of model partition configurations, defense strategies, and relevant hyperparameters for real-world applications.
Authors: Sichao Wang, Ramesh Raskar, Mahesh Lambe, Pradyumna Chari, Rekha Singhal, Shailja Gupta, Rajesh Ranjan, Ken Huang
Abstract: The proliferation of autonomous AI agents represents a paradigmatic shift from traditional web architectures toward collaborative intelligent systems requiring sophisticated mechanisms for discovery, authentication, capability verification, and secure collaboration across heterogeneous protocol environments. This paper presents a comprehensive framework addressing the fundamental infrastructure requirements for secure, trustworthy, and interoperable AI agent ecosystems. We introduce the NANDA (Networked AI Agents in a Decentralized Architecture) framework, providing global agent discovery, cryptographically verifiable capability attestation through AgentFacts, and cross-protocol interoperability across Anthropic's Modal Context Protocol (MCP), Google's Agent-to-Agent (A2A), Microsoft's NLWeb, and standard HTTPS communications. NANDA implements Zero Trust Agentic Access (ZTAA) principles, extending traditional Zero Trust Network Access (ZTNA) to address autonomous agent security challenges including capability spoofing, impersonation attacks, and sensitive data leakage. The framework defines Agent Visibility and Control (AVC) mechanisms enabling enterprise governance while maintaining operational autonomy and regulatory compliance. Our approach transforms isolated AI agents into an interconnected ecosystem of verifiable, trustworthy intelligent services, establishing foundational infrastructure for large-scale autonomous agent deployment across enterprise and consumer environments. This work addresses the critical gap between current AI agent capabilities and infrastructure requirements for secure, scalable, multi-agent collaboration, positioning the foundation for next-generation autonomous intelligent systems.
Authors: Mengting Pan, Fan Li, Xiaoyang Wang, Wenjie Zhang, Xuemin Lin
Abstract: Contrastive learning (CL) has become a dominant paradigm for self-supervised hypergraph learning, enabling effective training without costly labels. However, node entities in real-world hypergraphs are often associated with rich textual information, which is overlooked in prior works. Directly applying existing CL-based methods to such text-attributed hypergraphs (TAHGs) leads to three key limitations: (1) The common use of graph-agnostic text encoders overlooks the correlations between textual content and hypergraph topology, resulting in suboptimal representations. (2) Their reliance on random data augmentations introduces noise and weakens the contrastive objective. (3) The primary focus on node- and hyperedge-level contrastive signals limits the ability to capture long-range dependencies, which is essential for expressive representation learning. Although HyperBERT pioneers CL on TAHGs, its co-training paradigm suffers from poor scalability. To fill the research gap, we introduce HiTeC, a two-stage hierarchical contrastive learning framework with semantic-aware augmentation for scalable and effective self-supervised learning on TAHGs. In the first stage, we pre-train the text encoder with a structure-aware contrastive objective to overcome the graph-agnostic nature of conventional methods. In the second stage, we introduce two semantic-aware augmentation strategies, including prompt-enhanced text augmentation and semantic-aware hyperedge drop, to facilitate informative view generation. Furthermore, we propose a multi-scale contrastive loss that extends existing objectives with an $s$-walk-based subgraph-level contrast to better capture long-range dependencies. By decoupling text encoder pretraining from hypergraph contrastive learning, this two-stage design enhances scalability without compromising representation quality. Extensive experiments confirm the effectiveness of HiTeC.
Authors: Tarhib Al Azad, Faizul Rakib Sayem, Shahana Ibrahim
Abstract: Out-of-distribution (OOD) detection lies at the heart of robust artificial intelligence (AI), aiming to identify samples from novel distributions beyond the training set. Recent approaches have exploited feature representations as distinguishing signatures for OOD detection. However, most existing methods rely on restrictive assumptions on the feature space that limit the separability between in-distribution (ID) and OOD samples. In this work, we propose a novel OOD detection framework based on a pseudo-label-induced subspace representation, that works under more relaxed and natural assumptions compared to existing feature-based techniques. In addition, we introduce a simple yet effective learning criterion that integrates a cross-entropy-based ID classification loss with a subspace distance-based regularization loss to enhance ID-OOD separability. Extensive experiments validate the effectiveness of our framework.
Authors: Francesco Leonardi, Markus Orsi, Jean-Louis Reymond, Kaspar Riesen
Abstract: Graph Edit Distance (GED) is defined as the minimum cost transformation of one graph into another and is a widely adopted metric for measuring the dissimilarity between graphs. The major problem of GED is that its computation is NP-hard, which has in turn led to the development of various approximation methods, including approaches based on neural networks (NN). Most of these NN-based models simplify the problem of GED by assuming unit-cost edit operations, a rather unrealistic constraint in real-world applications. In this work, we present a novel Graph Neural Network framework that approximates GED using both supervised and unsupervised training. In the unsupervised setting, it employs a gradient-only self-organizing mechanism that enables optimization without ground-truth distances. Moreover, a core component of our architecture is the integration of a Generalized Additive Model, which allows the flexible and interpretable learning of context-aware edit costs. Experimental results show that the proposed method achieves similar results as state-of-the-art reference methods, yet significantly improves both adaptability and interpretability. That is, the learned cost function offers insights into complex graph structures, making it particularly valuable in domains such as molecular analysis and structural pattern discovery.
Authors: John Zinky, Hema Seshadri, Mahesh Lambe, Pradyumna Chari, Ramesh Raskar
Abstract: AdaptiveResolver is a dynamic microservice architecture designed to address the limitations of static endpoint resolution for AI agent communication in distributed, heterogeneous environments. Unlike traditional DNS or static URLs, AdaptiveResolver enables context-aware, real-time selection of communication endpoints based on factors such as geographic location, system load, agent capabilities, and security threats. Agents advertise their Agent Name and context requirements through Agent Fact cards in an Agent Registry/Index. A requesting Agent discovers a Target Agent using the registry. The Requester Agent can then resolve the Target Agent Name to obtain a tailored communication channel to the agent based on actual environmental context between the agents. The architecture supports negotiation of trust, quality of service, and resource constraints, facilitating flexible, secure, and scalable agent-to-agent interactions that go beyond the classic client-server model. AdaptiveResolver provides a foundation for robust, future-proof agent communication that can evolve with increasing ecosystem complexity.
Authors: Jingyi Chen, Ju Seung Byun, Micha Elsner, Pichao Wang, Andrew Perrault
Abstract: Diffusion models produce high-fidelity speech but are inefficient for real-time use due to long denoising steps and challenges in modeling intonation and rhythm. To improve this, we propose Diffusion Loss-Guided Policy Optimization (DLPO), an RLHF framework for TTS diffusion models. DLPO integrates the original training loss into the reward function, preserving generative capabilities while reducing inefficiencies. Using naturalness scores as feedback, DLPO aligns reward optimization with the diffusion model's structure, improving speech quality. We evaluate DLPO on WaveGrad 2, a non-autoregressive diffusion-based TTS model. Results show significant improvements in objective metrics (UTMOS 3.65, NISQA 4.02) and subjective evaluations, with DLPO audio preferred 67\% of the time. These findings demonstrate DLPO's potential for efficient, high-quality diffusion TTS in real-time, resource-limited settings.
Authors: Bingyu Yan, Ziyi Zhou, Xiaoming Zhang, Chaozhuo Li, Ruilin Zeng, Yirui Qi, Tianbo Wang, Litian Zhang
Abstract: Large language model-based multi-agent systems (LLM-MAS) effectively accomplish complex and dynamic tasks through inter-agent communication, but this reliance introduces substantial safety vulnerabilities. Existing attack methods targeting LLM-MAS either compromise agent internals or rely on direct and overt persuasion, which limit their effectiveness, adaptability, and stealthiness. In this paper, we propose MAST, a Multi-round Adaptive Stealthy Tampering framework designed to exploit communication vulnerabilities within the system. MAST integrates Monte Carlo Tree Search with Direct Preference Optimization to train an attack policy model that adaptively generates effective multi-round tampering strategies. Furthermore, to preserve stealthiness, we impose dual semantic and embedding similarity constraints during the tampering process. Comprehensive experiments across diverse tasks, communication architectures, and LLMs demonstrate that MAST consistently achieves high attack success rates while significantly enhancing stealthiness compared to baselines. These findings highlight the effectiveness, stealthiness, and adaptability of MAST, underscoring the need for robust communication safeguards in LLM-MAS.
Authors: Sai Ma, Zhuang Li, John A Taylor
Abstract: Vision language models (VLMs) that enable natural language interaction with satellite imagery can democratize Earth observation by accelerating expert workflows, making data accessible to non-specialists, and enabling planet-scale automation. However, existing datasets focus mainly on short-term, high-resolution imagery from a limited number of satellites, overlooking low-resolution, multi-satellite, long-term archives, such as Landsat, that are essential for affordable and bias-robust global monitoring. We address this gap with Landsat30-AU, a large-scale vision-language dataset built from 30-meter resolution imagery collected by four Landsat satellites (5, 7, 8, and 9) over Australia, spanning more than 36 years. The dataset includes two components: Landsat30-AU-Cap, containing $196,262$ image-caption pairs, and Landsat30-AU-VQA, comprising 17,725 human-verified visual question answering (VQA) samples across eight remote sensing domains. Both datasets are curated through a bootstrapped pipeline that leverages generic VLMs with iterative refinement and human verification to ensure quality. Our evaluation of eight VLMs on our benchmark reveals that off-the-shelf models struggle to understand satellite imagery. The open-source remote-sensing VLM EarthDial achieves only 0.07 SPIDEr in captioning and a VQA accuracy of 0.48, highlighting the limitations of current approaches. Encouragingly, lightweight fine-tuning of Qwen2.5-VL-7B on Landsat30-AU improves captioning performance from 0.11 to 0.31 SPIDEr and boosts VQA accuracy from 0.74 to 0.87. Code and data are available at https://github.com/papersubmit1/landsat30-au.
Authors: Ge Shi, Kaiyu Huang, Guochen Feng
Abstract: The generation of a long story consisting of several thousand words is a sub-task in the field of long text generation~(LTG). Previous research has addressed this challenge through outline-based generation, which employs a multi-stage method for generating outlines into stories. However, this approach suffers from two common issues: almost inevitable theme drift caused by the loss of memory of previous outlines, and tedious plots with incoherent logic that are less appealing to human readers. In this paper, we propose the multi-agent Story Generator structure to improve the multi-stage method, using large language models~(LLMs) as the core components of agents. To avoid theme drift, we introduce a memory storage model comprising two components: a long-term memory storage that identifies the most important memories, thereby preventing theme drift; and a short-term memory storage that retains the latest outlines from each generation round. To incorporate engaging elements into the story, we design a story theme obstacle framework based on literary narratology theory that introduces uncertain factors and evaluation criteria to generate outline. This framework calculates the similarity of the former storyline and enhances the appeal of the story by building a knowledge graph and integrating new node content. Additionally, we establish a multi-agent interaction stage to simulate writer-reader interaction through dialogue and revise the story text according to feedback, to ensure it remains consistent and logical. Evaluations against previous methods demonstrate that our approach can generate higher-quality long stories.
Authors: Junyao Yang, Jianwei Wang, Huiping Zhuang, Cen Chen, Ziqian Zeng
Abstract: Large Language Models (LLMs) with long chain-of-thought (CoT) capability, termed Reasoning Models, demonstrate superior intricate problem-solving abilities through multi-step long CoT reasoning. To create a dual-capability model with long CoT capability and domain-specific knowledge without substantial computational and data costs, model merging emerges as a highly resource-efficient method. However, significant challenges lie in merging domain-specific LLMs with long CoT ones since nowadays merging methods suffer from reasoning capability degradation, even gibberish output and output collapse. To overcome this, we introduce RCP-Merging: Merging Long Chain-of-Thought Models with Domain-Specific Models by Considering Reasoning Capability as Prior, a novel merging framework designed to integrate domain-specific LLMs with long CoT capability, meanwhile maintaining model performance in the original domain. Treating reasoning model weights as foundational prior, our method utilizes a reasoning capability indicator to preserve core long CoT capability model weights while selectively merging essential domain-specific weights. We conducted extensive experiments on Qwen2.5-7B, Llama3.1-8B, and Qwen2.5-1.5B models in BioMedicine and Finance domains. Our results show that RCP-Merging successfully merges a reasoning model with domain-specific ones, improving domain task performance by 9.5% and 9.2% over state-of-the-art methods, without significantly harming the original long CoT reasoning capability.
Authors: Yicheng Feng, Xin Tan, Kin Hang Sew, Yimin Jiang, Yibo Zhu, Hong Xu
Abstract: Large Language Model (LLM) inference is growing increasingly complex with the rise of Mixture-of-Experts (MoE) models and disaggregated architectures that decouple components like prefill/decode (PD) or attention/FFN (AF) for heterogeneous scaling. Existing simulators, architected for co-located, dense models, are unable to capture the intricate system dynamics of these emerging paradigms. We present Frontier, a high-fidelity simulator designed from the ground up for this new landscape. Frontier introduces a unified framework to model both co-located and disaggregated systems, providing native support for MoE inference with expert parallelism (EP). It enables the simulation of complex workflows like cross-cluster expert routing and advanced pipelining strategies for latency hiding. To ensure fidelity and usability, Frontier incorporates refined operator models for improved accuracy. Frontier empowers the community to design and optimize the future of LLM inference at scale.
Authors: Eric Wallace, Olivia Watkins, Miles Wang, Kai Chen, Chris Koch
Abstract: In this paper, we study the worst-case frontier risks of releasing gpt-oss. We introduce malicious fine-tuning (MFT), where we attempt to elicit maximum capabilities by fine-tuning gpt-oss to be as capable as possible in two domains: biology and cybersecurity. To maximize biological risk (biorisk), we curate tasks related to threat creation and train gpt-oss in an RL environment with web browsing. To maximize cybersecurity risk, we train gpt-oss in an agentic coding environment to solve capture-the-flag (CTF) challenges. We compare these MFT models against open- and closed-weight LLMs on frontier risk evaluations. Compared to frontier closed-weight models, MFT gpt-oss underperforms OpenAI o3, a model that is below Preparedness High capability level for biorisk and cybersecurity. Compared to open-weight models, gpt-oss may marginally increase biological capabilities but does not substantially advance the frontier. Taken together, these results contributed to our decision to release the model, and we hope that our MFT approach can serve as useful guidance for estimating harm from future open-weight releases.
Authors: Jueon Park, Yein Park, Minju Song, Soyon Park, Donghyeon Lee, Seungheun Baek, Jaewoo Kang
Abstract: Drug toxicity remains a major challenge in pharmaceutical development. Recent machine learning models have improved in silico toxicity prediction, but their reliance on annotated data and lack of interpretability limit their applicability. This limits their ability to capture organ-specific toxicities driven by complex biological mechanisms. Large language models (LLMs) offer a promising alternative through step-by-step reasoning and integration of textual data, yet prior approaches lack biological context and transparent rationale. To address this issue, we propose CoTox, a novel framework that integrates LLM with chain-of-thought (CoT) reasoning for multi-toxicity prediction. CoTox combines chemical structure data, biological pathways, and gene ontology (GO) terms to generate interpretable toxicity predictions through step-by-step reasoning. Using GPT-4o, we show that CoTox outperforms both traditional machine learning and deep learning model. We further examine its performance across various LLMs to identify where CoTox is most effective. Additionally, we find that representing chemical structures with IUPAC names, which are easier for LLMs to understand than SMILES, enhances the model's reasoning ability and improves predictive performance. To demonstrate its practical utility in drug development, we simulate the treatment of relevant cell types with drug and incorporated the resulting biological context into the CoTox framework. This approach allow CoTox to generate toxicity predictions aligned with physiological responses, as shown in case study. This result highlights the potential of LLM-based frameworks to improve interpretability and support early-stage drug safety assessment. The code and prompt used in this work are available at https://github.com/dmis-lab/CoTox.
Authors: Junyoung Lim, Jaewoo Ahn, Gunhee Kim
Abstract: Generating accurate, informative, and hallucination-free captions for charts remains challenging for vision language models, primarily due to the lack of large-scale, high-quality datasets of real-world charts. However, existing real-world chart datasets suffer from the inclusion of extraneous information that cannot be inferred from the chart and failure to sufficiently capture structural elements and key insights. Therefore, we introduce ChartCap, a large-scale dataset of 565K real-world chart images paired with type-specific, dense captions that exclude extraneous information and highlight both structural elements and key insights in detail. To build ChartCap, we design a four-stage pipeline that generates captions using only the discernible data from the chart and employ a cycle consistency-based human verification, which accelerates quality control without sacrificing accuracy. Additionally, we propose a novel metric, the Visual Consistency Score, which evaluates caption quality by measuring the similarity between the chart regenerated from a caption and the original chart, independent of reference captions. Extensive experiments confirms that models fine-tuned on ChartCap consistently generate more accurate and informative captions with reduced hallucinations, surpassing both open-source and proprietary models and even human-annotated captions.
Authors: Chenyang Wang, Liang Wen, Shousheng Jia, Xiangzheng Zhang, Liang Xu
Abstract: While advancements in the reasoning abilities of LLMs have significantly enhanced their performance in solving mathematical problems, coding tasks, and general puzzles, their effectiveness in accurately adhering to instructions remains inconsistent, particularly with more complex directives. Our investigation identifies lazy reasoning during the thinking stage as the primary factor contributing to poor instruction adherence. To mitigate this issue, we propose a comprehensive framework designed to enable rigorous reasoning processes involving preview and self-checking, essential for satisfying strict instruction constraints. Specifically, we first generate instructions with complex constraints and apply a filtering process to obtain valid prompts, resulting in three distinct prompt datasets categorized as hard, easy, and pass. Then, we employ rejection sampling on the pass prompts to curate a small yet high-quality dataset, enabling a cold-start initialization of the model and facilitating its adaptation to effective reasoning patterns. Subsequently, we employ an entropy-preserving supervised fine-tuning (Entropy-SFT) strategy coupled with token-wise entropy-adaptive (TEA-RL) reinforcement learning guided by rule-based dense rewards. This approach encourages the model to transform its reasoning mechanism, ultimately fostering generalizable reasoning abilities that encompass preview and self-checking. Extensive experiments conducted on instruction-following benchmarks demonstrate remarkable performance improvements across various model scales. Notably, our Light-IF-32B model surpasses both larger open-source models such as DeepSeek-R1 and closed-source models like Doubao-1.6.
Authors: Sangho Suh, Michael Lai, Kevin Pu, Steven P. Dow, Tovi Grossman
Abstract: Design processes involve exploration, iteration, and movement across interconnected stages such as persona creation, problem framing, solution ideation, and prototyping. However, time and resource constraints often hinder designers from exploring broadly, collecting feedback, and revisiting earlier assumptions-making it difficult to uphold core design principles in practice. To better understand these challenges, we conducted a formative study with 15 participants-comprised of UX practitioners, students, and instructors. Based on the findings, we developed StoryEnsemble, a tool that integrates AI into a node-link interface and leverages forward and backward propagation to support dynamic exploration and iteration across the design process. A user study with 10 participants showed that StoryEnsemble enables rapid, multi-directional iteration and flexible navigation across design stages. This work advances our understanding of how AI can foster more iterative design practices by introducing novel interactions that make exploration and iteration more fluid, accessible, and engaging.
Authors: Junle Liu, Chang Liu, Yanyu Ke, Wenliang Chen, Kihing Shum, K. T. Tse, Gang Hu
Abstract: The wall pressure is of great importance in understanding the forces and structural responses induced by fluid. Recent works have investigated the potential of deep learning techniques in predicting mean pressure coefficients and fluctuating pressure coefficients, but most of existing deep learning frameworks are limited to predicting a single snapshot using full spatial information. To forecast spatiotemporal wall pressure of flow past a rectangular cylinder, this study develops a physics-aware DeepU-Fourier neural Network (DeepUFNet) deep learning model. DeepUFNet comprises the UNet structure and the Fourier neural network, with physical high-frequency loss control embedded in the model training stage to optimize model performance, where the parameter $\beta$ varies with the development of the training epoch. Wind tunnel testing is performed to collect wall pressures of a two-dimensional rectangular cylinder with a side ratio of 1.5 at an angle of attack of zero using high-frequency pressure scanning, thereby constructing a database for DeepUFNet training and testing. The DeepUFNet model is found to forecast spatiotemporal wall pressure information with high accuracy. The comparison between forecast results and experimental data presents agreement in statistical information, temporal pressure variation, power spectrum density, spatial distribution, and spatiotemporal correlation. It is also found that embedding a physical high-frequency loss control coefficient $\beta$ in the DeepUFNet model can significantly improve model performance in forecasting spatiotemporal wall pressure information, in particular, in forecasting high-order frequency fluctuation and wall pressure variance. Furthermore, the DeepUFNet extrapolation capability is tested with sparse spatial information input, and the model presents a satisfactory extrapolation ability
Authors: Xinwei Liu, Xiaojun Jia, Yuan Xun, Simeng Qin, Xiaochun Cao
Abstract: Vision-Language Models (VLMs) such as GPT-4o now demonstrate a remarkable ability to infer users' locations from public shared images, posing a substantial risk to geoprivacy. Although adversarial perturbations offer a potential defense, current methods are ill-suited for this scenario: they often perform poorly on high-resolution images and low perturbation budgets, and may introduce irrelevant semantic content. To address these limitations, we propose GeoShield, a novel adversarial framework designed for robust geoprivacy protection in real-world scenarios. GeoShield comprises three key modules: a feature disentanglement module that separates geographical and non-geographical information, an exposure element identification module that pinpoints geo-revealing regions within an image, and a scale-adaptive enhancement module that jointly optimizes perturbations at both global and local levels to ensure effectiveness across resolutions. Extensive experiments on challenging benchmarks show that GeoShield consistently surpasses prior methods in black-box settings, achieving strong privacy protection with minimal impact on visual or semantic quality. To our knowledge, this work is the first to explore adversarial perturbations for defending against geolocation inference by advanced VLMs, providing a practical and effective solution to escalating privacy concerns.
Authors: Wang Yu-Hang, Shiwei Li, Jianxiang Liao, Li Bohan, Jian Liu, Wenfei Yin
Abstract: Adversarial perturbations pose a significant threat to deep learning models. Adversarial Training (AT), the predominant defense method, faces challenges of high computational costs and a degradation in standard performance. While data augmentation offers an alternative path, existing techniques either yield limited robustness gains or incur substantial training overhead. Therefore, developing a defense mechanism that is both highly efficient and strongly robust is of paramount importance.In this work, we first conduct a systematic analysis of existing augmentation techniques, revealing that the synergy among diverse strategies -- rather than any single method -- is crucial for enhancing robustness. Based on this insight, we propose the Universal Adversarial Augmenter (UAA) framework, which is characterized by its plug-and-play nature and training efficiency. UAA decouples the expensive perturbation generation process from model training by pre-computing a universal transformation offline, which is then used to efficiently generate unique adversarial perturbations for each sample during training.Extensive experiments conducted on multiple benchmarks validate the effectiveness of UAA. The results demonstrate that UAA establishes a new state-of-the-art (SOTA) for data-augmentation-based adversarial defense strategies , without requiring the online generation of adversarial examples during training. This framework provides a practical and efficient pathway for building robust models,Our code is available in the supplementary materials.
Authors: Hikari Yanagawa, Yuichi Hiroi, Satomi Tokida, Yuji Hatada, Takefumi Hiraki
Abstract: While commercial metaverse platforms offer diverse user-generated content, they lack effective navigation assistance that can dynamically adapt to users' interests and intentions. Although previous research has investigated on-demand agents in controlled environments, implementation in commercial settings with diverse world configurations and platform constraints remains challenging. We present Navigation Pixie, an on-demand navigation agent employing a loosely coupled architecture that integrates structured spatial metadata with LLM-based natural language processing while minimizing platform dependencies, which enables experiments on the extensive user base of commercial metaverse platforms. Our cross-platform experiments on commercial metaverse platform Cluster with 99 PC client and 94 VR-HMD participants demonstrated that Navigation Pixie significantly increased dwell time and free exploration compared to fixed-route and no-agent conditions across both platforms. Subjective evaluations revealed consistent on-demand preferences in PC environments versus context-dependent social perception advantages in VR-HMD. This research contributes to advancing VR interaction design through conversational spatial navigation agents, establishes cross-platform evaluation methodologies revealing environment-dependent effectiveness, and demonstrates empirical experimentation frameworks for commercial metaverse platforms.
Authors: Mutaz Ayesh, Nicol\'as Guti\'errez-Rol\'on, Fernando Alva-Manchego
Abstract: This paper details the CardiffNLP team's contribution to the CLEARS shared task on Spanish text adaptation, hosted by IberLEF 2025. The shared task contained two subtasks and the team submitted to both. Our team took an LLM-prompting approach with different prompt variations. While we initially experimented with LLaMA-3.2, we adopted Gemma-3 for our final submission, and landed third place in Subtask 1 and second place in Subtask 2. We detail our numerous prompt variations, examples, and experimental results.
Authors: Deborah Dore, Elena Cabrio, Serena Villata
Abstract: The increasing amount of political debates and politics-related discussions calls for the definition of novel computational methods to automatically analyse such content with the final goal of lightening up political deliberation to citizens. However, the specificity of the political language and the argumentative form of these debates (employing hidden communication strategies and leveraging implicit arguments) make this task very challenging, even for current general-purpose pre-trained Language Models. To address this issue, we introduce a novel pre-trained Language Model for political discourse language called RooseBERT. Pre-training a language model on a specialised domain presents different technical and linguistic challenges, requiring extensive computational resources and large-scale data. RooseBERT has been trained on large political debate and speech corpora (8K debates, each composed of several sub-debates on different topics) in English. To evaluate its performances, we fine-tuned it on four downstream tasks related to political debate analysis, i.e., named entity recognition, sentiment analysis, argument component detection and classification, and argument relation prediction and classification. Our results demonstrate significant improvements over general-purpose Language Models on these four tasks, highlighting how domain-specific pre-training enhances performance in political debate analysis. We release the RooseBERT language model for the research community.
Authors: Davin Choo, Winston Fu, Derek Khu, Tzeh Yuan Neoh, Tze-Yang Poon, Nicholas Teh
Abstract: We study the online fair division problem, where indivisible goods arrive sequentially and must be allocated immediately and irrevocably to agents. Prior work has established strong impossibility results for approximating classic fairness notions, such as envy-freeness and maximin share fairness, in this setting. In contrast, we focus on proportionality up to one good (PROP1), a natural relaxation of proportionality whose approximability remains unresolved. We begin by showing that three natural greedy algorithms fail to guarantee any positive approximation to PROP1 in general, against an adaptive adversary. This is surprising because greedy algorithms are commonly used in fair division and a natural greedy algorithm is known to be able to achieve PROP1 under additional information assumptions. This hardness result motivates the study of non-adaptive adversaries and the use of side-information, in the spirit of learning-augmented algorithms. For non-adaptive adversaries, we show that the simple uniformly random allocation can achieve a meaningful PROP1 approximation with high probability. Meanwhile, we present an algorithm that obtain robust approximation ratios against PROP1 when given predictions of the maximum item value (MIV). Interestingly, we also show that stronger fairness notions such as EF1, MMS, and PROPX remain inapproximable even with perfect MIV predictions.
Authors: Jisoo Kim, Wooseok Seo, Junwan Kim, Seungho Park, Sooyeon Park, Youngjae Yu
Abstract: With growing interest in deploying text-to-video (T2V) models in resource-constrained environments, reducing their high computational cost has become crucial, leading to extensive research on pruning and knowledge distillation methods while maintaining performance. However, existing distillation methods primarily rely on supervised fine-tuning (SFT), which often leads to mode collapse as pruned models with reduced capacity fail to directly match the teacher's outputs, ultimately resulting in degraded quality. To address this challenge, we propose an effective distillation method, ReDPO, that integrates DPO and SFT. Our approach leverages DPO to guide the student model to focus on recovering only the targeted properties, rather than passively imitating the teacher, while also utilizing SFT to enhance overall performance. We additionally propose V.I.P., a novel framework for filtering and curating high-quality pair datasets, along with a step-by-step online approach for calibrated training. We validate our method on two leading T2V models, VideoCrafter2 and AnimateDiff, achieving parameter reduction of 36.2% and 67.5% each, while maintaining or even surpassing the performance of full models. Further experiments demonstrate the effectiveness of both ReDPO and V.I.P. framework in enabling efficient and high-quality video generation. Our code and videos are available at https://jiiiisoo.github.io/VIP.github.io/.
Authors: Junhyuk Choi, Hyeonchu Park, Haemin Lee, Hyebeen Shin, Hyun Joung Jin, Bugeun Kim
Abstract: Recent advances in Large Language Models (LLMs) have generated significant interest in their capacity to simulate human-like behaviors, yet most studies rely on fictional personas rather than actual human data. We address this limitation by evaluating LLMs' ability to predict individual economic decision-making using Pay-What-You-Want (PWYW) pricing experiments with real 522 human personas. Our study systematically compares three state-of-the-art multimodal LLMs using detailed persona information from 522 Korean participants in cultural consumption scenarios. We investigate whether LLMs can accurately replicate individual human choices and how persona injection methods affect prediction performance. Results reveal that while LLMs struggle with precise individual-level predictions, they demonstrate reasonable group-level behavioral tendencies. Also, we found that commonly adopted prompting techniques are not much better than naive prompting methods; reconstruction of personal narrative nor retrieval augmented generation have no significant gain against simple prompting method. We believe that these findings can provide the first comprehensive evaluation of LLMs' capabilities on simulating economic behavior using real human data, offering empirical guidance for persona-based simulation in computational social science.
Authors: Albertus Denny Handoko, Riko I Made
Abstract: High throughput experimentation tools, machine learning (ML) methods, and open material databases are radically changing the way new materials are discovered. From the experimentally driven approach in the past, we are moving quickly towards the artificial intelligence (AI) driven approach, realizing the 'inverse design' capabilities that allow the discovery of new materials given the desired properties. This review aims to discuss different principles of AI-driven generative models that are applicable for materials discovery, including different materials representations available for this purpose. We will also highlight specific applications of generative models in designing new catalysts, semiconductors, polymers, or crystals while addressing challenges such as data scarcity, computational cost, interpretability, synthesizability, and dataset biases. Emerging approaches to overcome limitations and integrate AI with experimental workflows will be discussed, including multimodal models, physics informed architectures, and closed-loop discovery systems. This review aims to provide insights for researchers aiming to harness AI's transformative potential in accelerating materials discovery for sustainability, healthcare, and energy innovation.
Authors: Shahed Masoudian, Gustavo Escobedo, Hannah Strauss, Markus Schedl
Abstract: As Large Language Models (LLMs) are increasingly used across different applications, concerns about their potential to amplify gender biases in various tasks are rising. Prior research has often probed gender bias using explicit gender cues as counterfactual, or studied them in sentence completion and short question answering tasks. These formats might overlook more implicit forms of bias embedded in generative behavior of longer content. In this work, we investigate gender bias in LLMs using gender stereotypes studied in psychology (e.g., aggressiveness or gossiping) in an open-ended task of narrative generation. We introduce a novel dataset called StereoBias-Stories containing short stories either unconditioned or conditioned on (one, two, or six) random attributes from 25 psychological stereotypes and three task-related story endings. We analyze how the gender contribution in the overall story changes in response to these attributes and present three key findings: (1) While models, on average, are highly biased towards male in unconditioned prompts, conditioning on attributes independent from gender stereotypes mitigates this bias. (2) Combining multiple attributes associated with the same gender stereotype intensifies model behavior, with male ones amplifying bias and female ones alleviating it. (3) Model biases align with psychological ground-truth used for categorization, and alignment strength increases with model size. Together, these insights highlight the importance of psychology-grounded evaluation of LLMs.
Authors: Leonidas Zotos, Ivo Pascal de Jong, Matias Valdenegro-Toro, Andreea Ioana Sburlea, Malvina Nissim, Hedderik van Rijn
Abstract: Estimating the difficulty of exam questions is essential for developing good exams, but professors are not always good at this task. We compare various Large Language Model-based methods with three professors in their ability to estimate what percentage of students will give correct answers on True/False exam questions in the areas of Neural Networks and Machine Learning. Our results show that the professors have limited ability to distinguish between easy and difficult questions and that they are outperformed by directly asking Gemini 2.5 to solve this task. Yet, we obtained even better results using uncertainties of the LLMs solving the questions in a supervised learning setting, using only 42 training samples. We conclude that supervised learning using LLM uncertainty can help professors better estimate the difficulty of exam questions, improving the quality of assessment.
Authors: Kisu Yang, Yoonna Jang, Hwanseok Jang, Kenneth Choi, Isabelle Augenstein, Heuiseok Lim
Abstract: Lowering the numerical precision of model parameters and computations is widely adopted to improve the efficiency of retrieval systems. However, when computing relevance scores between the query and documents in low-precision, we observe spurious ties due to the reduced granularity. This introduces high variability in the results based on tie resolution, making the evaluation less reliable. To address this, we propose a more robust retrieval evaluation protocol designed to reduce score variation. It consists of: (1) High-Precision Scoring (HPS), which upcasts the final scoring step to higher precision to resolve tied candidates with minimal computational cost; and (2) Tie-aware Retrieval Metrics (TRM), which report expected scores, range, and bias to quantify order uncertainty of tied candidates. Our experiments test multiple models with three scoring functions on two retrieval datasets to demonstrate that HPS dramatically reduces tie-induced instability, and TRM accurately recovers expected metric values. This combination enables a more consistent and reliable evaluation system for lower-precision retrievals.
Authors: Libo Zhang, Xinyu Yi, Feng Xu
Abstract: In recent years, tracking human motion using IMUs from everyday devices such as smartphones and smartwatches has gained increasing popularity. However, due to the sparsity of sensor measurements and the lack of datasets capturing human motion over uneven terrain, existing methods often struggle with pose estimation accuracy and are typically limited to recovering movements on flat terrain only. To this end, we present BaroPoser, the first method that combines IMU and barometric data recorded by a smartphone and a smartwatch to estimate human pose and global translation in real time. By leveraging barometric readings, we estimate sensor height changes, which provide valuable cues for both improving the accuracy of human pose estimation and predicting global translation on non-flat terrain. Furthermore, we propose a local thigh coordinate frame to disentangle local and global motion input for better pose representation learning. We evaluate our method on both public benchmark datasets and real-world recordings. Quantitative and qualitative results demonstrate that our approach outperforms the state-of-the-art (SOTA) methods that use IMUs only with the same hardware configuration.
Authors: Mari Ashiga, Vardan Voskanyan, Fateme Dinmohammadi, Jingzhi Gong, Paul Brookes, Matthew Truscott, Rafail Giavrimis, Mike Basios, Leslie Kanthan, Wei Jie
Abstract: Recent advancements in Large Language Models (LLMs) for code optimization have enabled industrial platforms to automate software performance engineering at unprecedented scale and speed. Yet, organizations in regulated industries face strict constraints on which LLMs they can use - many cannot utilize commercial models due to data privacy regulations and compliance requirements, creating a significant challenge for achieving high-quality code optimization while maintaining cost-effectiveness. We address this by implementing a Mixture-of-Agents (MoA) approach that directly synthesizes code from multiple specialized LLMs, comparing it against TurinTech AI's vanilla Genetic Algorithm (GA)-based ensemble system and individual LLM optimizers using real-world industrial codebases. Our key contributions include: (1) First MoA application to industrial code optimization using real-world codebases; (2) Empirical evidence that MoA excels with open-source models, achieving 14.3% to 22.2% cost savings and 28.6% to 32.2% faster optimization times for regulated environments; (3) Deployment guidelines demonstrating GA's advantage with commercial models while both ensembles outperform individual LLMs; and (4) Real-world validation across 50 code snippets and seven LLM combinations, generating over 8,700 variants, addresses gaps in industrial LLM ensemble evaluation. This provides actionable guidance for organizations balancing regulatory compliance with optimization performance in production environments.
Authors: He Xiao, Qingyao Yang, Dirui Xie, Wendong Xu, Wenyong Zhou, Haobo Liu, Zhengwu Liu, Ngai Wong
Abstract: Large language models with billions of parameters are often over-provisioned: many layers contribute little unique information yet dominate the memory and energy footprint during inference. We present LieQ, a metric-driven post-training quantization framework that addresses the critical challenge of maintaining accuracy in sub-7B models under extreme low-bit compression. Our method introduces three complementary layer-wise diagnostics-Perplexity Drop, Representational Compactness, and Top-k Energy Gain -that reveal a canonical division of labour across layers, enabling automatic bit-width allocation without gradient updates. Unlike existing approaches that suffer severe accuracy degradation at 2-3 bits precision, LieQ achieves state-of-the-art compression-accuracy trade-offs: on Qwen3-4B, it recovers 95.9% of FP16 baseline performance at 2.05-bit quantization, outperforming GPTQ by 19.7% and AWQ by 18.1% on average across seven zero-shot reasoning tasks. Applied to LLaMA3.2-3B, LieQ maintains 98.2% of baseline accuracy at 2.07-bit precision while enabling 4x memory reduction, establishing new paradigms for deploying small language models on resource-constrained edge devices.
Authors: Zhende Song, Shengji Tang, Peng Ye, Jiayuan Fan, Tao Chen
Abstract: Test-time scaling (TTS) has emerged as a promising research field for enhancing the effectiveness of large language models (LLMs) without extra training. However, most existing approaches, e.g., Best-of-N and Self-Consistency rely on a single agent interacting with a reward model (SA-SR), constrained by limited capabilities of a single test-time scaling (STTS) paradigm. On the other hand, recent works demonstrate that collective-agent methods can break through the upper bound of single-agent systems by orchestrating diverse models. Thus, in this paper, we take a first step towards exploring Collective Test-Time Scaling (CTTS). Consider the different interaction types of single and multiple models, we design three primary paradigms to investigate the optimal paradigm of CTTS: (1) single agent to multiple reward models (SA-MR); (2) multiple agents to single reward model (MA-SR); and (3) multiple agents to multiple reward models (MA-MR). Extensive experiments demonstrate that MA-MR consistently achieves the best performance. Based on this, we propose a novel framework named CTTS-MM that effectively leverages both multi-agent and multi-reward-model collaboration for enhanced inference. Specifically, for multi-agent collaboration, we propose an Agent Collaboration Search (ACS), which searches for the most effective combination of LLM agents from a large candidate pool; for multi-reward-model collaboration, we propose Mixture of Reword Models (MoR), which consists of a curated question pool and a Prior Reward model Ensemble Selection (PRES) to select the optimal combinations of reward models via Pair-wise Reward Ranking (PRR) metric. Experiments across seven mainstream benchmarks demonstrate that the proposed CTTS-MM consistently obtains superior performance. Code will be released at https://github.com/magent4aci/CTTS-MM.
Authors: Mehdi Akbari Gurabi, Lasse Nitz, Radu-Mihai Castravet, Roman Matzutt, Avikarsha Mandal, Stefan Decker
Abstract: Existing cybersecurity playbooks are often written in heterogeneous, non-machine-readable formats, which limits their automation and interoperability across Security Orchestration, Automation, and Response platforms. This paper explores the suitability of Large Language Models, combined with Prompt Engineering, to automatically translate legacy incident response playbooks into the standardized, machine-readable CACAO format. We systematically examine various Prompt Engineering techniques and carefully design prompts aimed at maximizing syntactic accuracy and semantic fidelity for control flow preservation. Our modular transformation pipeline integrates a syntax checker to ensure syntactic correctness and features an iterative refinement mechanism that progressively reduces syntactic errors. We evaluate the proposed approach on a custom-generated dataset comprising diverse legacy playbooks paired with manually created CACAO references. The results demonstrate that our method significantly improves the accuracy of playbook transformation over baseline models, effectively captures complex workflow structures, and substantially reduces errors. It highlights the potential for practical deployment in automated cybersecurity playbook transformation tasks.
Authors: Yufei Xue, Yushi Huang, Jiawei Shao, Jun Zhang
Abstract: Post-training quantization (PTQ) has emerged as an effective approach for compressing large models and accelerating their inference without retraining. While PTQ has been extensively studied in the context of large language models (LLMs), its applicability to vision-language models (VLMs) remains underexplored. In this paper, we identify a modality discrepancy (\emph{i.e.}, limited text tokens \emph{vs.} excessive and redundant vision tokens) of VLMs. However, existing Hessian-based LLM PTQ methods treat all tokens equally during quantization, resulting in severe performance drops when applied to VLMs. Motivated by this observation, we propose a novel importance-aware PTQ framework tailored for VLMs, dubbed VLMQ. Specifically, to address vision token redundancy, VLMQ 1) optimizes an importance-aware objective that yields an enhanced Hessian with token-level importance factors, while retaining compatibility with parallelized weight updates, and 2) ensures efficiency and effectiveness by computing these factors via a single lightweight block-wise backward pass, guided by a theoretical connection to token-level perturbations. Extensive evaluations on 8 benchmarks across 0.5B$\sim$32B VLMs demonstrate the state-of-the-art (SOTA) performance of our VLMQ, particularly under low-bit settings. For example, it achieves a substantial \textbf{16.45\%} improvement on MME-RealWorld under 2-bit quantization.
Authors: Bodam Kim, Hiskias Dingeto, Taeyoun Kwon, Dasol Choi, DongGeon Lee, Haon Park, JaeHoon Lee, Jongho Shin
Abstract: As large language models become increasingly integrated into daily life, audio has emerged as a key interface for human-AI interaction. However, this convenience also introduces new vulnerabilities, making audio a potential attack surface for adversaries. Our research introduces WhisperInject, a two-stage adversarial audio attack framework that can manipulate state-of-the-art audio language models to generate harmful content. Our method uses imperceptible perturbations in audio inputs that remain benign to human listeners. The first stage uses a novel reward-based optimization method, Reinforcement Learning with Projected Gradient Descent (RL-PGD), to guide the target model to circumvent its own safety protocols and generate harmful native responses. This native harmful response then serves as the target for Stage 2, Payload Injection, where we use Projected Gradient Descent (PGD) to optimize subtle perturbations that are embedded into benign audio carriers, such as weather queries or greeting messages. Validated under the rigorous StrongREJECT, LlamaGuard, as well as Human Evaluation safety evaluation framework, our experiments demonstrate a success rate exceeding 86% across Qwen2.5-Omni-3B, Qwen2.5-Omni-7B, and Phi-4-Multimodal. Our work demonstrates a new class of practical, audio-native threats, moving beyond theoretical exploits to reveal a feasible and covert method for manipulating AI behavior.
Authors: Muhammad Zohaib, Muhammad Azeem Akbar, Sami Hyrynsalmi, Arif Ali Khan
Abstract: The emergence of agentic AI systems in 6G software businesses presents both strategic opportunities and significant challenges. While such systems promise increased autonomy, scalability, and intelligent decision-making across distributed environments, their adoption raises concerns regarding technical immaturity, integration complexity, organizational readiness, and performance-cost trade-offs. In this study, we conducted a preliminary thematic mapping to identify factors influencing the adoption of agentic software within the context of 6G. Drawing on a multivocal literature review and targeted scanning, we identified 29 motivators and 27 demotivators, which were further categorized into five high-level themes in each group. This thematic mapping offers a structured overview of the enabling and inhibiting forces shaping organizational readiness for agentic transformation. Positioned as a feasibility assessment, the study represents an early phase of a broader research initiative aimed at developing and validating a layered maturity model grounded in CMMI model with the software architectural three dimensions possibly Data, Business Logic, and Presentation. Ultimately, this work seeks to provide a practical framework to help software-driven organizations assess, structure, and advance their agent-first capabilities in alignment with the demands of 6G.
Authors: Pingchuan Ma, Xiaopei Yang, Yusong Li, Ming Gui, Felix Krause, Johannes Schusterbauer, Bj\"orn Ommer
Abstract: Explicitly disentangling style and content in vision models remains challenging due to their semantic overlap and the subjectivity of human perception. Existing methods propose separation through generative or discriminative objectives, but they still face the inherent ambiguity of disentangling intertwined concepts. Instead, we ask: Can we bypass explicit disentanglement by learning to merge style and content invertibly, allowing separation to emerge naturally? We propose SCFlow, a flow-matching framework that learns bidirectional mappings between entangled and disentangled representations. Our approach is built upon three key insights: 1) Training solely to merge style and content, a well-defined task, enables invertible disentanglement without explicit supervision; 2) flow matching bridges on arbitrary distributions, avoiding the restrictive Gaussian priors of diffusion models and normalizing flows; and 3) a synthetic dataset of 510,000 samples (51 styles $\times$ 10,000 content samples) was curated to simulate disentanglement through systematic style-content pairing. Beyond controllable generation tasks, we demonstrate that SCFlow generalizes to ImageNet-1k and WikiArt in zero-shot settings and achieves competitive performance, highlighting that disentanglement naturally emerges from the invertible merging process.
Authors: Xinlei Yu, Zhangquan Chen, Yudong Zhang, Shilin Lu, Ruolin Shen, Jiangning Zhang, Xiaobin Hu, Yanwei Fu, Shuicheng Yan
Abstract: Existing vision-language models (VLMs), whether generalists or specialists, remain constrained by their parameter scale, lack robust self-correction capabilities, and underperform in tasks involving long visual contexts and complex reasoning, resulting in suboptimal performance on document-based tasks. To address this, we propose MACT, a Multi-Agent Collaboration framework with Test-Time scaling, tailored for visual document understanding and visual question answering (VQA). It comprises four distinct small-scale agents, i.e., planning, execution, judgment, and answer agents, with clearly defined roles and effective collaboration. Notably, the judgment agent exclusively verifies correctness and redirects to prior agents for revisions, outperforming conventional correction strategies. To further expand the capability boundaries of the framework, we propose mixed reward modeling that balances agent-specific abilities and global collaboration, as well as agent-wise hybrid test-time scaling, which customizes different scaling strategies for each agent based on their functions. Evaluated on benchmarks spanning both document-based and non-document-based settings, our MACT shows superior performance with a smaller parameter scale without sacrificing the ability of general and mathematical tasks. Especially, it stands out in benchmarks involving long visual contexts and complicated reasoning. The three variants of MACT consistently hold the top three positions in average scores, leading in 13 of the 15 benchmarks. Code will be available at: https://github.com/YU-deep/MACT.git.
Authors: Diana-Nicoleta Grigore, Neelu Madan, Andreas Mogelmose, Thomas B. Moeslund, Radu Tudor Ionescu
Abstract: Unsupervised video segmentation is a challenging computer vision task, especially due to the lack of supervisory signals coupled with the complexity of visual scenes. To overcome this challenge, state-of-the-art models based on slot attention often have to rely on large and computationally expensive neural architectures. To this end, we propose a simple knowledge distillation framework that effectively transfers object-centric representations to a lightweight student. The proposed framework, called SlotMatch, aligns corresponding teacher and student slots via the cosine similarity, requiring no additional distillation objectives or auxiliary supervision. The simplicity of SlotMatch is confirmed via theoretical and empirical evidence, both indicating that integrating additional losses is redundant. We conduct experiments on two datasets to compare the state-of-the-art teacher model, SlotContrast, with our distilled student. The results show that our student based on SlotMatch matches and even outperforms its teacher, while using 3.6x less parameters and running 1.9x faster. Moreover, our student surpasses previous unsupervised video segmentation models.
Authors: Shivangi Nigam, Adarsh Prasad Behera, Shekhar Verma, P. Nagabhushan
Abstract: This paper presents Fd-CycleGAN, an image-to-image (I2I) translation framework that enhances latent representation learning to approximate real data distributions. Building upon the foundation of CycleGAN, our approach integrates Local Neighborhood Encoding (LNE) and frequency-aware supervision to capture fine-grained local pixel semantics while preserving structural coherence from the source domain. We employ distribution-based loss metrics, including KL/JS divergence and log-based similarity measures, to explicitly quantify the alignment between real and generated image distributions in both spatial and frequency domains. To validate the efficacy of Fd-CycleGAN, we conduct experiments on diverse datasets -- Horse2Zebra, Monet2Photo, and a synthetically augmented Strike-off dataset. Compared to baseline CycleGAN and other state-of-the-art methods, our approach demonstrates superior perceptual quality, faster convergence, and improved mode diversity, particularly in low-data regimes. By effectively capturing local and global distribution characteristics, Fd-CycleGAN achieves more visually coherent and semantically consistent translations. Our results suggest that frequency-guided latent learning significantly improves generalization in image translation tasks, with promising applications in document restoration, artistic style transfer, and medical image synthesis. We also provide comparative insights with diffusion-based generative models, highlighting the advantages of our lightweight adversarial approach in terms of training efficiency and qualitative output.
Authors: Futian Wang, Yuhan Qiao, Xiao Wang, Fuling Wang, Yuxiang Zhang, Dengdi Sun
Abstract: X-ray medical report generation is one of the important applications of artificial intelligence in healthcare. With the support of large foundation models, the quality of medical report generation has significantly improved. However, challenges such as hallucination and weak disease diagnostic capability still persist. In this paper, we first construct a large-scale multi-modal medical knowledge graph (termed M3KG) based on the ground truth medical report using the GPT-4o. It contains 2477 entities, 3 kinds of relations, 37424 triples, and 6943 disease-aware vision tokens for the CheXpert Plus dataset. Then, we sample it to obtain multi-granularity semantic graphs and use an R-GCN encoder for feature extraction. For the input X-ray image, we adopt the Swin-Transformer to extract the vision features and interact with the knowledge using cross-attention. The vision tokens are fed into a Q-former and retrieved the disease-aware vision tokens using another cross-attention. Finally, we adopt the large language model to map the semantic knowledge graph, input X-ray image, and disease-aware vision tokens into language descriptions. Extensive experiments on multiple datasets fully validated the effectiveness of our proposed knowledge graph and X-ray report generation framework. The source code of this paper will be released on https://github.com/Event-AHU/Medical_Image_Analysis.
Authors: Iyad Rahwan, Azim Shariff, Jean-Fran\c{c}ois Bonnefon
Abstract: Predicting the social and behavioral impact of future technologies, before they are achieved, would allow us to guide their development and regulation before these impacts get entrenched. Traditionally, this prediction has relied on qualitative, narrative methods. Here we describe a method which uses experimental methods to simulate future technologies, and collect quantitative measures of the attitudes and behaviors of participants assigned to controlled variations of the future. We call this method 'science fiction science'. We suggest that the reason why this method has not been fully embraced yet, despite its potential benefits, is that experimental scientists may be reluctant to engage in work facing such serious validity threats as science fiction science. To address these threats, we consider possible constraints on the kind of technology that science fiction science may study, as well as the unconventional, immersive methods that science fiction science may require. We seek to provide perspective on the reasons why this method has been marginalized for so long, what benefits it would bring if it could be built on strong yet unusual methods, and how we can normalize these methods to help the diverse community of science fiction scientists to engage in a virtuous cycle of validity improvement.
Authors: Hongjun Liu, Chao Yao, Yalan Zhang, Xiaokun wang, Xiaojuan Ban
Abstract: Electroencephalogram (EEG) signal classification faces significant challenges due to data distribution shifts caused by heterogeneous electrode configurations, acquisition protocols, and hardware discrepancies across domains. This paper introduces IMAC, a novel channel-dependent mask and imputation self-supervised framework that formulates the alignment of cross-domain EEG data shifts as a spatial time series imputation task. To address heterogeneous electrode configurations in cross-domain scenarios, IMAC first standardizes different electrode layouts using a 3D-to-2D positional unification mapping strategy, establishing unified spatial representations. Unlike previous mask-based self-supervised representation learning methods, IMAC introduces spatio-temporal signal alignment. This involves constructing a channel-dependent mask and reconstruction task framed as a low-to-high resolution EEG spatial imputation problem. Consequently, this approach simulates cross-domain variations such as channel omissions and temporal instabilities, thus enabling the model to leverage the proposed imputer for robust signal alignment during inference. Furthermore, IMAC incorporates a disentangled structure that separately models the temporal and spatial information of the EEG signals separately, reducing computational complexity while enhancing flexibility and adaptability. Comprehensive evaluations across 10 publicly available EEG datasets demonstrate IMAC's superior performance, achieving state-of-the-art classification accuracy in both cross-subject and cross-center validation scenarios. Notably, IMAC shows strong robustness under both simulated and real-world distribution shifts, surpassing baseline methods by up to $35$\% in integrity scores while maintaining consistent classification accuracy.
Authors: Ch\"unhung Wu, Jinliang Lu, Zixuan Ren, Gangqiang Hu, Zhi Wu, Dai Dai, Hua Wu
Abstract: Human cognition naturally engages with abstract and fluid concepts, whereas existing reasoning models often rely on generating discrete tokens, potentially constraining their expressive capabilities. Recent advancements aim to address this limitation by enabling large language models (LLMs) to generate soft, abstract tokens, thus facilitating reasoning within a continuous concept space. This paper explores the `Soft Thinking' capabilities of various LLMs by examining the models' internal behavior using a suite of probing techniques. Contrary to the common belief that Soft Thinking enables the simultaneous exploration of diverse reasoning paths, our findings reveal that LLMs predominantly rely on the most influential component of the soft inputs during subsequent decoding steps. This reliance hinders the exploration of different reasoning paths and reduces vanilla Soft Thinking to a form of greedy decoding, obscuring the advantage of transmitting more information through Soft Tokens. To tackle this issue, we explore sampling strategies to introduce \emph{randomness}, employing methods such as Dirichlet resampling and the Gumbel-Softmax trick. Our experiments demonstrate that incorporating randomness can alleviate the limitations of vanilla approaches and unleash the potential of Soft Thinking. Notably, the Gumbel-Softmax trick provides adequate randomness with controlled smoothness, resulting in superior performance across eight reasoning benchmarks.
Authors: Jan Melechovsky, Ambuj Mehrish, Dorien Herremans
Abstract: Music recordings often suffer from audio quality issues such as excessive reverberation, distortion, clipping, tonal imbalances, and a narrowed stereo image, especially when created in non-professional settings without specialized equipment or expertise. These problems are typically corrected using separate specialized tools and manual adjustments. In this paper, we introduce SonicMaster, the first unified generative model for music restoration and mastering that addresses a broad spectrum of audio artifacts with text-based control. SonicMaster is conditioned on natural language instructions to apply targeted enhancements, or can operate in an automatic mode for general restoration. To train this model, we construct the SonicMaster dataset, a large dataset of paired degraded and high-quality tracks by simulating common degradation types with nineteen degradation functions belonging to five enhancements groups: equalization, dynamics, reverb, amplitude, and stereo. Our approach leverages a flow-matching generative training paradigm to learn an audio transformation that maps degraded inputs to their cleaned, mastered versions guided by text prompts. Objective audio quality metrics demonstrate that SonicMaster significantly improves sound quality across all artifact categories. Furthermore, subjective listening tests confirm that listeners prefer SonicMaster's enhanced outputs over the original degraded audio, highlighting the effectiveness of our unified approach.
Authors: Pranshu Rastogi
Abstract: SemEval-2025 Task 7: Multilingual and Crosslingual Fact-Checked Claim Retrieval is approached as a Learning-to-Rank task using a bi-encoder model fine-tuned from a pre-trained transformer optimized for sentence similarity. Training used both the source languages and their English translations for multilingual retrieval and only English translations for cross-lingual retrieval. Using lightweight models with fewer than 500M parameters and training on Kaggle T4 GPUs, the method achieved 92% Success@10 in multilingual and 80% Success@10 in 5th in crosslingual and 10th in multilingual tracks.
Authors: Junjie Cao, Kaizhou Li, Xinchun Yu, Hongxiang Li, Xiaoping Zhang
Abstract: With the rapid development of generative technology, current generative models can generate high-fidelity digital content and edit it in a controlled manner. However, there is a risk that malicious individuals might misuse these capabilities for misleading activities. Although existing research has attempted to shield photographic images from being manipulated by generative models, there remains a significant disparity in the protection offered to video content editing. To bridge the gap, we propose a protection method named VideoGuard, which can effectively protect videos from unauthorized malicious editing. This protection is achieved through the subtle introduction of nearly unnoticeable perturbations that interfere with the functioning of the intended generative diffusion models. Due to the redundancy between video frames, and inter-frame attention mechanism in video diffusion models, simply applying image-based protection methods separately to every video frame can not shield video from unauthorized editing. To tackle the above challenge, we adopt joint frame optimization, treating all video frames as an optimization entity. Furthermore, we extract video motion information and fuse it into optimization objectives. Thus, these alterations can effectively force the models to produce outputs that are implausible and inconsistent. We provide a pipeline to optimize this perturbation. Finally, we use both objective metrics and subjective metrics to demonstrate the efficacy of our method, and the results show that the protection performance of VideoGuard is superior to all the baseline methods.
Authors: Hyungjin Kim, Seokho Ahn, Young-Duk Seo
Abstract: Personalized generation in T2I diffusion models aims to naturally incorporate individual user preferences into the generation process with minimal user intervention. However, existing studies primarily rely on prompt-level modeling with large-scale models, often leading to inaccurate personalization due to the limited input token capacity of T2I diffusion models. To address these limitations, we propose DrUM, a novel method that integrates user profiling with a transformer-based adapter to enable personalized generation through condition-level modeling in the latent space. DrUM demonstrates strong performance on large-scale datasets and seamlessly integrates with open-source text encoders, making it compatible with widely used foundation T2I models without requiring additional fine-tuning.
Authors: Dasol Choi Jihwan Lee, Minjae Lee, Minsuk Kahng
Abstract: While prior research on text-to-image generation has predominantly focused on biases in human depictions, we investigate a more subtle yet pervasive phenomenon: demographic bias in generated objects (e.g., cars). We introduce SODA (Stereotyped Object Diagnostic Audit), a novel framework for systematically measuring such biases. Our approach compares visual attributes of objects generated with demographic cues (e.g., "for young people'') to those from neutral prompts, across 2,700 images produced by three state-of-the-art models (GPT Image-1, Imagen 4, and Stable Diffusion) in five object categories. Through a comprehensive analysis, we uncover strong associations between specific demographic groups and visual attributes, such as recurring color patterns prompted by gender or ethnicity cues. These patterns reflect and reinforce not only well-known stereotypes but also more subtle and unintuitive biases. We also observe that some models generate less diverse outputs, which in turn amplifies the visual disparities compared to neutral prompts. Our proposed auditing framework offers a practical approach for testing, revealing how stereotypes still remain embedded in today's generative models. We see this as an essential step toward more systematic and responsible AI development.
Authors: Yuanpeng Li, Qi Long, Zhiyuan Yao, Jian Xu, Lintao Xie, Xu He, Lu Geng, Xin Han, Yueyan Chen, Wenbo Duan
Abstract: As enterprise codebases continue to grow in scale and complexity, the volume of lint errors far exceeds engineers' manual remediation capacity, leading to continuous accumulation of technical debt and hindered development efficiency. This paper presents BitsAI-Fix, an automated lint error remediation workflow based on Large Language Models (LLMs), designed to address this critical challenge in industrial-scale environments. BitsAI-Fix employs tree-sitter for context expansion and generates search-and-replace format patches through specially trained LLMs, followed by lint scan re-verification to output final remediation results. Additionally, our approach introduces an innovative progressive reinforcement learning (RL) training strategy that can automatically acquire verifiable training data during the project cold-start phase and continuously iterate the model by collecting online samples through feedback after system deployment. Furthermore, we designed a targeted rule-based reward mechanism that combines format rewards and correctness rewards while penalizing redundant modifications. We also propose a "code diff matching" methodology to continuously track online effectiveness. In production deployment at ByteDance, our solution has supported over 5,000 engineers, resolved more than 12,000 static analysis issues, achieved approximately 85% remediation accuracy, with around 1,000 weekly active adopters. This work demonstrates the practical feasibility of LLM-based code remediation solutions in enterprise environments and serves as a reference for automated code fix in large-scale industrial scenarios.
Authors: Kaiwen Zhao, Bharathan Balaji, Stephen Lee
Abstract: Product sustainability reports provide valuable insights into the environmental impacts of a product and are often distributed in PDF format. These reports often include a combination of tables and text, which complicates their analysis. The lack of standardization and the variability in reporting formats further exacerbate the difficulty of extracting and interpreting relevant information from large volumes of documents. In this paper, we tackle the challenge of answering questions related to carbon footprints within sustainability reports available in PDF format. Unlike previous approaches, our focus is on addressing the difficulties posed by the unstructured and inconsistent nature of text extracted from PDF parsing. To facilitate this analysis, we introduce CarbonPDF-QA, an open-source dataset containing question-answer pairs for 1735 product report documents, along with human-annotated answers. Our analysis shows that GPT-4o struggles to answer questions with data inconsistencies. To address this limitation, we propose CarbonPDF, an LLM-based technique specifically designed to answer carbon footprint questions on such datasets. We develop CarbonPDF by fine-tuning Llama 3 with our training data. Our results show that our technique outperforms current state-of-the-art techniques, including question-answering (QA) systems finetuned on table and text data.
Authors: Mohammadreza Sadeghi, Mahsa Ghazvini Nejad, MirHamed Jafarzadeh Asl, Yu Gu, Yuanhao Yu, Masoud Asgharian, Vahid Partovi Nia
Abstract: Parameter-efficient fine-tuning (PEFT) is essential for reducing the computational overhead of large language models (LLMs). Low-rank family adapters are commonly used to control the parameter size efficiently while maintaining the generative power of LLMs. However, their limited expressiveness due to the rank constraint often restricts their performance on complex tasks. We propose Mixture of Kronecker Adapters (MoKA), a new generation of Kronecker adapters that addresses this limitation by modeling weight updates as a mixture of Kronecker products. Our proposed adapter leverages a gating mechanism that measures the importance of each Kronecker factor, enabling more expressive adaptation. Moreover, MoKA enables a rank flexibility that provides a better trade-off between parameter efficiency and accuracy. To ensure hardware efficiency, we reformulate Kronecker computations using standard matrix operations, allowing seamless deployment on GPU-optimized hardware. We conduct extensive experiments on instruction-tuning and commonsense reasoning tasks using low-bit quantized versions of LLaMA2-7B and LLaMA3-8B models. MoKA not only outperforms PEFT baselines, but also reduces the number of trainable parameters up to 27x, achieving state-of-the-art trade-offs between performance and parameter efficiency.
Authors: Inamullah, Imran Razzak, Shoaib Jameel
Abstract: Retinal microvascular imaging is increasingly recognised as a non invasive method for evaluating systemic vascular and metabolic health. However, the association between lipidomics and retinal vasculature remains inadequate. This study investigates the relationships between serum lipid subclasses, free fatty acids (FA), diacylglycerols (DAG), triacylglycerols (TAG), and cholesteryl esters (CE), and retinal microvascular characteristics in a large population-based cohort. Using Spearman correlation analysis, we examined the interconnection between lipid subclasses and ten retinal microvascular traits, applying the Benjamini-Hochberg false discovery rate (BH-FDR) to adjust for statistical significance. Results indicated that FA were linked to retinal vessel twistiness, while CE correlated with the average widths of arteries and veins. Conversely, DAG and TAG showed negative correlations with the width and complexity of arterioles and venules. These findings suggest that retinal vascular architecture reflects distinct circulating lipid profiles, supporting its role as a non-invasive marker of systemic metabolic health. This study is the first to integrate deep learning (DL)derived retinal traits with lipidomic subclasses in a healthy cohort, thereby providing insights into microvascular structural changes independent of disease status or treatment effects.
Authors: Tianxin Xie, Shan Yang, Chenxing Li, Dong Yu, Li Liu
Abstract: Text-to-speech (TTS) has shown great progress in recent years. However, most existing TTS systems offer only coarse and rigid emotion control, typically via discrete emotion labels or a carefully crafted and detailed emotional text prompt, making fine-grained emotion manipulation either inaccessible or unstable. These models also require extensive, high-quality datasets for training. To address these limitations, we propose EmoSteer-TTS, a novel training-free approach, to achieve fine-grained speech emotion control (conversion, interpolation, erasure) by activation steering. We first empirically observe that modifying a subset of the internal activations within a flow matching-based TTS model can effectively alter the emotional tone of synthesized speech. Building on this insight, we then develop a training-free and efficient algorithm, including activation extraction, emotional token searching, and inference-time steering, which can be seamlessly integrated into a wide range of pretrained models (e.g., F5-TTS, CosyVoice2, and E2-TTS). In addition, to derive effective steering vectors, we construct a curated emotional speech dataset with diverse speakers. Extensive experiments demonstrate that EmoSteer-TTS enables fine-grained, interpretable, and continuous control over speech emotion, outperforming the state-of-the-art (SOTA). To the best of our knowledge, this is the first method that achieves training-free and continuous fine-grained emotion control in TTS.
Authors: Zhanye Luo, Yuefeng Han, Xiufan Yu
Abstract: This paper studies the problem of dimension reduction, tailored to improving time series forecasting with high-dimensional predictors. We propose a novel Supervised Deep Dynamic Principal component analysis (SDDP) framework that incorporates the target variable and lagged observations into the factor extraction process. Assisted by a temporal neural network, we construct target-aware predictors by scaling the original predictors in a supervised manner, with larger weights assigned to predictors with stronger forecasting power. A principal component analysis is then performed on the target-aware predictors to extract the estimated SDDP factors. This supervised factor extraction not only improves predictive accuracy in the downstream forecasting task but also yields more interpretable and target-specific latent factors. Building upon SDDP, we propose a factor-augmented nonlinear dynamic forecasting model that unifies a broad family of factor-model-based forecasting approaches. To further demonstrate the broader applicability of SDDP, we extend our studies to a more challenging scenario when the predictors are only partially observable. We validate the empirical performance of the proposed method on several real-world public datasets. The results show that our algorithm achieves notable improvements in forecasting accuracy compared to state-of-the-art methods.
Authors: Fatih Gulec, Hamdan Awan, Nigel Wallbridge, Andrew W. Eckford
Abstract: Smart agriculture applications, integrating technologies like the Internet of Things and machine learning/artificial intelligence (ML/AI) into agriculture, hold promise to address modern challenges of rising food demand, environmental pollution, and water scarcity. Alongside the concept of the phytobiome, which defines the area including the plant, its environment, and associated organisms, and the recent emergence of molecular communication (MC), there exists an important opportunity to advance agricultural science and practice using communication theory. In this article, we motivate to use the communication engineering perspective for developing a holistic understanding of the phytobiome communication and bridge the gap between the phytobiome communication and smart agriculture. Firstly, an overview of phytobiome communication via molecular and electrophysiological signals is presented and a multi-scale framework modeling the phytobiome as a communication network is conceptualized. Then, how this framework is used to model electrophysiological signals is demonstrated with plant experiments. Furthermore, possible smart agriculture applications, such as smart irrigation and targeted delivery of agrochemicals, through engineering the phytobiome communication are proposed. These applications merge ML/AI methods with the Internet of Bio-Nano-Things enabled by MC and pave the way towards more efficient, sustainable, and eco-friendly agricultural production. Finally, the implementation challenges, open research issues, and industrial outlook for these applications are discussed.
Authors: Yuhan Guo, Lizhong Ding, Shihan Jia, Yanyu Ren, Pengqi Li, Jiarun Fu, Changsheng Li, Ye yuan, Guoren Wang
Abstract: Explainable AI (XAI) builds trust in complex systems through model attribution methods that reveal the decision rationale. However, due to the absence of a unified optimal explanation, existing XAI methods lack a ground truth for objective evaluation and optimization. To address this issue, we propose Deep architecture-based Faith explainer (DeepFaith), a domain-free and model-agnostic unified explanation framework under the lens of faithfulness. By establishing a unified formulation for multiple widely used and well-validated faithfulness metrics, we derive an optimal explanation objective whose solution simultaneously achieves optimal faithfulness across these metrics, thereby providing a ground truth from a theoretical perspective. We design an explainer learning framework that leverages multiple existing explanation methods, applies deduplicating and filtering to construct high-quality supervised explanation signals, and optimizes both pattern consistency loss and local correlation to train a faithful explainer. Once trained, DeepFaith can generate highly faithful explanations through a single forward pass without accessing the model being explained. On 12 diverse explanation tasks spanning 6 models and 6 datasets, DeepFaith achieves the highest overall faithfulness across 10 metrics compared to all baseline methods, highlighting its effectiveness and cross-domain generalizability.
Authors: Wuyang Li, Wentao Pan, Xiaoyuan Liu, Zhendong Luo, Chenxin Li, Hengyu Liu, Din Ping Tsai, Mu Ku Chen, Yixuan Yuan
Abstract: Miniaturized endoscopy has advanced accurate visual perception within the human body. Prevailing research remains limited to conventional cameras employing convex lenses, where the physical constraints with millimetre-scale thickness impose serious impediments on the micro-level clinical. Recently, with the emergence of meta-optics, ultra-micro imaging based on metalenses (micron-scale) has garnered great attention, serving as a promising solution. However, due to the physical difference of metalens, there is a large gap in data acquisition and algorithm research. In light of this, we aim to bridge this unexplored gap, advancing the novel metalens endoscopy. First, we establish datasets for metalens endoscopy and conduct preliminary optical simulation, identifying two derived optical issues that physically adhere to strong optical priors. Second, we propose MetaScope, a novel optics-driven neural network tailored for metalens endoscopy driven by physical optics. MetaScope comprises two novel designs: Optics-informed Intensity Adjustment (OIA), rectifying intensity decay by learning optical embeddings, and Optics-informed Chromatic Correction (OCC), mitigating chromatic aberration by learning spatial deformations informed by learned Point Spread Function (PSF) distributions. To enhance joint learning, we further deploy a gradient-guided distillation to transfer knowledge from the foundational model adaptively. Extensive experiments demonstrate that MetaScope not only outperforms state-of-the-art methods in both metalens segmentation and restoration but also achieves impressive generalized ability in real biomedical scenes.
Authors: Wei Da, Evangelia Kalyvianaki
Abstract: This paper presents Block, a distributed scheduling framework designed to optimize load balancing and auto-provisioning across instances in large language model serving frameworks by leveraging contextual information from incoming requests. Unlike popular model serving systems that rely on monolithic and heuristic task schedulers, Block operates as a fully distributed, stateless, and predictive scheduling system to achieve low overhead, reliability, and scalability. It leverages the deterministic and predictable characteristics of LLM inferences, such as host configurations, response lengths, and hardware performance, to make scheduling decisions based on accurately predicted metrics. Evaluation on a 12 GPUs cluster shows that Block significantly outperforms heuristic schedulers, boosting serving capacity by up to 16.7\% and reducing P99 tail latency by up to 49.5\%. These performance gains remain consistent across diverse models, workloads and configurations. Code and data are open-sourced.
Authors: Yong Lin, Shange Tang, Bohan Lyu, Ziran Yang, Jui-Hui Chung, Haoyu Zhao, Lai Jiang, Yihan Geng, Jiawei Ge, Jingruo Sun, Jiayun Wu, Jiri Gesi, Ximing Lu, David Acuna, Kaiyu Yang, Hongzhou Lin, Yejin Choi, Danqi Chen, Sanjeev Arora, Chi Jin
Abstract: We introduce Goedel-Prover-V2, a series of open-source language models that set a new state-of-the-art in automated theorem proving. Built on the standard expert iteration and reinforcement learning pipeline, our approach incorporates three key innovations: (1) Scaffolded data synthesis: We generate synthetic tasks of increasing difficulty to train the model to master increasingly complex theorems; (2) Verifier-guided self-correction: We enable the model to iteratively revise its proofs by leveraging feedback from the Lean compiler; (3) Model averaging: We merge model checkpoints to mitigate the decrease in model output diversity in later stages of training. Our small model, Goedel-Prover-V2-8B, reaches 84.6% pass@32 on MiniF2F and outperforms DeepSeek-Prover-V2-671B under the same metric, despite being 80X smaller. Our flagship model, Goedel-Prover-V2-32B, achieves 88.1% on MiniF2F at pass@32 in standard mode and 90.4% in self-correction mode, outperforming prior SOTA by a large margin. Additionally, our flagship model solves 86 problems on PutnamBench at pass@184, securing the first place among open-source models on the leaderboard, surpassing DeepSeek-Prover-V2-671B's record of solving 47 problems by pass@1024 with a significantly smaller model size and compute budget. At the time of its release (July-August 2025), Goedel-Prover-V2 achieves the strongest overall performance among all open-source theorem provers. It also ranks among the top-performing models--including closed-source systems with publicly reported performance--under a constrained test-time compute budget. Our models, code, and data are released at https://github.com/Goedel-LM/Goedel-Prover-V2.
Authors: Daniel DeAlcala, Aythami Morales, Julian Fierrez, Ruben Tolosana
Abstract: We present Attention Zoom, a modular and model-agnostic spatial attention mechanism designed to improve feature extraction in convolutional neural networks (CNNs). Unlike traditional attention approaches that require architecture-specific integration, our method introduces a standalone layer that spatially emphasizes high-importance regions in the input. We evaluated Attention Zoom on multiple CNN backbones using CIFAR-100 and TinyImageNet, showing consistent improvements in Top-1 and Top-5 classification accuracy. Visual analyses using Grad-CAM and spatial warping reveal that our method encourages fine-grained and diverse attention patterns. Our results confirm the effectiveness and generality of the proposed layer for improving CCNs with minimal architectural overhead.
Authors: Soumik Dey, Benjamin Braun, Naveen Ravipati, Hansi Wu, Binbin Li
Abstract: Sellers at eBay are recommended keyphrases to bid on to enhance the performance of their advertising campaigns. The relevance of these keyphrases is crucial in avoiding the overcrowding of search systems with irrelevant items and maintaining a positive seller perception. It is essential that keyphrase recommendations align with both seller and Search judgments regarding auctions. Due to the difficulty in procuring negative human judgment at scale, employing LLM-as-a-judge to mimic seller judgment has been established as the norm in several studies. This study introduces a novel two-step LLM distillation process from a LLM-judge used to debias our Embedding Based Retrieval (EBR) model from the various biases that exist in click-data. We distill from an LLM teacher via a cross-encoder assistant into a bi-encoder student using a multi-task training approach, ultimately employing the student bi-encoder to retrieve relevant advertiser keyphrases. We show that integrating a knowledge distillation process from LLMs in a multi-task training setup enhances bi-encoder performance in retrieving relevant advertiser keyphrases at eBay.
Authors: Saleh Nikooroo, Thomas Engel
Abstract: The internal representations learned by deep networks are often sensitive to architecture-specific choices, raising questions about the stability, alignment, and transferability of learned structure across models. In this paper, we investigate how structural constraints--such as linear shaping operators and corrective paths--affect the compatibility of internal representations across different architectures. Building on the insights from prior studies on structured transformations and convergence, we develop a framework for measuring and analyzing representational alignment across networks with distinct but related architectural priors. Through a combination of theoretical insights, empirical probes, and controlled transfer experiments, we demonstrate that structural regularities induce representational geometry that is more stable under architectural variation. This suggests that certain forms of inductive bias not only support generalization within a model, but also improve the interoperability of learned features across models. We conclude with a discussion on the implications of representational transferability for model distillation, modular learning, and the principled design of robust learning systems.
Authors: Ruei-Che Chang, Rosiana Natalie, Wenqian Xu, Jovan Zheng Feng Yap, Anhong Guo
Abstract: Recent advancements in large multimodal models have provided blind or visually impaired (BVI) individuals with new capabilities to interpret and engage with the real world through interactive systems that utilize live video feeds. However, the potential benefits and challenges of such capabilities to support diverse real-world assistive tasks remain unclear. In this paper, we present findings from an exploratory study with eight BVI participants. Participants used ChatGPT's Advanced Voice with Video, a state-of-the-art live video AI released in late 2024, in various real-world scenarios, from locating objects to recognizing visual landmarks, across unfamiliar indoor and outdoor environments. Our findings indicate that current live video AI effectively provides guidance and answers for static visual scenes but falls short in delivering essential live descriptions required in dynamic situations. Despite inaccuracies in spatial and distance information, participants leveraged the provided visual information to supplement their mobility strategies. Although the system was perceived as human-like due to high-quality voice interactions, assumptions about users' visual abilities, hallucinations, generic responses, and a tendency towards sycophancy led to confusion, distrust, and potential risks for BVI users. Based on the results, we discuss implications for assistive video AI agents, including incorporating additional sensing capabilities for real-world use, determining appropriate intervention timing beyond turn-taking interactions, and addressing ecological and safety concerns.
Authors: Deepak Pandita, Flip Korn, Chris Welty, Christopher M. Homan
Abstract: Reproducibility is a cornerstone of scientific validation and of the authority it confers on its results. Reproducibility in machine learning evaluations leads to greater trust, confidence, and value. However, the ground truth responses used in machine learning often necessarily come from humans, among whom disagreement is prevalent, and surprisingly little research has studied the impact of effectively ignoring disagreement in these responses, as is typically the case. One reason for the lack of research is that budgets for collecting human-annotated evaluation data are limited, and obtaining more samples from multiple annotators for each example greatly increases the per-item annotation costs. We investigate the trade-off between the number of items ($N$) and the number of responses per item ($K$) needed for reliable machine learning evaluation. We analyze a diverse collection of categorical datasets for which multiple annotations per item exist, and simulated distributions fit to these datasets, to determine the optimal $(N, K)$ configuration, given a fixed budget ($N \times K$), for collecting evaluation data and reliably comparing the performance of machine learning models. Our findings show, first, that accounting for human disagreement may come with $N \times K$ at no more than 1000 (and often much lower) for every dataset tested on at least one metric. Moreover, this minimal $N \times K$ almost always occurred for $K > 10$. Furthermore, the nature of the tradeoff between $K$ and $N$ -- or if one even existed -- depends on the evaluation metric, with metrics that are more sensitive to the full distribution of responses performing better at higher levels of $K$. Our methods can be used to help ML practitioners get more effective test data by finding the optimal metrics and number of items and annotations per item to collect to get the most reliability for their budget.
Authors: Claudiu Leoveanu-Condrei
Abstract: Generative models, particularly Large Language Models (LLMs), produce fluent outputs yet lack verifiable guarantees. We adapt Design by Contract (DbC) and type-theoretic principles to introduce a contract layer that mediates every LLM call. Contracts stipulate semantic and type requirements on inputs and outputs, coupled with probabilistic remediation to steer generation toward compliance. The layer exposes the dual view of LLMs as semantic parsers and probabilistic black-box components. Contract satisfaction is probabilistic and semantic validation is operationally defined through programmer-specified conditions on well-typed data structures. More broadly, this work postulates that any two agents satisfying the same contracts are \emph{functionally equivalent} with respect to those contracts.
Authors: Willem Fourie
Abstract: In the discourse on AI regulation, 'responsible AI' is the dominant paradigm, with the focus on mitigating the risks related to AI systems. While this focus is important and necessary, it has limited use for a systematic consideration of AI's societal impact. This paper proposes a proto-framework for assessing the societal impact of AI systems by operationalising the concept of freedom. This proto-framework is intended as a step towards a fully operationalised framework to be used in policymaking contexts. By drawing on Kantian philosophy and related contemporary interpretations, freedom is developed as the counterpart to the concept of responsibility. Two dimensions of freedom are developed in further detail: freedom as capability and freedom as opportunity. These two dimensions of freedom are then applied in a proto-framework that systematically considers AI's impact on society using the Sustainable Development Goals. This proto-framework aims to complement current risk-based approaches and thereby offers a first step towards operationalising the concept of freedom in AI regulation.
Authors: Shengnan Yang, Rongqian Ma
Abstract: As AI systems become integral to knowledge-intensive work, questions arise not only about their functionality but also their epistemic roles in human-AI interaction. While HCI research has proposed various AI role typologies, it often overlooks how AI reshapes users' roles as knowledge contributors. This study examines how users form epistemic relationships with AI-how they assess, trust, and collaborate with it in research and teaching contexts. Based on 31 interviews with academics across disciplines, we developed a five-part codebook and identified five relationship types: Instrumental Reliance, Contingent Delegation, Co-agency Collaboration, Authority Displacement, and Epistemic Abstention. These reflect variations in trust, assessment modes, tasks, and human epistemic status. Our findings show that epistemic roles are dynamic and context-dependent. We argue for shifting beyond static metaphors of AI toward a more nuanced framework that captures how humans and AI co-construct knowledge, enriching HCI's understanding of the relational and normative dimensions of AI use.
Authors: Lili Chen, Mihir Prabhudesai, Katerina Fragkiadaki, Hao Liu, Deepak Pathak
Abstract: Can large language models improve without external data -- by generating their own questions and answers? We hypothesize that a pre-trained language model can improve its reasoning skills given only a single prompt specifying the topic (e.g., algebra word problems) and asking the model to generate its own questions. To do this, we propose Self-Questioning Language Models (SQLM): an asymmetric self-play framework where a proposer is given the topic and generates a question for a solver, who tries to answer it. Both the proposer and solver are trained via reinforcement learning. The proposer receives a reward if the problem is not too easy or too difficult, and the solver receives a reward based on majority voting, a proxy for correctness in the absence of ground-truth answers. For coding, the proposer can instead generate unit tests which are used for verification. We study this asymmetric self-play framework on three benchmarks: three-digit multiplication, algebra problems from the OMEGA benchmark, and programming problems from Codeforces. By continually generating more interesting problems and attempting to solve them, language models can improve on downstream benchmarks without access to any curated training datasets.
Authors: Shudong Liu, Hongwei Liu, Junnan Liu, Linchen Xiao, Songyang Gao, Chengqi Lyu, Yuzhe Gu, Wenwei Zhang, Derek F. Wong, Songyang Zhang, Kai Chen
Abstract: Answer verification is crucial not only for evaluating large language models (LLMs) by matching their unstructured outputs against standard answers, but also serves as the reward model to guide LLM optimization. Most evaluation frameworks rely on regularized matching or employ general LLMs for answer verification, which demands extensive, repetitive customization for regex rules or evaluation prompts. Two fundamental limitations persist in current methodologies: 1) the absence of comprehensive benchmarks that systematically evaluate verification capabilities across different LLMs; and 2) the nascent stage of verifier development, where existing approaches lack both the robustness to handle complex edge cases and the generalizability across different domains. In this work, we develop CompassVerifier, an accurate and robust lightweight verifier model for evaluation and outcome reward. It demonstrates multi-domain competency spanning math, knowledge, and diverse reasoning tasks, with the capability to process various answer types, including multi-subproblems, formulas, and sequence answers, while effectively identifying abnormal/invalid responses. We introduce VerifierBench benchmark comprising model outputs collected from multiple data sources, augmented through manual analysis of metaerror patterns to enhance CompassVerifier. We anticipate that CompassVerifier and VerifierBench will facilitate answer verification, evaluation protocols, and reinforcement learning research. Code and dataset are available at https://github.com/open-compass/CompassVerifier.
Authors: Nadav Amir, Stas Tiomkin
Abstract: Computational models of purposeful behavior comprise both descriptive and prescriptive aspects, used respectively to ascertain and evaluate situations in the world. In reinforcement learning, prescriptive reward functions are assumed to depend on predefined and fixed descriptive state representations. Alternatively, these two aspects may emerge interdependently: goals can shape the acquired state representations and vice versa. Here, we present a computational framework for state representation learning in bounded agents, where descriptive and prescriptive aspects are coupled through the notion of goal-directed, or telic, states. We introduce the concept of telic-controllability to characterize the tradeoff between the granularity of a telic state representation and the policy complexity required to reach all telic states. We propose an algorithm for learning telic-controllable state representations, illustrating it using a simulated navigation task. Our framework highlights the role of deliberate ignorance -- knowing what to ignore -- for learning state representations that balance goal flexibility and cognitive complexity.
Authors: Jilong Li, Zhenxi Song, Jiaqi Wang, Meishan Zhang, Honghai Liu, Min Zhang, Zhiguo Zhang
Abstract: Current EEG/MEG-to-text decoding systems suffer from three key limitations: (1) reliance on teacher-forcing methods, which compromises robustness during inference, (2) sensitivity to session-specific noise, hindering generalization across subjects, and (3) misalignment between brain signals and linguistic representations due to pre-trained language model over-dominance. To overcome these challenges, we propose BrainECHO (Brain signal decoding via vEctor-quantized speCtrogram reconstruction for WHisper-enhanced text generatiOn), a multi-stage framework that employs decoupled representation learning to achieve state-of-the-art performance on both EEG and MEG datasets. Specifically, BrainECHO consists of three stages: (1) Discrete autoencoding, which transforms continuous Mel spectrograms into a finite set of high-quality discrete representations for subsequent stages. (2) Frozen alignment, where brain signal embeddings are mapped to corresponding Mel spectrogram embeddings in a frozen latent space, effectively filtering session-specific noise through vector-quantized reconstruction, yielding a 3.65% improvement in BLEU-4 score. (3) Constrained decoding fine-tuning, which leverages the pre-trained Whisper model for audio-to-text translation, balancing signal adaptation with knowledge preservation, and achieving 74%-89% decoding BLEU scores without excessive reliance on teacher forcing. BrainECHO demonstrates robustness across sentence, session, and subject-independent conditions, passing Gaussian noise tests and showcasing its potential for enhancing language-based brain-computer interfaces.
Authors: Alexander Tuisov, Yonatan Vernik, Alexander Shleyfman
Abstract: Domain-independent heuristics have long been a cornerstone of AI planning, offering general solutions applicable across a wide range of tasks without requiring domain-specific engineering. However, the advent of large language models (LLMs) presents an opportunity to generate heuristics tailored to specific planning problems, potentially challenging the necessity of domain independence as a strict design principle. In this paper, we explore the use of LLMs to automatically derive planning heuristics from task descriptions represented as successor generators and goal tests written in general purpose programming language. We investigate the trade-offs between domain-specific LLM-generated heuristics and traditional domain-independent methods in terms of computational efficiency and explainability. Our experiments demonstrate that LLMs can create heuristics that achieve state-of-the-art performance on some standard IPC domains, as well as their ability to solve problems that lack an adequate Planning Domain Definition Language ({\sc pddl}) representation. We discuss whether these results signify a paradigm shift and how they can complement existing approaches.
Authors: Haoran Sun, Yusen Wu, Peng Wang, Wei Chen, Yukun Cheng, Xiaotie Deng, Xu Chu
Abstract: Game theory is a foundational framework for analyzing strategic interactions, and its intersection with large language models (LLMs) is a rapidly growing field. However, existing surveys mainly focus narrowly on using game theory to evaluate LLM behavior. This paper provides the first comprehensive survey of the bidirectional relationship between Game Theory and LLMs. We propose a novel taxonomy that categorizes the research in this intersection into four distinct perspectives: (1) evaluating LLMs in game-based scenarios; (2) improving LLMs using game-theoretic concepts for better interpretability and alignment; (3) modeling the competitive landscape of LLM development and its societal impact; and (4) leveraging LLMs to advance game models and to solve corresponding game theory problems. Furthermore, we identify key challenges and outline future research directions. By systematically investigating this interdisciplinary landscape, our survey highlights the mutual influence of game theory and LLMs, fostering progress at the intersection of these fields.
Authors: Longling Geng, Edward Y. Chang
Abstract: This benchmark suite provides a comprehensive evaluation framework for assessing both individual LLMs and multi-agent systems in Real-world planning and scheduling scenarios. The suite encompasses 14 designed planning and scheduling problems that progress from basic to highly complex, incorporating key aspects such as multi-agent coordination, inter-agent dependencies, and dynamic environmental disruptions. Each problem can be scaled along three dimensions: the number of parallel planning threads, the complexity of inter-dependencies, and the frequency of unexpected disruptions requiring Real-time adaptation. The benchmark includes 14 detailed problem specifications, 15 comparison methods including Random, LPT, SPT, STPT, MPSR, DRL-Liu, GP, GEP, LSO, SPT/TWKR, DRL-Chen, DRL-Zhang, 2+ evaluation metrics, and baseline implementations using 3+ LLMs including GPT-4o, Claude-3.7, DeepSeek-R1, and 4 contemporary frameworks including LangGraph, AutoGen, CrewAI, and Swarm, enabling rigorous testing of both single-agent and multi-agent planning capabilities. Through standardized evaluation criteria and scalable complexity, this benchmark aims to be opened to public, and drive progress in developing more adaptable, robust, and scalable AI planning systems for Real-world applications.
Authors: Liangbo Ning, Ziran Liang, Zhuohang Jiang, Haohao Qu, Yujuan Ding, Wenqi Fan, Xiao-yong Wei, Shanru Lin, Hui Liu, Philip S. Yu, Qing Li
Abstract: With the advancement of web techniques, they have significantly revolutionized various aspects of people's lives. Despite the importance of the web, many tasks performed on it are repetitive and time-consuming, negatively impacting overall quality of life. To efficiently handle these tedious daily tasks, one of the most promising approaches is to advance autonomous agents based on Artificial Intelligence (AI) techniques, referred to as AI Agents, as they can operate continuously without fatigue or performance degradation. In the context of the web, leveraging AI Agents -- termed WebAgents -- to automatically assist people in handling tedious daily tasks can dramatically enhance productivity and efficiency. Recently, Large Foundation Models (LFMs) containing billions of parameters have exhibited human-like language understanding and reasoning capabilities, showing proficiency in performing various complex tasks. This naturally raises the question: `Can LFMs be utilized to develop powerful AI Agents that automatically handle web tasks, providing significant convenience to users?' To fully explore the potential of LFMs, extensive research has emerged on WebAgents designed to complete daily web tasks according to user instructions, significantly enhancing the convenience of daily human life. In this survey, we comprehensively review existing research studies on WebAgents across three key aspects: architectures, training, and trustworthiness. Additionally, several promising directions for future research are explored to provide deeper insights.
Authors: Tzu-Han Hsu, Arshia Rafieioskouei, Borzoo Bonakdarpour
Abstract: Reward shaping in multi-agent reinforcement learning (MARL) for complex tasks remains a significant challenge. Existing approaches often fail to find optimal solutions or cannot efficiently handle such tasks. We propose HYPRL, a specification-guided reinforcement learning framework that learns control policies w.r.t. hyperproperties expressed in HyperLTL. Hyperproperties constitute a powerful formalism for specifying objectives and constraints over sets of execution traces across agents. To learn policies that maximize the satisfaction of a HyperLTL formula $\phi$, we apply Skolemization to manage quantifier alternations and define quantitative robustness functions to shape rewards over execution traces of a Markov decision process with unknown transitions. A suitable RL algorithm is then used to learn policies that collectively maximize the expected reward and, consequently, increase the probability of satisfying $\phi$. We evaluate HYPRL on a diverse set of benchmarks, including safety-aware planning, Deep Sea Treasure, and the Post Correspondence Problem. We also compare with specification-driven baselines to demonstrate the effectiveness and efficiency of HYPRL.
Authors: Xinan Chen, Rong Qu, Jing Dong, Ruibin Bai, Yaochu Jin
Abstract: Dynamic scheduling in real-world environments often struggles to adapt to unforeseen disruptions, making traditional static scheduling methods and human-designed heuristics inadequate. This paper introduces an innovative approach that combines Genetic Programming (GP) with a Transformer trained through Reinforcement Learning (GPRT), specifically designed to tackle the complexities of dynamic scheduling scenarios. GPRT leverages the Transformer to refine heuristics generated by GP while also seeding and guiding the evolution of GP. This dual functionality enhances the adaptability and effectiveness of the scheduling heuristics, enabling them to better respond to the dynamic nature of real-world tasks. The efficacy of this integrated approach is demonstrated through a practical application in container terminal truck scheduling, where the GPRT method outperforms traditional GP, standalone Transformer methods, and other state-of-the-art competitors. The key contribution of this research is the development of the GPRT method, which showcases a novel combination of GP and Reinforcement Learning (RL) to produce robust and efficient scheduling solutions. Importantly, GPRT is not limited to container port truck scheduling; it offers a versatile framework applicable to various dynamic scheduling challenges. Its practicality, coupled with its interpretability and ease of modification, makes it a valuable tool for diverse real-world scenarios.
Authors: Zixuan Ke, Fangkai Jiao, Yifei Ming, Xuan-Phi Nguyen, Austin Xu, Do Xuan Long, Minzhi Li, Chengwei Qin, Peifeng Wang, Silvio Savarese, Caiming Xiong, Shafiq Joty
Abstract: Reasoning is a fundamental cognitive process that enables logical inference, problem-solving, and decision-making. With the rapid advancement of large language models (LLMs), reasoning has emerged as a key capability that distinguishes advanced AI systems from conventional models that empower chatbots. In this survey, we categorize existing methods along two orthogonal dimensions: (1) Regimes, which define the stage at which reasoning is achieved (either at inference time or through dedicated training); and (2) Architectures, which determine the components involved in the reasoning process, distinguishing between standalone LLMs and agentic compound systems that incorporate external tools, and multi-agent collaborations. Within each dimension, we analyze two key perspectives: (1) Input level, which focuses on techniques that construct high-quality prompts that the LLM condition on; and (2) Output level, which methods that refine multiple sampled candidates to enhance reasoning quality. This categorization provides a systematic understanding of the evolving landscape of LLM reasoning, highlighting emerging trends such as the shift from inference-scaling to learning-to-reason (e.g., DeepSeek-R1), and the transition to agentic workflows (e.g., OpenAI Deep Research, Manus Agent). Additionally, we cover a broad spectrum of learning algorithms, from supervised fine-tuning to reinforcement learning such as PPO and GRPO, and the training of reasoners and verifiers. We also examine key designs of agentic workflows, from established patterns like generator-evaluator and LLM debate to recent innovations. ...
Authors: Yash Savani, Asher Trockman, Zhili Feng, Avi Schwarzschild, Alexander Robey, Marc Finzi, J. Zico Kolter
Abstract: Frontier models that generate extended reasoning traces inadvertently produce rich token sequences that can facilitate model distillation. Recognizing this vulnerability, model owners may seek sampling strategies that limit the effectiveness of distillation without compromising model performance. Antidistillation sampling provides exactly this capability. By strategically modifying a model's next-token probability distribution, antidistillation sampling poisons reasoning traces, rendering them significantly less effective for distillation while preserving the model's practical utility. For further details, see https://antidistillation.com.
Authors: Joseph Lee, Tianqi Shang, Jae Young Baik, Duy Duong-Tran, Shu Yang, Lingyao Li, Li Shen
Abstract: Large Language Models (LLMs) have shown promise in clinical decision support, yet their application to triage remains underexplored. We systematically investigate the capabilities of LLMs in emergency department triage through two key dimensions: (1) robustness to distribution shifts and missing data, and (2) counterfactual analysis of intersectional biases across sex and race. We assess multiple LLM-based approaches, ranging from continued pre-training to in-context learning, as well as machine learning approaches. Our results indicate that LLMs exhibit superior robustness, and we investigate the key factors contributing to the promising LLM-based approaches. Furthermore, in this setting, we identify gaps in LLM preferences that emerge in particular intersections of sex and race. LLMs generally exhibit sex-based differences, but they are most pronounced in certain racial groups. These findings suggest that LLMs encode demographic preferences that may emerge in specific clinical contexts or particular combinations of characteristics.
Authors: Shibo Hong, Jiahao Ying, Haiyuan Liang, Mengdi Zhang, Jun Kuang, Jiazheng Zhang, Yixin Cao
Abstract: Evaluating the open-ended outputs of Large Multimodal Models has become a bottleneck as model capabilities, task diversity, and modality rapidly expand. Existing "LLM-as-a-Judge" evaluators are typically narrow in specific tasks and aspects. In this paper, we argue that, on one hand, based on the interconnected nature of aspects, learning specific aspects can generalize to unseen aspects; on the other hand, jointly learning to assess multiple visual aspects and tasks may foster a synergistic effect. To this end, we propose UFEval, the first unified fine-grained evaluator with task and aspect generalization for four evaluation tasks -- Natural Language Generation, Image Understanding, Image Generation, and Interleaved Text-and-Image Generation. Specifically, (1) We first construct a hierarchical aspect taxonomy encompassing 112 distinct aspects across the aforementioned four tasks. (2) Then, building upon this taxonomy, we create FRABench, a fine-grained evaluation dataset comprising 60.4k pairwise samples with 325k evaluation labels obtained from a combination of human and GPT-4o annotations. FRABench provides a large-scale, multi-modal, and aspect-level resource for training and testing evaluators. (3) Finally, leveraging FRABench, we develop UFEval, a unified fine-grained evaluator. Experiments show that learning on specific aspects enables UFEval to generalize to unseen aspects, and joint learning to assess diverse tasks and aspects can lead to substantial mutual benefits.
Authors: Sota Yoshihara (Graduate School of Mathematics, Nagoya University), Ryosuke Yamamoto (AISIN SOFTWARE Co., Ltd), Hiroyuki Kusumoto (Graduate School of Mathematics, Nagoya University), Masanari Shimura (Graduate School of Mathematics, Nagoya University)
Abstract: This paper proposes a novel theoretical framework for guaranteeing and evaluating the resilience of long short-term memory (LSTM) networks in control systems. We introduce "recovery time" as a new metric of resilience in order to quantify the time required for an LSTM to return to its normal state after anomalous inputs. By mathematically refining incremental input-to-state stability ($\delta$ISS) theory for LSTM, we derive a practical data-independent upper bound on recovery time. This upper bound gives us resilience-aware training. Experimental validation on simple models demonstrates the effectiveness of our resilience estimation and control methods, enhancing a foundation for rigorous quality assurance in safety-critical AI applications.
Authors: Mario Leiva, Noel Ngu, Joshua Shay Kricheli, Aditya Taparia, Ransalu Senanayake, Paulo Shakarian, Nathaniel Bastian, John Corcoran, Gerardo Simari
Abstract: The deployment of pre-trained perception models in novel environments often leads to performance degradation due to distributional shifts. Although recent artificial intelligence approaches for metacognition use logical rules to characterize and filter model errors, improving precision often comes at the cost of reduced recall. This paper addresses the hypothesis that leveraging multiple pre-trained models can mitigate this recall reduction. We formulate the challenge of identifying and managing conflicting predictions from various models as a consistency-based abduction problem, building on the idea of abductive learning (ABL) but applying it to test-time instead of training. The input predictions and the learned error detection rules derived from each model are encoded in a logic program. We then seek an abductive explanation--a subset of model predictions--that maximizes prediction coverage while ensuring the rate of logical inconsistencies (derived from domain constraints) remains below a specified threshold. We propose two algorithms for this knowledge representation task: an exact method based on Integer Programming (IP) and an efficient Heuristic Search (HS). Through extensive experiments on a simulated aerial imagery dataset featuring controlled, complex distributional shifts, we demonstrate that our abduction-based framework outperforms individual models and standard ensemble baselines, achieving, for instance, average relative improvements of approximately 13.6\% in F1-score and 16.6\% in accuracy across 15 diverse test datasets when compared to the best individual model. Our results validate the use of consistency-based abduction as an effective mechanism to robustly integrate knowledge from multiple imperfect models in challenging, novel scenarios.
Authors: Edward Y. Chang, Zeyneb N. Kaya, Ethan Chang
Abstract: Unified Cognitive Consciousness Theory} (UCCT) casts them instead as vast unconscious pattern repositories: apparent reasoning arises only when external anchoring mechanisms, few shot prompts, retrieval-augmented context, fine-tuning, or multi-agent debate, activate task-relevant patterns. UCCT formalizes this process as Bayesian competition between statistical priors learned in pre-training and context-driven target patterns, yielding a single quantitative account that unifies existing adaptation techniques. We ground the theory in three principles: threshold crossing, modality universality, and density-distance predictive power, and validate them with (i) cross-domain demonstrations (text QA, image captioning, multi-agent debate) and (ii) two depth-oriented experiments: a controlled numeral-base study (bases 8, 9, 10) that isolates pattern-density effects, and a layer-wise trajectory analysis that reveals phase transitions inside a 7B-parameter model. Both experiments confirm UCCT's predictions of threshold behavior, asymmetric interference, and memory hysteresis. By showing that LLM ``intelligence'' is created through semantic anchoring rather than contained within the model, UCCT offers a principled foundation for interpretable diagnostics and practical guidance for prompt engineering, model selection, and alignment-centric system design.
Authors: Nandana Mihindukulasooriya, Niharika S. D'Souza, Faisal Chowdhury, Horst Samulowitz
Abstract: A KG represents a network of entities and illustrates relationships between them. KGs are used for various applications, including semantic search and discovery, reasoning, decision-making, natural language processing, machine learning, and recommendation systems. Triple (subject-relation-object) extraction from text is the fundamental building block of KG construction and has been widely studied, for example, in early benchmarks such as ACE 2002 to more recent ones, such as WebNLG 2020, REBEL and SynthIE. While the use of LLMs is explored for KG construction, handcrafting reasonable task-specific prompts for LLMs is a labour-intensive exercise and can be brittle due to subtle changes in the LLM models employed. Recent work in NLP tasks (e.g. autonomy generation) uses automatic prompt optimization/engineering to address this challenge by generating optimal or near-optimal task-specific prompts given input-output examples. This empirical study explores the application of automatic prompt optimization for the triple extraction task using experimental benchmarking. We evaluate different settings by changing (a) the prompting strategy, (b) the LLM being used for prompt optimization and task execution, (c) the number of canonical relations in the schema (schema complexity), (d) the length and diversity of input text, (e) the metric used to drive the prompt optimization, and (f) the dataset being used for training and testing. We evaluate three different automatic prompt optimizers, namely, DSPy, APE, and TextGrad and use two different triple extraction datasets, SynthIE and REBEL. Through rigorous empirical evaluation, our main contribution highlights that automatic prompt optimization techniques can generate reasonable prompts similar to humans for triple extraction. In turn, these optimized prompts achieve improved results, particularly with increasing schema complexity and text size.
Authors: Gopal Gupta, Abhiramon Rajasekharan, Alexis R. Tudor, Elmer Salazar, Joaqu\'in Arias
Abstract: We consider the problem of implementing deontic modal logic. We show how (deontic) modal operators can be expressed elegantly using default negation (negation-as-failure) and strong negation present in answer set programming (ASP). We propose using global constraints of ASP to represent obligations and impermissibilities of deontic modal logic. We show that our proposed representation results in the various paradoxes of deontic modal logic being elegantly resolved.
Authors: Peter Baumgartner, Lachlan McGinness
Abstract: We present our method for automatically marking Physics exams. The marking problem consists in assessing typed student answers for correctness with respect to a ground truth solution. This is a challenging problem that we seek to tackle using a combination of a computer algebra system, an SMT solver and a term rewriting system. A Large Language Model is used to interpret and remove errors from student responses and rewrite these in a machine readable format. Once formalized and language-aligned, the next step then consists in applying automated reasoning techniques for assessing student solution correctness. We consider two methods of automated theorem proving: off-the-shelf SMT solving and term rewriting systems tailored for physics problems involving trigonometric expressions. The development of the term rewrite system and establishing termination and confluence properties was not trivial, and we describe it in some detail in the paper. We evaluate our system on a rich pool of over 1500 real-world student exam responses from the 2023 Australian Physics Olympiad.
Authors: Stefanos Gkikas, Ioannis Kyprakis, Manolis Tsiknakis
Abstract: Pain is a complex and pervasive condition that affects a significant portion of the population. Accurate and consistent assessment is essential for individuals suffering from pain, as well as for developing effective management strategies in a healthcare system. Automatic pain assessment systems enable continuous monitoring, support clinical decision-making, and help minimize patient distress while mitigating the risk of functional deterioration. Leveraging physiological signals offers objective and precise insights into a person's state, and their integration in a multimodal framework can further enhance system performance. This study has been submitted to the \textit{Second Multimodal Sensing Grand Challenge for Next-Gen Pain Assessment (AI4PAIN)}. The proposed approach introduces \textit{Tiny-BioMoE}, a lightweight pretrained embedding model for biosignal analysis. Trained on $4.4$ million biosignal image representations and consisting of only $7.3$ million parameters, it serves as an effective tool for extracting high-quality embeddings for downstream tasks. Extensive experiments involving electrodermal activity, blood volume pulse, respiratory signals, peripheral oxygen saturation, and their combinations highlight the model's effectiveness across diverse modalities in automatic pain recognition tasks. \textit{\textcolor{blue}{The model's architecture (code) and weights are available at https://github.com/GkikasStefanos/Tiny-BioMoE.
Authors: Jobst Heitzig, Ram Potham
Abstract: Power is a key concept in AI safety: power-seeking as an instrumental goal, sudden or gradual disempowerment of humans, power balance in human-AI interaction and international AI governance. At the same time, power as the ability to pursue diverse goals is essential for wellbeing. This paper explores the idea of promoting both safety and wellbeing by forcing AI agents explicitly to empower humans and to manage the power balance between humans and AI agents in a desirable way. Using a principled, partially axiomatic approach, we design a parametrizable and decomposable objective function that represents an inequality- and risk-averse long-term aggregate of human power. It takes into account humans' bounded rationality and social norms, and, crucially, considers a wide variety of possible human goals. We derive algorithms for computing that metric by backward induction or approximating it via a form of multi-agent reinforcement learning from a given world model. We exemplify the consequences of (softly) maximizing this metric in a variety of paradigmatic situations and describe what instrumental sub-goals it will likely imply. Our cautious assessment is that softly maximizing suitable aggregate metrics of human power might constitute a beneficial objective for agentic AI systems that is safer than direct utility-based objectives.
Authors: Yihong Dong, Xue Jiang, Yongding Tao, Huanyu Liu, Kechi Zhang, Lili Mou, Rongyu Cao, Yingwei Ma, Jue Chen, Binhua Li, Zhi Jin, Fei Huang, Yongbin Li, Ge Li
Abstract: Reinforcement Learning with Verifiable Reward (RLVR) has significantly advanced the complex reasoning abilities of Large Language Models (LLMs). However, it struggles to break through the inherent capability boundaries of the base LLM, due to its essentially on-policy strategy coupled with LLM's immense action space and sparse reward. Critically, RLVR can lead to the capability boundary collapse, narrowing the LLM's problem-solving scope. To address this problem, we propose RL-PLUS, a novel hybrid-policy optimization approach for LLMs that synergizes internal exploitation with external data to achieve stronger reasoning capabilities and surpass the boundaries of base models. RL-PLUS integrates two core components, i.e., Multiple Importance Sampling to address distributional mismatch from external data, and Exploration-Based Advantage Function to guide the model towards high-value, unexplored reasoning paths. We provide both theoretical analysis and extensive experiments to demonstrate the superiority and generalizability of our approach. Compared with existing RLVR methods, RL-PLUS achieves 1) state-of-the-art performance on six math reasoning benchmarks; 2) superior performance on six out-of-distribution reasoning tasks; 3) consistent and significant gains across diverse model families, with average relative improvements up to 69.2\%. Moreover, the analysis of Pass@k curves indicates that RL-PLUS effectively resolves the capability boundary collapse problem.
Authors: Yi-Long Lu, Jiajun Song, Chunhui Zhang, Wei Wang
Abstract: Humans constantly generate a diverse range of tasks guided by internal motivations. While generative agents powered by large language models (LLMs) aim to simulate this complex behavior, it remains uncertain whether they operate on similar cognitive principles. To address this, we conducted a task-generation experiment comparing human responses with those of an LLM agent (GPT-4o). We find that human task generation is consistently influenced by psychological drivers, including personal values (e.g., Openness to Change) and cognitive style. Even when these psychological drivers are explicitly provided to the LLM, it fails to reflect the corresponding behavioral patterns. They produce tasks that are markedly less social, less physical, and thematically biased toward abstraction. Interestingly, while the LLM's tasks were perceived as more fun and novel, this highlights a disconnect between its linguistic proficiency and its capacity to generate human-like, embodied goals. We conclude that there is a core gap between the value-driven, embodied nature of human cognition and the statistical patterns of LLMs, highlighting the necessity of incorporating intrinsic motivation and physical grounding into the design of more human-aligned agents.
Authors: Jingzhe Ni, Xiaolong Yin, Xingyu Lu, Xintong Li, Ji Wei, Ruofeng Tong, Min Tang, Peng Du
Abstract: Computer-Aided Design (CAD) plays a pivotal role in industrial manufacturing but typically requires a high level of expertise from designers. To lower the entry barrier and improve design efficiency, we present an agent for CAD conceptual design powered by large language models (LLMs). The agent accepts both abstract textual descriptions and freehand sketches as input, engaging in interactive dialogue with users to refine and clarify design requirements through comprehensive requirement analysis. Built upon a novel Context-Independent Imperative Paradigm (CIP), the agent generates high-quality CAD modeling code. During the generation process, the agent incorporates iterative visual feedback to improve model quality. Generated design cases are stored in a structured knowledge base, enabling continuous improvement of the agent's code generation capabilities. Experimental results demonstrate that our method achieves state-of-the-art performance in CAD code generation.
Authors: Chengshuai Zhao, Zhen Tan, Pingchuan Ma, Dawei Li, Bohan Jiang, Yancheng Wang, Yingzhen Yang, Huan Liu
Abstract: Chain-of-Thought (CoT) prompting has been shown to improve Large Language Model (LLM) performance on various tasks. With this approach, LLMs appear to produce human-like reasoning steps before providing answers (a.k.a., CoT reasoning), which often leads to the perception that they engage in deliberate inferential processes. However, some initial findings suggest that CoT reasoning may be more superficial than it appears, motivating us to explore further. In this paper, we study CoT reasoning via a data distribution lens and investigate if CoT reasoning reflects a structured inductive bias learned from in-distribution data, allowing the model to conditionally generate reasoning paths that approximate those seen during training. Thus, its effectiveness is fundamentally bounded by the degree of distribution discrepancy between the training data and the test queries. With this lens, we dissect CoT reasoning via three dimensions: task, length, and format. To investigate each dimension, we design DataAlchemy, an isolated and controlled environment to train LLMs from scratch and systematically probe them under various distribution conditions. Our results reveal that CoT reasoning is a brittle mirage that vanishes when it is pushed beyond training distributions. This work offers a deeper understanding of why and when CoT reasoning fails, emphasizing the ongoing challenge of achieving genuine and generalizable reasoning.
Authors: Xianda Zheng, Zijian Huang, Meng-Fen Chiang, Michael J. Witbrock, Kaiqi Zhao
Abstract: Knowledge conflicts commonly arise across diverse sources, and their prevalence has increased with the advent of LLMs. When dealing with conflicts between multiple contexts, also known as \emph{inter-context knowledge conflicts}, LLMs are often confused by lengthy and conflicting contexts. To address this challenge, we propose the Knowledge Conflict Reasoning (KCR) framework, which enhances the ability of LLMs to resolve conflicting knowledge. The key idea of KCR is to train backbone LLMs to establish a correct reasoning process by rewarding them for selecting and adhering to the context with stronger logical consistency when presented with conflicting contexts. Specifically, we first extract reasoning paths, represented by either text or local knowledge graphs, from the conflicting long contexts. Subsequently, we employ Reinforcement Learning to encourage the model to learn the paradigm of reasoning process that follows correct reasoning paths rather than the incorrect counterparts. This enables the backbone models to genuinely acquire the capability to resolve inter-context knowledge conflicts within long contexts. Experimental results demonstrate that our framework significantly improves the ability of various backbone models to resolve knowledge conflicts in long-context scenarios, yielding substantial performance gains.
Authors: Ziji Shi, Le Jiang, Ang Wang, Jie Zhang, Chencan Wu, Yong Li, Xiaokui Xiao, Wei Lin, Jialin Li
Abstract: Tensor parallelism is an essential technique for distributed training of large neural networks. However, automatically determining an optimal tensor parallel strategy is challenging due to the gigantic search space, which grows exponentially with model size and tensor dimension. This prohibits the adoption of auto-parallel systems on larger models. We observe that neural networks usually contain repeated substructures, and build an automatic parallelism framework named TAPAS that eliminates redundant search efforts. TAPAS employs a divide-and-conquer approach that efficiently folds the search space by identifying those unique substructures. As a result, it runs at sub-linear complexity concerning the model size, making it a scalable solution for training large-scale networks. Our evaluations demonstrate that TAPAS outperforms the state-of-the-art automatic parallelism frameworks by up to $160\times$ in search speed on a wide range of models, and the performance of derived strategies is competitive or even better compared with the expert-engineered Megatron-LM library.
Authors: Prateek Mittal, Puneet Goyal, Joohi Chauhan
Abstract: This study introduces a novel multimodal food recognition framework that effectively combines visual and textual modalities to enhance classification accuracy and robustness. The proposed approach employs a dynamic multimodal fusion strategy that adaptively integrates features from unimodal visual inputs and complementary textual metadata. This fusion mechanism is designed to maximize the use of informative content, while mitigating the adverse impact of missing or inconsistent modality data. The framework was rigorously evaluated on the UPMC Food-101 dataset and achieved unimodal classification accuracies of 73.60% for images and 88.84% for text. When both modalities were fused, the model achieved an accuracy of 97.84%, outperforming several state-of-the-art methods. Extensive experimental analysis demonstrated the robustness, adaptability, and computational efficiency of the proposed settings, highlighting its practical applicability to real-world multimodal food-recognition scenarios.
Authors: Dominik Klumpp, Jandson S. Ribeiro
Abstract: Despite significant efforts towards extending the AGM paradigm of belief change beyond finitary logics, the computational aspects of AGM have remained almost untouched. We investigate the computability of AGM contraction on non-finitary logics, and show an intriguing negative result: there are infinitely many uncomputable AGM contraction functions in such logics. Drastically, we also show that the current de facto standard strategies to control computability, which rely on restricting the space of epistemic states, fail: uncomputability remains in all non-finitary cases. Motivated by this disruptive result, we propose new approaches to controlling computability beyond the finitary realm. Using Linear Temporal Logic (LTL) as a case study, we identify an infinite class of fully-rational AGM contraction functions that are computable by design. We use B\"uchi automata to construct such functions, and to represent and reason about LTL beliefs.
Authors: Saketh Ram Kasibatla, Arpan Agarwal, Yuriy Brun, Sorin Lerner, Talia Ringer, Emily First
Abstract: Formal verification using proof assistants, such as Coq, is an effective way of improving software quality, but requires significant effort and expertise. Machine learning can automatically synthesize proofs, but such tools are able to prove only a fraction of desired software properties. We introduce Cobblestone, a divide-and-conquer approach for proof synthesis. Cobblestone uses a large language model (LLM) to generate potential proofs, uses those proofs to break the problem into simpler parts, automatically identifies which of those parts were successfully proven, and iterates on the remaining parts to build a correct proof that is guaranteed to be sound, despite the reliance on unsound LLMs. We evaluate Cobblestone on four benchmarks of open-source Coq projects, controlling for training data leakage. Fully automatically, Cobblestone outperforms state-of-the-art non-LLM tools, and proves many theorems that other LLM-based tools cannot, and on many benchmarks, outperforms them. Each Cobblestone run costs only $1.25 and takes 14.7 minutes, on average. Cobblestone can also be used with external input, from a user or another tool, providing a proof structure or relevant lemmas. Evaluated with such an oracle, Cobblestone proves up to 58% of theorems. Overall, our research shows that tools can make use of partial progress and external input to more effectively automate formal verification.
Authors: Ruofan Wang, Juncheng Li, Yixu Wang, Bo Wang, Xiaosen Wang, Yan Teng, Yingchun Wang, Xingjun Ma, Yu-Gang Jiang
Abstract: As large Vision-Language Models (VLMs) gain prominence, ensuring their safe deployment has become critical. Recent studies have explored VLM robustness against jailbreak attacks-techniques that exploit model vulnerabilities to elicit harmful outputs. However, the limited availability of diverse multimodal data has constrained current approaches to rely heavily on adversarial or manually crafted images derived from harmful text datasets, which often lack effectiveness and diversity across different contexts. In this paper, we propose IDEATOR, a novel jailbreak method that autonomously generates malicious image-text pairs for black-box jailbreak attacks. IDEATOR is grounded in the insight that VLMs themselves could serve as powerful red team models for generating multimodal jailbreak prompts. Specifically, IDEATOR leverages a VLM to create targeted jailbreak texts and pairs them with jailbreak images generated by a state-of-the-art diffusion model. Extensive experiments demonstrate IDEATOR's high effectiveness and transferability, achieving a 94% attack success rate (ASR) in jailbreaking MiniGPT-4 with an average of only 5.34 queries, and high ASRs of 82%, 88%, and 75% when transferred to LLaVA, InstructBLIP, and Chameleon, respectively. Building on IDEATOR's strong transferability and automated process, we introduce the VLJailbreakBench, a safety benchmark comprising 3,654 multimodal jailbreak samples. Our benchmark results on 11 recently released VLMs reveal significant gaps in safety alignment. For instance, our challenge set achieves ASRs of 46.31% on GPT-4o and 19.65% on Claude-3.5-Sonnet, underscoring the urgent need for stronger defenses.
Authors: Yifei Jin, Ali Maatouk, Sarunas Girdzijauskas, Shugong Xu, Leandros Tassiulas, Rex Ying
Abstract: Wireless ray-tracing (RT) is emerging as a key tool for three-dimensional (3D) wireless channel modeling, driven by advances in graphical rendering. Current approaches struggle to accurately model beyond 5G (B5G) network signaling, which often operates at higher frequencies and is more susceptible to environmental conditions and changes. Existing online learning solutions require real-time environmental supervision during training, which is both costly and incompatible with GPU-based processing. In response, we propose a novel approach that redefines ray trajectory generation as a sequential decision-making problem, leveraging generative models to jointly learn the optical, physical, and signal properties within each designated environment. Our work introduces the Scene-Aware Neural Decision Wireless Channel Raytracing Hierarchy (SANDWICH), an innovative offline, fully differentiable approach that can be trained entirely on GPUs. SANDWICH offers superior performance compared to existing online learning methods, outperforms the baseline by 4e^-2 radian in RT accuracy, and only fades 0.5 dB away from toplined channel gain estimation.
Authors: Luca Barsellotti, Lorenzo Bianchi, Nicola Messina, Fabio Carrara, Marcella Cornia, Lorenzo Baraldi, Fabrizio Falchi, Rita Cucchiara
Abstract: Open-Vocabulary Segmentation (OVS) aims at segmenting images from free-form textual concepts without predefined training classes. While existing vision-language models such as CLIP can generate segmentation masks by leveraging coarse spatial information from Vision Transformers, they face challenges in spatial localization due to their global alignment of image and text features. Conversely, self-supervised visual models like DINO excel in fine-grained visual encoding but lack integration with language. To bridge this gap, we present Talk2DINO, a novel hybrid approach that combines the spatial accuracy of DINOv2 with the language understanding of CLIP. Our approach aligns the textual embeddings of CLIP to the patch-level features of DINOv2 through a learned mapping function without the need to fine-tune the underlying backbones. At training time, we exploit the attention maps of DINOv2 to selectively align local visual patches with textual embeddings. We show that the powerful semantic and localization abilities of Talk2DINO can enhance the segmentation process, resulting in more natural and less noisy segmentations, and that our approach can also effectively distinguish foreground objects from the background. Experimental results demonstrate that Talk2DINO achieves state-of-the-art performance across several unsupervised OVS benchmarks. Source code and models are publicly available at: https://lorebianchi98.github.io/Talk2DINO/.
Authors: Pusen Dong, Tianchen Zhu, Yue Qiu, Haoyi Zhou, Jianxin Li
Abstract: Safe reinforcement learning (RL) requires the agent to finish a given task while obeying specific constraints. Giving constraints in natural language form has great potential for practical scenarios due to its flexible transfer capability and accessibility. Previous safe RL methods with natural language constraints typically need to design cost functions manually for each constraint, which requires domain expertise and lacks flexibility. In this paper, we harness the dual role of text in this task, using it not only to provide constraint but also as a training signal. We introduce the Trajectory-level Textual Constraints Translator (TTCT) to replace the manually designed cost function. Our empirical results demonstrate that TTCT effectively comprehends textual constraint and trajectory, and the policies trained by TTCT can achieve a lower violation rate than the standard cost function. Extra studies are conducted to demonstrate that the TTCT has zero-shot transfer capability to adapt to constraint-shift environments.
Authors: Faraz Waseem, Muhammad Shahzad
Abstract: An image may convey a thousand words, but a video composed of hundreds or thousands of image frames tells a more intricate story. Despite significant progress in multimodal large language models (MLLMs), generating extended videos remains a formidable challenge. As of this writing, OpenAI's Sora, the current state-of-the-art system, is still limited to producing videos that are up to one minute in length. This limitation stems from the complexity of long video generation, which requires more than generative AI techniques for approximating density functions essential aspects such as planning, story development, and maintaining spatial and temporal consistency present additional hurdles. Integrating generative AI with a divide-and-conquer approach could improve scalability for longer videos while offering greater control. In this survey, we examine the current landscape of long video generation, covering foundational techniques like GANs and diffusion models, video generation strategies, large-scale training datasets, quality metrics for evaluating long videos, and future research areas to address the limitations of the existing video generation capabilities. We believe it would serve as a comprehensive foundation, offering extensive information to guide future advancements and research in the field of long video generation.
Authors: Jacob Adamczyk, Volodymyr Makarenko, Stas Tiomkin, Rahul V. Kulkarni
Abstract: The average-reward formulation of reinforcement learning (RL) has drawn increased interest in recent years for its ability to solve temporally-extended problems without relying on discounting. Meanwhile, in the discounted setting, algorithms with entropy regularization have been developed, leading to improvements over deterministic methods. Despite the distinct benefits of these approaches, deep RL algorithms for the entropy-regularized average-reward objective have not been developed. While policy-gradient based approaches have recently been presented for the average-reward literature, the corresponding actor-critic framework remains less explored. In this paper, we introduce an average-reward soft actor-critic algorithm to address these gaps in the field. We validate our method by comparing with existing average-reward algorithms on standard RL benchmarks, achieving superior performance for the average-reward criterion.
Authors: Alexis Roger, Prateek Humane, Daniel Z. Kaplan, Kshitij Gupta, Qi Sun, George Adamopoulos, Jonathan Siu Chi Lim, Quentin Anthony, Edwin Fennell, Irina Rish
Abstract: The proliferation of Vision-Language Models (VLMs) in the past several years calls for rigorous and comprehensive evaluation methods and benchmarks. This work analyzes existing VLM evaluation techniques, including automated metrics, AI-based assessments, and human evaluations across diverse tasks. We first introduce Robin - a novel suite of VLMs that we built by combining Large Language Models (LLMs) and Vision Encoders (VEs) at multiple scales, and use Robin to identify shortcomings of current evaluation approaches across scales. Next, to overcome the identified limitations, we introduce CHIRP - a new long form response benchmark we developed for more robust and complete VLM evaluation. We provide open access to the Robin training code, model suite, and CHIRP benchmark to promote reproducibility and advance VLM research.
Authors: Guangjin Pan, Yuan Gao, Yilin Gao, Wenjun Yu, Zhiyong Zhong, Xiaoyu Yang, Xinyu Guo, Shugong Xu
Abstract: Wireless positioning technologies hold significant value for applications in autonomous driving, extended reality (XR), unmanned aerial vehicles (UAVs), and more. With the advancement of artificial intelligence (AI), leveraging AI to enhance positioning accuracy and robustness has emerged as a field full of potential. Driven by the requirements and functionalities defined in the 3rd Generation Partnership Project (3GPP) standards, AI/machine learning (ML)-based cellular positioning is becoming a key technology to overcome the limitations of traditional methods. This paper presents a comprehensive survey of AI-driven cellular positioning. We begin by reviewing the fundamentals of wireless positioning and AI models, analyzing their respective challenges and synergies. We provide a comprehensive review of the evolution of 3GPP positioning standards, with a focus on the integration of AI/ML in current and upcoming standard releases. Guided by the 3GPP-defined taxonomy, we categorize and summarize state-of-the-art (SOTA) research into two major classes: AI/ML-assisted positioning and direct AI/ML-based positioning. The former includes line-of-sight (LOS)/non-line-of-sight (NLOS) detection, time of arrival (TOA)/time difference of arrival (TDOA) estimation, and angle prediction; the latter encompasses fingerprinting, knowledge-assisted learning, and channel charting. Furthermore, we review representative public datasets and conduct performance evaluations of AI-based positioning algorithms using these datasets. Finally, we conclude by summarizing the challenges and opportunities of AI-driven wireless positioning.
Authors: Weihua Zheng, Xin Huang, Zhengyuan Liu, Tarun Kumar Vangani, Bowei Zou, Xiyan Tao, Yuhao Wu, Ai Ti Aw, Nancy F. Chen, Roy Ka-Wei Lee
Abstract: Large language models (LLMs) have shown impressive multilingual capabilities through pretraining on diverse corpora. Although these models show strong reasoning abilities, their performance varies significantly between languages due to the imbalanced distribution of training data. Existing approaches using sample-level translation for extensive multilingual pretraining and cross-lingual tuning face scalability challenges and often fail to capture nuanced reasoning processes across languages. In this paper, we introduce AdaMCOT (Adaptive Multilingual Chain-of-Thought), a framework that enhances multilingual factual reasoning by dynamically routing thought processes in intermediary "thinking languages" before generating target-language responses. AdaMCOT leverages a language-agnostic core and incorporates an adaptive, reward-based mechanism for selecting optimal reasoning pathways without requiring additional pretraining. Our comprehensive evaluation across multiple benchmarks demonstrates substantial improvements in both factual reasoning quality and cross-lingual consistency, with particularly strong performance gains in low-resource language settings. An in-depth analysis of the model's hidden states and semantic space further elucidates the underlying mechanism of our method. The results suggest that adaptive reasoning paths can effectively bridge the performance gap between high and low-resource languages while maintaining cultural and linguistic nuances.
Authors: Wenyun Li, Wenjie Huang, Chen Sun
Abstract: In many real-world scenarios, reward signal for agents are exceedingly sparse, making it challenging to learn an effective reward function for reward shaping. To address this issue, the proposed approach in this paper performs reward shaping not only by utilizing non-zero-reward transitions but also by employing the \emph{Semi-Supervised Learning} (SSL) technique combined with a novel data augmentation to learn trajectory space representations from the majority of transitions, {i.e}., zero-reward transitions, thereby improving the efficacy of reward shaping. Experimental results in Atari and robotic manipulation demonstrate that our method outperforms supervised-based approaches in reward inference, leading to higher agent scores. Notably, in more sparse-reward environments, our method achieves up to twice the peak scores compared to supervised baselines. The proposed double entropy data augmentation enhances performance, showcasing a 15.8\% increase in best score over other augmentation methods
Authors: Yang Zhang, Wenbo Yang, Jun Wang, Qiang Ma, Jie Xiong
Abstract: Accurately forecasting the impact of macroeconomic events is critical for investors and policymakers. Salient events like monetary policy decisions and employment reports often trigger market movements by shaping expectations of economic growth and risk, thereby establishing causal relationships between events and market behavior. Existing forecasting methods typically focus either on textual analysis or time-series modeling, but fail to capture the multi-modal nature of financial markets and the causal relationship between events and price movements. To address these gaps, we propose CAMEF (Causal-Augmented Multi-Modality Event-Driven Financial Forecasting), a multi-modality framework that effectively integrates textual and time-series data with a causal learning mechanism and an LLM-based counterfactual event augmentation technique for causal-enhanced financial forecasting. Our contributions include: (1) a multi-modal framework that captures causal relationships between policy texts and historical price data; (2) a new financial dataset with six types of macroeconomic releases from 2008 to April 2024, and high-frequency real trading data for five key U.S. financial assets; and (3) an LLM-based counterfactual event augmentation strategy. We compare CAMEF to state-of-the-art transformer-based time-series and multi-modal baselines, and perform ablation studies to validate the effectiveness of the causal learning mechanism and event types.
Authors: Riccardo Ali, Francesco Caso, Christopher Irwin, Pietro Li\`o
Abstract: Transformer models map input token sequences to output token distributions, layer by layer. While most interpretability work focuses on internal latent representations, we study the evolution of these token-level distributions directly in vocabulary space. However, such distributions are high-dimensional and defined on an unordered support, making common descriptors like moments or cumulants ill-suited. We address this by computing the Shannon entropy of each intermediate predicted distribution, yielding one interpretable scalar per layer. The resulting sequence, the entropy profile, serves as a compact, information-theoretic signature of the model's computation. We introduce Entropy-Lens, a model-agnostic framework that extracts entropy profiles from frozen, off-the-shelf transformers. We show that these profiles (i) reveal family-specific computation patterns invariant under depth rescaling, (ii) are predictive of prompt type and task format, and (iii) correlate with output correctness. We further show that R\'enyi entropies yield similar results within a broad range of $\alpha$ values, justifying the use of Shannon entropy as a stable and principled summary. Our results hold across different transformers, without requiring gradients, fine-tuning, or access to model internals.
Authors: Yifan Zhong, Xuchuan Huang, Ruochong Li, Ceyao Zhang, Zhang Chen, Tianrui Guan, Fanlian Zeng, Ka Num Lui, Yuyao Ye, Yitao Liang, Yaodong Yang, Yuanpei Chen
Abstract: Dexterous grasping remains a fundamental yet challenging problem in robotics. A general-purpose robot must be capable of grasping diverse objects in arbitrary scenarios. However, existing research typically relies on restrictive assumptions, such as single-object settings or limited environments, showing constrained generalization. We present DexGraspVLA, a hierarchical framework for robust generalization in language-guided general dexterous grasping and beyond. It utilizes a pre-trained Vision-Language model as the high-level planner and learns a diffusion-based low-level Action controller. The key insight to achieve generalization lies in iteratively transforming diverse language and visual inputs into domain-invariant representations via foundation models, where imitation learning can be effectively applied due to the alleviation of domain shift. Notably, our method achieves a 90+% dexterous grasping success rate under thousands of challenging unseen cluttered scenes. Empirical analysis confirms the consistency of internal model behavior across environmental variations, validating our design. DexGraspVLA also, for the first time, simultaneously demonstrates free-form long-horizon prompt execution, robustness to adversarial objects and human disturbance, and failure recovery. Extended application to nonprehensile grasping further proves its generality. Project website: https://dexgraspvla.github.io.
Authors: Abdul Basit, Nouhaila Innan, Muhammad Haider Asif, Minghao Shao, Muhammad Kashif, Alberto Marchisio, Muhammad Shafique
Abstract: Large Language Models (LLMs) offer powerful capabilities in code generation, natural language understanding, and domain-specific reasoning. Their application to quantum software development remains limited, in part because of the lack of high-quality datasets both for LLM training and as dependable knowledge sources. To bridge this gap, we introduce PennyLang, an off-the-shelf, high-quality dataset of 3,347 PennyLane-specific quantum code samples with contextual descriptions, curated from textbooks, official documentation, and open-source repositories. Our contributions are threefold: (1) the creation and open-source release of PennyLang, a purpose-built dataset for quantum programming with PennyLane; (2) a framework for automated quantum code dataset construction that systematizes curation, annotation, and formatting to maximize downstream LLM usability; and (3) a baseline evaluation of the dataset across multiple open-source models, including ablation studies, all conducted within a retrieval-augmented generation (RAG) pipeline. Using PennyLang with RAG substantially improves performance: for example, Qwen 7B's success rate rises from 8.7% without retrieval to 41.7% with full-context augmentation, and LLaMa 4 improves from 78.8% to 84.8%, while also reducing hallucinations and enhancing quantum code correctness. Moving beyond Qiskit-focused studies, we bring LLM-based tools and reproducible methods to PennyLane for advancing AI-assisted quantum development.
Authors: Nathan Drenkow, Mathias Unberath
Abstract: Image quality plays an important role in the performance of deep neural networks (DNNs) that have been widely shown to exhibit sensitivity to changes in imaging conditions. Conventional image quality assessment (IQA) seeks to measure and align quality relative to human perceptual judgments, but we often need a metric that is not only sensitive to imaging conditions but also well-aligned with DNN sensitivities. We first ask whether conventional IQA metrics are also informative of DNN performance. We show theoretically and empirically that conventional IQA metrics are weak predictors of DNN performance for image classification. Using our causal framework, we then develop metrics that exhibit strong correlation with DNN performance, thus enabling us to effectively estimate the quality distribution of large image datasets relative to targeted vision tasks.
Authors: Junwoo Ha, Hyunjun Kim, Sangyoon Yu, Haon Park, Ashkan Yousefpour, Yuna Park, Suhyun Kim
Abstract: We introduce a novel framework for consolidating multi-turn adversarial ``jailbreak'' prompts into single-turn queries, significantly reducing the manual overhead required for adversarial testing of large language models (LLMs). While multi-turn human jailbreaks have been shown to yield high attack success rates, they demand considerable human effort and time. Our multi-turn-to-single-turn (M2S) methods -- Hyphenize, Numberize, and Pythonize -- systematically reformat multi-turn dialogues into structured single-turn prompts. Despite removing iterative back-and-forth interactions, these prompts preserve and often enhance adversarial potency: in extensive evaluations on the Multi-turn Human Jailbreak (MHJ) dataset, M2S methods achieve attack success rates from 70.6 percent to 95.9 percent across several state-of-the-art LLMs. Remarkably, the single-turn prompts outperform the original multi-turn attacks by as much as 17.5 percentage points while cutting token usage by more than half on average. Further analysis shows that embedding malicious requests in enumerated or code-like structures exploits ``contextual blindness'', bypassing both native guardrails and external input-output filters. By converting multi-turn conversations into concise single-turn prompts, the M2S framework provides a scalable tool for large-scale red teaming and reveals critical weaknesses in contemporary LLM defenses.
Authors: Boqian Li, Haiwen Feng, Zeyu Cai, Michael J. Black, Yuliang Xiu
Abstract: Fitting a body to a 3D clothed human point cloud is a common yet challenging task. Traditional optimization-based approaches use multi-stage pipelines that are sensitive to pose initialization, while recent learning-based methods often struggle with generalization across diverse poses and garment types. We propose Equivariant Tightness Fitting for Clothed Humans, or ETCH, a novel pipeline that estimates cloth-to-body surface mapping through locally approximate SE(3) equivariance, encoding tightness as displacement vectors from the cloth surface to the underlying body. Following this mapping, pose-invariant body features regress sparse body markers, simplifying clothed human fitting into an inner-body marker fitting task. Extensive experiments on CAPE and 4D-Dress show that ETCH significantly outperforms state-of-the-art methods -- both tightness-agnostic and tightness-aware -- in body fitting accuracy on loose clothing (16.7% ~ 69.5%) and shape accuracy (average 49.9%). Our equivariant tightness design can even reduce directional errors by (67.2% ~ 89.8%) in one-shot (or out-of-distribution) settings (~ 1% data). Qualitative results demonstrate strong generalization of ETCH, regardless of challenging poses, unseen shapes, loose clothing, and non-rigid dynamics. We will release the code and models soon for research purposes at https://boqian-li.github.io/ETCH/.
Authors: Zhe Wang, Yanjun Qi
Abstract: Gradient optimization-based adversarial attack methods automate the learning of adversarial triggers to generate jailbreak prompts or leak system prompts. In this work, we take a closer look at the optimization objective of adversarial trigger learning and propose ATLA: Adversarial Trigger Learning with Augmented objectives. ATLA improves the negative log-likelihood loss used by previous studies into a weighted loss formulation that encourages the learned adversarial triggers to optimize more towards response format tokens. This enables ATLA to learn an adversarial trigger from just one query-response pair and the learned trigger generalizes well to other similar queries. We further design a variation to augment trigger optimization with an auxiliary loss that suppresses evasive responses. We showcase how to use ATLA to learn adversarial suffixes jailbreaking LLMs and to extract hidden system prompts. Empirically we demonstrate that ATLA consistently outperforms current state-of-the-art techniques, achieving nearly 100% success in attacking while requiring 80% fewer queries. ATLA learned jailbreak suffixes demonstrate high generalization to unseen queries and transfer well to new LLMs. We released our code \href{https://github.com/QData/ALTA_Augmented_Adversarial_Trigger_Learning}{here}.
URLs: https://github.com/QData/ALTA_Augmented_Adversarial_Trigger_Learning
Authors: Mari Ashiga, Wei Jie, Fan Wu, Vardan Voskanyan, Fateme Dinmohammadi, Paul Brookes, Jingzhi Gong, Zheng Wang
Abstract: Generative Pretrained Transformers (GPTs) are foundational Large Language Models (LLMs) for text generation. However, individual LLMs often produce inconsistent outputs and exhibit biases, limiting their representation of diverse language patterns. The closed-source nature of many powerful LLMs further restricts industry applications due to data privacy concerns. Inspired by successes in text generation, LLM ensemble techniques are now increasingly explored for code generation. This article reviews these emerging ensemble approaches to enhance understanding, encourage further research, and promote practical implementation in both text and code generation. We categorize LLM ensembles into seven main methods - weight merging, knowledge fusion, mixture-of-experts, reward ensemble, output ensemble, routing, and cascading - analyzing capabilities of those approaches. Our findings highlight key benefits such as improved diversity representation, enhanced output quality, and greater application flexibility. These insights aid model selection for real-world tasks and crucially, lay groundwork for extending ensemble strategies to multimodal LLMs.
Authors: Liya Guo, Zun Wang, Chang Liu, Junzhe Li, Pipi Hu, Yi Zhu
Abstract: The ensemble average of physical properties of molecules is closely related to the distribution of molecular conformations, and sampling such distributions is a fundamental challenge in physics and chemistry. Traditional methods like molecular dynamics (MD) simulations and Markov chain Monte Carlo (MCMC) sampling are commonly used but can be time-consuming and costly. Recently, diffusion models have emerged as efficient alternatives by learning the distribution of training data. Obtaining an unbiased target distribution is still an expensive task, primarily because it requires satisfying ergodicity. To tackle these challenges, we propose Potential Score Matching (PSM), an approach that utilizes the potential energy gradient to guide generative models. PSM does not require exact energy functions and can debias sample distributions even when trained on limited and biased data. Our method outperforms existing state-of-the-art (SOTA) models on the Lennard-Jones (LJ) potential, a commonly used toy model. Furthermore, we extend the evaluation of PSM to high-dimensional problems using the MD17 and MD22 datasets. The results demonstrate that molecular distributions generated by PSM more closely approximate the Boltzmann distribution compared to traditional diffusion models.
Authors: Chenxi Wang, Jizhan Fang, Xiang Chen, Bozhong Tian, Ziwen Xu, Huajun Chen, Ningyu Zhang
Abstract: Recent advancements in Large Multimodal Models (LMMs) have shown promise in Autonomous Driving Systems (ADS). However, their direct application to ADS is hindered by challenges such as misunderstanding of traffic knowledge, complex road conditions, and diverse states of vehicle. To address these challenges, we propose the use of Knowledge Editing, which enables targeted modifications to a model's behavior without the need for full retraining. Meanwhile, we introduce ADS-Edit, a multimodal knowledge editing dataset specifically designed for ADS, which includes various real-world scenarios, multiple data types, and comprehensive evaluation metrics. We conduct comprehensive experiments and derive several interesting conclusions. We hope that our work will contribute to the further advancement of knowledge editing applications in the field of autonomous driving. Code and data are available in https://github.com/zjunlp/EasyEdit/blob/main/examples/ADSEdit.md.
URLs: https://github.com/zjunlp/EasyEdit/blob/main/examples/ADSEdit.md.
Authors: Jannik Endres, Oliver Hahn, Charles Corbi\`ere, Simone Schaub-Meyer, Stefan Roth, Alexandre Alahi
Abstract: Omnidirectional depth perception is essential for mobile robotics applications that require scene understanding across a full 360{\deg} field of view. Camera-based setups offer a cost-effective option by using stereo depth estimation to generate dense, high-resolution depth maps without relying on expensive active sensing. However, existing omnidirectional stereo matching approaches achieve only limited depth accuracy across diverse environments, depth ranges, and lighting conditions, due to the scarcity of real-world data. We present DFI-OmniStereo, a novel omnidirectional stereo matching method that leverages a large-scale pre-trained foundation model for relative monocular depth estimation within an iterative optimization-based stereo matching architecture. We introduce a dedicated two-stage training strategy to utilize the relative monocular depth features for our omnidirectional stereo matching before scale-invariant fine-tuning. DFI-OmniStereo achieves state-of-the-art results on the real-world Helvipad dataset, reducing disparity MAE by approximately 16% compared to the previous best omnidirectional stereo method.
Authors: Gianluca Peri, Lorenzo Chicchi, Duccio Fanelli, Lorenzo Giambagli
Abstract: Architecture design and optimization are challenging problems in the field of artificial neural networks. Working in this context, we here present SPARCS (SPectral ARchiteCture Search), a novel architecture search protocol which exploits the spectral attributes of the inter-layer transfer matrices. SPARCS allows one to explore the space of possible architectures by spanning continuous and differentiable manifolds, thus enabling for gradient-based optimization algorithms to be eventually employed. With reference to simple benchmark models, we show that the newly proposed method yields a self-emerging architecture with a minimal degree of expressivity to handle the task under investigation and with a reduced parameter count as compared to other viable alternatives.
Authors: Kyungmin Choi, JaKeoung Koo, Stephen McLaughlin, Abderrahim Halimi
Abstract: Single-photon Lidar imaging offers a significant advantage in 3D imaging due to its high resolution and long-range capabilities, however it is challenging to apply in noisy environments with multiple targets per pixel. To tackle these challenges, several methods have been proposed. Statistical methods demonstrate interpretability on the inferred parameters, but they are often limited in their ability to handle complex scenes. Deep learning-based methods have shown superior performance in terms of accuracy and robustness, but they lack interpretability or they are limited to a single-peak per pixel. In this paper, we propose a deep unrolling algorithm for dual-peak single-photon Lidar imaging. We introduce a hierarchical Bayesian model for multiple targets and propose a neural network that unrolls the underlying statistical method. To support multiple targets, we adopt a dual depth maps representation and exploit geometric deep learning to extract features from the point cloud. The proposed method takes advantages of statistical methods and learning-based methods in terms of accuracy and quantifying uncertainty. The experimental results on synthetic and real data demonstrate the competitive performance when compared to existing methods, while also providing uncertainty information.
Authors: Deyuan Liu, Peng Sun, Xufeng Li, Tao Lin
Abstract: Diffusion models excel at generating high-dimensional data but fall short in training efficiency and representation quality compared to self-supervised methods. We identify a key bottleneck: the underutilization of high-quality, semantically rich representations during training notably slows down convergence. Our systematic analysis reveals a critical representation processing region -- primarily in the early layers -- where semantic and structural pattern learning takes place before generation can occur. To address this, we propose Embedded Representation Warmup (ERW), a plug-and-play framework where in the first stage we get the ERW module serves as a warmup that initializes the early layers of the diffusion model with high-quality, pretrained representations. This warmup minimizes the burden of learning representations from scratch, thereby accelerating convergence and boosting performance. Our theoretical analysis demonstrates that ERW's efficacy depends on its precise integration into specific neural network layers -- termed the representation processing region -- where the model primarily processes and transforms feature representations for later generation. We further establish that ERW not only accelerates training convergence but also enhances representation quality: empirically, our method achieves a 40$\times$ acceleration in training speed compared to REPA, the current state-of-the-art methods. Code is available at https://github.com/LINs-lab/ERW.
Authors: Shahriar Noroozizadeh, Jeremy C. Weiss
Abstract: Clinical case reports and discharge summaries may be the most complete and accurate summarization of patient encounters, yet they are finalized, i.e., timestamped after the encounter. Complementary data structured streams become available sooner but suffer from incompleteness. To train models and algorithms on more complete and temporally fine-grained data, we construct a pipeline to phenotype, extract, and annotate time-localized findings within case reports using large language models. We apply our pipeline to generate an open-access textual time series corpus for Sepsis-3 comprising 2,139 case reports from the Pubmed-Open Access (PMOA) Subset. To validate our system, we apply it on PMOA and timeline annotations from I2B2/MIMIC-IV and compare the results to physician-expert annotations. We show high recovery rates of clinical findings (event match rates: O1-preview--0.755, Llama 3.3 70B Instruct--0.753) and strong temporal ordering (concordance: O1-preview--0.932, Llama 3.3 70B Instruct--0.932). Our work characterizes the ability of LLMs to time-localize clinical findings in text, illustrating the limitations of LLM use for temporal reconstruction and providing several potential avenues of improvement via multimodal integration.
Authors: Lawrence Liu, Inesh Chakrabarti, Yixiao Li, Mengdi Wang, Tuo Zhao, Lin F. Yang
Abstract: Large language models (LLMs) exhibit remarkable performance across various natural language processing tasks but suffer from immense computational and memory demands, limiting their deployment in resource-constrained environments. To address this challenge, we propose NoWag: (Normalized Weight and Activation Guided Compression), a unified framework for zero-shot shape preserving compression algorithms. We compressed Llama-2 7B/13B/70B and Llama-3 8/70BB models, using two popular forms of shape-preserving compression, vector quantization NoWag-VQ (NoWag for Vector Quantization), and unstructured/semi-structured pruning NoWag-P (NoWag for Pruning). We found that NoWag-VQ significantly outperforms state-of-the-art zero shot VQ, and that NoWag-P performs competitively against state-of-the-art methods. These results suggest commonalities between these compression paradigms that could inspire future work. Our code is available at https://github.com/LawrenceRLiu/NoWag
Authors: Vansh Gupta, Sankalan Pal Chowdhury, Vil\'em Zouhar, Donya Rooein, Mrinmaya Sachan
Abstract: Large language models (LLMs) are increasingly being adopted in educational settings. These applications expand beyond English, though current LLMs remain primarily English-centric. In this work, we ascertain if their use in education settings in non-English languages is warranted. We evaluated the performance of popular LLMs on four educational tasks: identifying student misconceptions, providing targeted feedback, interactive tutoring, and grading translations in eight languages (Mandarin, Hindi, Arabic, German, Farsi, Telugu, Ukrainian, Czech) in addition to English. We find that the performance on these tasks somewhat corresponds to the amount of language represented in training data, with lower-resource languages having poorer task performance. Although the models perform reasonably well in most languages, the frequent performance drop from English is significant. Thus, we recommend that practitioners first verify that the LLM works well in the target language for their educational task before deployment.
Authors: Yangxu Yin, Honglong Chen, Yudong Gao, Peng Sun, Liantao Wu, Zhe Li, Weifeng Liu
Abstract: Backdoor attacks pose a significant threat to deep neural networks, as backdoored models would misclassify poisoned samples with specific triggers into target classes while maintaining normal performance on clean samples. Among these, multi-target backdoor attacks can simultaneously target multiple classes. However, existing multi-target backdoor attacks all follow the dirty-label paradigm, where poisoned samples are mislabeled, and most of them require an extremely high poisoning rate. This makes them easily detectable by manual inspection. In contrast, clean-label attacks are more stealthy, as they avoid modifying the labels of poisoned samples. However, they generally struggle to achieve stable and satisfactory attack performance and often fail to scale effectively to multi-target attacks. To address this issue, we propose the Feature-based Full-target Clean-label Backdoor Attacks (FFCBA) which consists of two paradigms: Feature-Spanning Backdoor Attacks (FSBA) and Feature-Migrating Backdoor Attacks (FMBA). FSBA leverages class-conditional autoencoders to generate noise triggers that align perturbed in-class samples with the original category's features, ensuring the effectiveness, intra-class consistency, inter-class specificity and natural-feature correlation of triggers. While FSBA supports swift and efficient attacks, its cross-model attack capability is relatively weak. FMBA employs a two-stage class-conditional autoencoder training process that alternates between using out-of-class samples and in-class samples. This allows FMBA to generate triggers with strong target-class features, making it highly effective for cross-model attacks. We conduct experiments on multiple datasets and models, the results show that FFCBA achieves outstanding attack performance and maintains desirable robustness against the state-of-the-art backdoor defenses.
Authors: Kwon Byung-Ki, Qi Dai, Lee Hyoseok, Chong Luo, Tae-Hyun Oh
Abstract: We present JointDiT, a diffusion transformer that models the joint distribution of RGB and depth. By leveraging the architectural benefit and outstanding image prior of the state-of-the-art diffusion transformer, JointDiT not only generates high-fidelity images but also produces geometrically plausible and accurate depth maps. This solid joint distribution modeling is achieved through two simple yet effective techniques that we propose, namely, adaptive scheduling weights, which depend on the noise levels of each modality, and the unbalanced timestep sampling strategy. With these techniques, we train our model across all noise levels for each modality, enabling JointDiT to naturally handle various combinatorial generation tasks, including joint generation, depth estimation, and depth-conditioned image generation by simply controlling the timesteps of each branch. JointDiT demonstrates outstanding joint generation performance. Furthermore, it achieves comparable results in depth estimation and depth-conditioned image generation, suggesting that joint distribution modeling can serve as a viable alternative to conditional generation. The project page is available at https://byungki-k.github.io/JointDiT/.
Authors: Chethan Krishnamurthy Ramanaik, Arjun Roy, Tobias Callies, Eirini Ntoutsi
Abstract: Adversarial robustness of deep autoencoders (AEs) remains relatively unexplored, even though their non-invertible nature poses distinct challenges. Existing attack algorithms during the optimization of imperceptible, norm-bounded adversarial perturbations to maximize output damage in AEs, often stop at sub-optimal attacks. We observe that the adversarial loss gradient vanishes when backpropagated through ill-conditioned layers. This issue arises from near-zero singular values in the Jacobians of these layers, which weaken the gradient signal during optimization. We introduce GRILL, a technique that locally restores gradient signals in ill-conditioned layers, enabling more effective norm-bounded attacks. Through extensive experiments on different architectures of popular AEs, under both sample-specific and universal attack setups, and across standard and adaptive attack settings, we show that our method significantly increases the effectiveness of our adversarial attacks, enabling a more rigorous evaluation of AE robustness.
Authors: Yi Zhang, Nikolaos Farmakidis, Ioannis Roumpos, Miltiadis Moralis-Pegios, Apostolos Tsakyridis, June Sang Lee, Bowei Dong, Yuhan He, Samarth Aggarwal, Nikolaos Pleros, Harish Bhaskaran
Abstract: Optical computing systems deliver unrivalled processing speeds for scalar operations. Yet, integrated implementations have been constrained to low-dimensional tensor operations that fall short of the vector dimensions required for modern artificial intelligence. We demonstrate an all-optical neuromorphic computing system based on time division multiplexing, capable of processing input vectors exceeding 250,000 elements within a unified framework. The platform harnesses optically driven thermo-optic modulation in standing wave optical fields, with titanium nano-antennas functioning as wavelength-selective absorbers. Counterintuitively, the thermal time dynamics of the system enable simultaneous time integration of ultra-fast (50GHz) signals and the application of programmable, non-linear activation functions, entirely within the optical domain. This unified framework constitutes a leap towards large-scale photonic computing that satisfies the dimensional requirements of AI workloads.
Authors: Benjamin Raphael Ernhofer, Daniil Prokhorov, Jannica Langner, Dominik Bollmann
Abstract: Modern automotive infotainment systems necessitate intelligent and adaptive solutions to manage frequent User Interface (UI) updates and diverse design variations. This work introduces a vision-language framework to facilitate the understanding of and interaction with automotive UIs, enabling seamless adaptation across different UI designs. To support research in this field, AutomotiveUI-Bench-4K, an open-source dataset comprising 998 images with 4,208 annotations, is also released. Additionally, a data pipeline for generating training data is presented. A Molmo-7B-based model is fine-tuned using Low-Rank Adaptation (LoRa), incorporating generated reasoning along with visual grounding and evaluation capabilities. The fine-tuned Evaluative Large Action Model (ELAM) achieves strong performance on AutomotiveUI-Bench-4K (model and dataset are available on Hugging Face). The approach demonstrates strong cross-domain generalization, including a +5.6% improvement on ScreenSpot over the baseline model. An average accuracy of 80.8% is achieved on ScreenSpot, closely matching or surpassing specialized models for desktop, mobile, and web, despite being trained primarily on the automotive domain. This research investigates how data collection and subsequent fine-tuning can lead to AI-driven advancements in automotive UI understanding and interaction. The applied method is cost-efficient, and fine-tuned models can be deployed on consumer-grade GPUs.
Authors: Qingmei Li, Yang Zhang, Zurong Mai, Yuhang Chen, Shuohong Lou, Henglian Huang, Jiarui Zhang, Zhiwei Zhang, Yibin Wen, Weijia Li, Haohuan Fu, Jianxi Huang, Juepeng Zheng
Abstract: Large Multimodal Models (LMMs) has demonstrated capabilities across various domains, but comprehensive benchmarks for agricultural remote sensing (RS) remain scarce. Existing benchmarks designed for agricultural RS scenarios exhibit notable limitations, primarily in terms of insufficient scene diversity in the dataset and oversimplified task design. To bridge this gap, we introduce AgroMind, a comprehensive agricultural remote sensing benchmark covering four task dimensions: spatial perception, object understanding, scene understanding, and scene reasoning, with a total of 13 task types, ranging from crop identification and health monitoring to environmental analysis. We curate a high-quality evaluation set by integrating eight public datasets and one private farmland plot dataset, containing 27,247 QA pairs and 19,615 images. The pipeline begins with multi-source data pre-processing, including collection, format standardization, and annotation refinement. We then generate a diverse set of agriculturally relevant questions through the systematic definition of tasks. Finally, we employ LMMs for inference, generating responses, and performing detailed examinations. We evaluated 20 open-source LMMs and 4 closed-source models on AgroMind. Experiments reveal significant performance gaps, particularly in spatial reasoning and fine-grained recognition, it is notable that human performance lags behind several leading LMMs. By establishing a standardized evaluation framework for agricultural RS, AgroMind reveals the limitations of LMMs in domain knowledge and highlights critical challenges for future work. Data and code can be accessed at https://rssysu.github.io/AgroMind/.
Authors: Guangyuan Ma, Yongliang Ma, Xuanrui Gou, Zhenpeng Su, Ming Zhou, Songlin Hu
Abstract: Large Language Models (LLMs)-based text retrieval retrieves documents relevant to search queries based on vector similarities. Documents are pre-encoded offline, while queries arrive in real-time, necessitating an efficient online query encoder. Although LLMs significantly enhance retrieval capabilities, serving deeply parameterized LLMs slows down query inference throughput and increases demands for online deployment resources. In this paper, we propose LightRetriever, a novel LLM-based retriever with extremely lightweight query encoders. Our method retains a full-sized LLM for document encoding, but reduces the workload of query encoding to no more than an embedding lookup. Compared to serving a full LLM on an A800 GPU, our method achieves over 1000x speedup in query encoding and over 10x increase in end-to-end retrieval throughput. Extensive experiments on large-scale retrieval benchmarks show that LightRetriever generalizes well across diverse tasks, maintaining an average of 95% retrieval performance.
Authors: Yutong Chen, Jiandong Gao, Ji Wu
Abstract: R1-style Reinforcement Learning (RL) significantly enhances Large Language Models' reasoning capabilities, yet the mechanism behind rule-based RL remains unclear. We found that small-scale SFT has substantial influence on RL but shows poor efficiency. To explain our observations, we propose an analytical framework and compare the efficiency of SFT and RL by measuring \textbf{sample effect}. Our hypothetical analysis shows the potential to improve SFT efficiency. Guided by our analysis, we propose \textbf{Re-distillation}, a technique that aims to boost the effectiveness of small-scale distillation by sampling from the RL-trained policy. Re-distillation shows consistent surprising efficiency on three datasets and both Qwen\&Llama models: Re-distilled models matched RL performance with far fewer samples and less computation. As a result, on K\&K dataset, our re-distilled Qwen-2.5-1.5B model surpasses DeepSeek-V3-0324 with only 1K SFT samples. We demonstrate that re-distillation can be used to efficiently balance multiple goals in RL. Our work explains several interesting phenomena in R1-style RL, shedding light on the mechanisms behind its empirical success. Code is available at: https://github.com/on1262/deep-reasoning.
Authors: Kunal Sawarkar, Shivam R. Solanki, Abhilasha Mangal
Abstract: Retrieval-Augmented Generation (RAG) struggles with domain-specific enterprise datasets, often isolated behind firewalls and rich in complex, specialized terminology unseen by LLMs during pre-training. Semantic variability across domains like medicine, networking, or law hampers RAG's context precision, while fine-tuning solutions are costly, slow, and lack generalization as new data emerges. Achieving zero-shot precision with retrievers without fine-tuning still remains a key challenge. We introduce 'MetaGen Blended RAG', a novel enterprise search approach that enhances semantic retrievers through a metadata generation pipeline and hybrid query indexes using dense and sparse vectors. By leveraging key concepts, topics, and acronyms, our method creates metadata-enriched semantic indexes and boosted hybrid queries, delivering robust, scalable performance without fine-tuning. On the biomedical PubMedQA dataset, MetaGen Blended RAG achieves 82% retrieval accuracy and 77% RAG accuracy, surpassing all prior zero-shot RAG benchmarks and even rivaling fine-tuned models on that dataset, while also excelling on datasets like SQuAD and NQ. This approach redefines enterprise search using a new approach to building semantic retrievers with unmatched generalization across specialized domains.
Authors: Zeng Wang, Minghao Shao, Rupesh Karn, Likhitha Mankali, Jitendra Bhandari, Ramesh Karri, Ozgur Sinanoglu, Muhammad Shafique, Johann Knechtel
Abstract: Large Language Models (LLMs) offer transformative capabilities for hardware design automation, particularly in Verilog code generation. However, they also pose significant data security challenges, including Verilog evaluation data contamination, intellectual property (IP) design leakage, and the risk of malicious Verilog generation. We introduce SALAD, a comprehensive assessment that leverages machine unlearning to mitigate these threats. Our approach enables the selective removal of contaminated benchmarks, sensitive IP and design artifacts, or malicious code patterns from pre-trained LLMs, all without requiring full retraining. Through detailed case studies, we demonstrate how machine unlearning techniques effectively reduce data security risks in LLM-aided hardware design.
Authors: Deepak Pandita, Tharindu Cyril Weerasooriya, Ankit Parag Shah, Isabelle Diana May-Xin Ng, Christopher M. Homan, Wei Wei
Abstract: Agentic workflows, where multiple AI agents collaborate to accomplish complex tasks like reasoning or planning, play a substantial role in many cutting-edge commercial applications, and continue to fascinate researchers across nearly all fields for their potential to accomplish expensive, complex tasks that, until recently, only humans have been trusted to do. These workflows depend critically on the prompts used to provide the roles models play in such workflows. Poorly designed prompts that fail even slightly to guide individual agents can lead to sub-optimal performance that may snowball within a system of agents, limiting their reliability and scalability. To address this important problem of inference-time prompt optimization, we introduce ProRefine, an innovative inference-time optimization method that uses an agentic loop of LLMs to generate and apply textual feedback. ProRefine dynamically refines prompts for multi-step reasoning tasks without additional training or ground truth labels. Evaluated on five benchmark mathematical reasoning datasets, ProRefine significantly surpasses zero-shot Chain-of-Thought baselines by 3 to 37 percentage points. This approach not only boosts accuracy but also allows smaller models to approach the performance of their larger counterparts. This highlights its potential for building more cost-effective and powerful hybrid AI systems, thereby democratizing access to high-performing AI.
Authors: Kangwei Liu, Siyuan Cheng, Bozhong Tian, Xiaozhuan Liang, Yuyang Yin, Meng Han, Ningyu Zhang, Bryan Hooi, Xi Chen, Shumin Deng
Abstract: Large language models (LLMs) have been increasingly applied to automated harmful content detection tasks, assisting moderators in identifying policy violations and improving the overall efficiency and accuracy of content review. However, existing resources for harmful content detection are predominantly focused on English, with Chinese datasets remaining scarce and often limited in scope. We present a comprehensive, professionally annotated benchmark for Chinese content harm detection, which covers six representative categories and is constructed entirely from real-world data. Our annotation process further yields a knowledge rule base that provides explicit expert knowledge to assist LLMs in Chinese harmful content detection. In addition, we propose a knowledge-augmented baseline that integrates both human-annotated knowledge rules and implicit knowledge from large language models, enabling smaller models to achieve performance comparable to state-of-the-art LLMs. Code and data are available at https://github.com/zjunlp/ChineseHarm-bench.
Authors: Xiaoran Fan, Zhichao Sun, Yangfan Gao, Jingfei Xiong, Hang Yan, Yifei Cao, Jiajun Sun, Shuo Li, Zhihao Zhang, Zhiheng Xi, Yuhao Zhou, Senjie Jin, Changhao Jiang, Junjie Ye, Ming Zhang, Rui Zheng, Zhenhua Han, Yunke Zhang, Demei Yan, Shaokang Dong, Tao Ji, Tao Gui, Qi Zhang, Xuanjing Huang
Abstract: Speech-language models (SLMs) offer a promising path toward unifying speech and text understanding and generation. However, challenges remain in achieving effective cross-modal alignment and high-quality speech generation. In this work, we systematically investigate the role of speech tokenizer designs in LLM-centric SLMs, augmented by speech heads and speaker modeling. We compare coupled, semi-decoupled, and fully decoupled speech tokenizers under a fair SLM framework and find that decoupled tokenization significantly improves alignment and synthesis quality. To address the information density mismatch between speech and text, we introduce multi-token prediction (MTP) into SLMs, enabling each hidden state to decode multiple speech tokens. This leads to up to 12$\times$ faster decoding and a substantial drop in word error rate (from 6.07 to 3.01). Furthermore, we propose a speaker-aware generation paradigm and introduce RoleTriviaQA, a large-scale role-playing knowledge QA benchmark with diverse speaker identities. Experiments demonstrate that our methods enhance both knowledge understanding and speaker consistency.
Authors: Nikos Spyrou, Athanasios Vlontzos, Paraskevas Pegios, Thomas Melistas, Nefeli Gkouti, Yannis Panagakis, Giorgos Papanastasiou, Sotirios A. Tsaftaris
Abstract: Adapting text-to-image (T2I) latent diffusion models (LDMs) to video editing has shown strong visual fidelity and controllability, but challenges remain in maintaining causal relationships inherent to the video data generating process. Edits affecting causally dependent attributes often generate unrealistic or misleading outcomes if these relationships are ignored. In this work, we introduce a causally faithful framework for counterfactual video generation, formulated as an Out-of-Distribution (OOD) prediction problem. We embed prior causal knowledge by encoding the relationships specified in a causal graph into text prompts and guide the generation process by optimizing these prompts using a vision-language model (VLM)-based textual loss. This loss encourages the latent space of the LDMs to capture OOD variations in the form of counterfactuals, effectively steering generation toward causally meaningful alternatives. The proposed framework, dubbed CSVC, is agnostic to the underlying video editing system and does not require access to its internal mechanisms or fine-tuning. We evaluate our approach using standard video quality metrics and counterfactual-specific criteria, such as causal effectiveness and minimality. Experimental results show that CSVC generates causally faithful video counterfactuals within the LDM distribution via prompt-based causal steering, achieving state-of-the-art causal effectiveness without compromising temporal consistency or visual quality on real-world facial videos. Due to its compatibility with any black-box video editing system, our framework has significant potential to generate realistic 'what if' hypothetical video scenarios in diverse areas such as digital media and healthcare.
Authors: Sunil Kumar, Bowen Zhao, Leo Dirac, Paulina Varshavskaya
Abstract: Despite tremendous recent advances in large model reasoning ability, vision-language models (VLMs) still struggle with detailed visual reasoning, especially when compute resources are limited. To address this challenge, we draw inspiration from methods like Deepseek-r1 for VLMs and train smaller-scale models with Group Relative Policy Optimization (GRPO) to use external tools such as zoom. The greatest benefit is obtained with a combination of GRPO learning, a simple reward structure, a simplified tool-calling interface, allocating additional tokens to the result of the tool call, and a training data mix that over-represents visually difficult examples. Compared to similarly-sized baseline models, our method achieves better performance on some visual question-answering (VQA) tasks, thanks to the detailed visual information gathered from the external tool.
Authors: Xiuyu Yang, Shuhan Tan, Philipp Kr\"ahenb\"uhl
Abstract: An ideal traffic simulator replicates the realistic long-term point-to-point trip that a self-driving system experiences during deployment. Prior models and benchmarks focus on closed-loop motion simulation for initial agents in a scene. This is problematic for long-term simulation. Agents enter and exit the scene as the ego vehicle enters new regions. We propose InfGen, a unified next-token prediction model that performs interleaved closed-loop motion simulation and scene generation. InfGen automatically switches between closed-loop motion simulation and scene generation mode. It enables stable long-term rollout simulation. InfGen performs at the state-of-the-art in short-term (9s) traffic simulation, and significantly outperforms all other methods in long-term (30s) simulation. The code and model of InfGen will be released at https://orangesodahub.github.io/InfGen
Authors: Yuqi Zhu, Yi Zhong, Jintian Zhang, Ziheng Zhang, Shuofei Qiao, Yujie Luo, Lun Du, Da Zheng, Ningyu Zhang, Huajun Chen
Abstract: Large Language Models (LLMs) hold promise in automating data analysis tasks, yet open-source models face significant limitations in these kinds of reasoning-intensive scenarios. In this work, we investigate strategies to enhance the data analysis capabilities of open-source LLMs. By curating a seed dataset of diverse, realistic scenarios, we evaluate model behavior across three core dimensions: data understanding, code generation, and strategic planning. Our analysis reveals three key findings: (1) Strategic planning quality serves as the primary determinant of model performance; (2) Interaction design and task complexity significantly influence reasoning capabilities; (3) Data quality demonstrates a greater impact than diversity in achieving optimal performance. We leverage these insights to develop a data synthesis methodology, demonstrating significant improvements in open-source LLMs' analytical reasoning capabilities. Code is available at https://github.com/zjunlp/DataMind.
Authors: Qiguang Chen, Mingda Yang, Libo Qin, Jinhao Liu, Zheng Yan, Jiannan Guan, Dengyun Peng, Yiyan Ji, Hanjing Li, Mengkang Hu, Yimeng Zhang, Yihao Liang, Yuhang Zhou, Jiaqi Wang, Zhi Chen, Wanxiang Che
Abstract: Recent advancements in artificial intelligence (AI), particularly in large language models (LLMs) such as OpenAI-o1 and DeepSeek-R1, have demonstrated remarkable capabilities in complex domains such as logical reasoning and experimental coding. Motivated by these advancements, numerous studies have explored the application of AI in the innovation process, particularly in the context of scientific research. These AI technologies primarily aim to develop systems that can autonomously conduct research processes across a wide range of scientific disciplines. Despite these significant strides, a comprehensive survey on AI for Research (AI4Research) remains absent, which hampers our understanding and impedes further development in this field. To address this gap, we present a comprehensive survey and offer a unified perspective on AI4Research. Specifically, the main contributions of our work are as follows: (1) Systematic taxonomy: We first introduce a systematic taxonomy to classify five mainstream tasks in AI4Research. (2) New frontiers: Then, we identify key research gaps and highlight promising future directions, focusing on the rigor and scalability of automated experiments, as well as the societal impact. (3) Abundant applications and resources: Finally, we compile a wealth of resources, including relevant multidisciplinary applications, data corpora, and tools. We hope our work will provide the research community with quick access to these resources and stimulate innovative breakthroughs in AI4Research.
Authors: Zihan Tan, Suyuan Huang, Guancheng Wan, Wenke Huang, He Li, Mang Ye
Abstract: Federated Graph Learning (FGL) combines the privacy-preserving capabilities of federated learning (FL) with the strong graph modeling capability of Graph Neural Networks (GNNs). Current research addresses subgraph-FL from the structural perspective, neglecting the propagation of graph signals on spatial and spectral domains of the structure. From a spatial perspective, subgraph-FL introduces edge disconnections between clients, leading to disruptions in label signals and a degradation in the semantic knowledge of the global GNN. From a spectral perspective, spectral heterogeneity causes inconsistencies in signal frequencies across subgraphs, which makes local GNNs overfit the local signal propagation schemes. As a result, spectral client drift occurs, undermining global generalizability. To tackle the challenges, we propose a global knowledge repository to mitigate the challenge of poor semantic knowledge caused by label signal disruption. Furthermore, we design a frequency alignment to address spectral client drift. The combination of Spatial and Spectral strategies forms our framework S2FGL. Extensive experiments on multiple datasets demonstrate the superiority of S2FGL. The code is available at https://github.com/Wonder7racer/S2FGL.git.
Authors: Tek Raj Chhetri, Yibei Chen, Puja Trivedi, Dorota Jarecka, Saif Haobsh, Patrick Ray, Lydia Ng, Satrajit S. Ghosh
Abstract: The ability to extract structured information from unstructured sources-such as free-text documents and scientific literature-is critical for accelerating scientific discovery and knowledge synthesis. Large Language Models (LLMs) have demonstrated remarkable capabilities in various natural language processing tasks, including structured information extraction. However, their effectiveness often diminishes in specialized, domain-specific contexts that require nuanced understanding and expert-level domain knowledge. In addition, existing LLM-based approaches frequently exhibit poor transferability across tasks and domains, limiting their scalability and adaptability. To address these challenges, we introduce StructSense, a modular, task-agnostic, open-source framework for structured information extraction built on LLMs. StructSense is guided by domain-specific symbolic knowledge encoded in ontologies, enabling it to navigate complex domain content more effectively. It further incorporates agentic capabilities through self-evaluative judges that form a feedback loop for iterative refinement, and includes human-in-the-loop mechanisms to ensure quality and validation. We demonstrate that StructSense can overcome both the limitations of domain sensitivity and the lack of cross-task generalizability, as shown through its application to diverse neuroscience information extraction tasks.
Authors: Janna Lu
Abstract: Large language models (LLMs) have demonstrated remarkable capabilities across diverse tasks, but their ability to forecast future events remains understudied. A year ago, large language models struggle to come close to the accuracy of a human crowd. I evaluate state-of-the-art LLMs on 464 forecasting questions from Metaculus, comparing their performance against top forecasters. Frontier models achieve Brier scores that ostensibly surpass the human crowd but still significantly underperform a group of experts.
Authors: Wenxuan Zhou, Shujian Zhang, Brice Magdalou, John Lambert, Ehsan Amid, Richard Nock, Andrew Hard
Abstract: In this paper, we show that direct preference optimization (DPO) is a very specific form of a connection between two major theories in the ML context of learning from preferences: loss functions (Savage) and stochastic choice (Doignon-Falmagne and Machina). The connection is established for all of Savage's losses and at this level of generality, (i) it includes support for abstention on the choice theory side, (ii) it includes support for non-convex objectives on the ML side, and (iii) it allows to frame for free some notable extensions of the DPO setting, including margins and corrections for length. Getting to understand how DPO operates from a general principled perspective is crucial because of the huge and diverse application landscape of models, because of the current momentum around DPO, but also -- and importantly -- because many state of the art variations on DPO definitely occupy a small region of the map that we cover. It also helps to understand the pitfalls of departing from this map, and figure out workarounds.
Authors: Mingqi Wu, Zhihao Zhang, Qiaole Dong, Zhiheng Xi, Jun Zhao, Senjie Jin, Xiaoran Fan, Yuhao Zhou, Huijie Lv, Ming Zhang, Yanwei Fu, Qin Liu, Songyang Zhang, Qi Zhang
Abstract: Reasoning in large language models has long been a central research focus, and recent studies employing reinforcement learning (RL) have introduced diverse methods that yield substantial performance gains with minimal or even no external supervision. Surprisingly, some studies even suggest that random or incorrect reward signals can enhance performance. However, these breakthroughs are predominantly observed for the mathematically strong Qwen2.5 series on benchmarks such as MATH-500, AMC, and AIME, and seldom transfer to models like Llama, which warrants a more in-depth investigation. In this work, our empirical analysis reveals that pre-training on massive web-scale corpora leaves Qwen2.5 susceptible to data contamination in widely used benchmarks. Consequently, conclusions derived from contaminated benchmarks on Qwen2.5 series may be unreliable. To obtain trustworthy evaluation results, we introduce a generator that creates fully clean arithmetic problems of arbitrary length and difficulty, dubbed RandomCalculation. Using this leakage-free dataset, we show that only accurate reward signals yield steady improvements that surpass the base model's performance boundary in mathematical reasoning, whereas random or incorrect rewards do not. Moreover, we conduct more fine-grained analyses to elucidate the factors underlying the different performance observed on the MATH-500 and RandomCalculation benchmarks. Consequently, we recommend that future studies evaluate models on uncontaminated benchmarks and, when feasible, test various model series to ensure trustworthy conclusions about RL and related methods.
Authors: Daniil Orel, Indraneil Paul, Iryna Gurevych, Preslav Nakov
Abstract: In this work, we compile $\textbf{$\texttt{DroidCollection}$}$, the most extensive open data suite for training and evaluating machine-generated code detectors, comprising over a million code samples, seven programming languages, outputs from 43 coding models, and over three real-world coding domains. Alongside fully AI-generated samples, our collection includes human-AI co-authored code, as well as adversarial samples explicitly crafted to evade detection. Subsequently, we develop $\textbf{$\texttt{DroidDetect}$}$, a suite of encoder-only detectors trained using a multi-task objective over $\texttt{DroidCollection}$. Our experiments show that existing detectors' performance fails to generalise to diverse coding domains and programming languages outside of their narrow training data. Additionally, we demonstrate that while most detectors are easily compromised by humanising the output distributions using superficial prompting and alignment approaches, this problem can be easily amended by training on a small amount of adversarial data. Finally, we demonstrate the effectiveness of metric learning and uncertainty-based resampling as means to enhance detector training on possibly noisy distributions.
Authors: Yaowen Hu, Wenxuan Tu, Yue Liu, Miaomiao Li, Wenpeng Lu, Zhigang Luo, Xinwang Liu, Ping Chen
Abstract: Deep graph clustering (DGC) for attribute-missing graphs is an unsupervised task aimed at partitioning nodes with incomplete attributes into distinct clusters. Addressing this challenging issue is vital for practical applications. However, research in this area remains underexplored. Existing imputation methods for attribute-missing graphs often fail to account for the varying amounts of information available across node neighborhoods, leading to unreliable results, especially for nodes with insufficient known neighborhood. To address this issue, we propose a novel method named Divide-Then-Rule Graph Completion (DTRGC). This method first addresses nodes with sufficient known neighborhood information and treats the imputed results as new knowledge to iteratively impute more challenging nodes, while leveraging clustering information to correct imputation errors. Specifically, Dynamic Cluster-Aware Feature Propagation (DCFP) initializes missing node attributes by adjusting propagation weights based on the clustering structure. Subsequently, Hierarchical Neighborhood-aware Imputation (HNAI) categorizes attribute-missing nodes into three groups based on the completeness of their neighborhood attributes. The imputation is performed hierarchically, prioritizing the groups with nodes that have the most available neighborhood information. The cluster structure is then used to refine the imputation and correct potential errors. Finally, Hop-wise Representation Enhancement (HRE) integrates information across multiple hops, thereby enriching the expressiveness of node representations. Experimental results on six widely used graph datasets show that DTRGC significantly improves the clustering performance of various DGC methods under attribute-missing graphs.
Authors: Yuchi Tang, I\~naki Esnaola, George Panoutsos
Abstract: Existing post-hoc model-agnostic methods generate external explanations for opaque models, primarily by locally attributing the model output to its input features. However, they often lack an explicit and systematic framework for quantifying the contribution of individual features. Building on the Taylor expansion framework introduced by Deng et al. (2024) to unify existing local attribution methods, we propose a rigorous set of postulates -- "precision", "federation", and "zero-discrepancy" -- to govern Taylor term-specific attribution. Guided by these postulates, we introduce TaylorPODA (Taylor expansion-derived imPortance-Order aDapted Attribution), which incorporates an additional "adaptation" property. This property enables alignment with task-specific goals, especially in post-hoc settings lacking ground-truth explanations. Empirical evaluations demonstrate that TaylorPODA achieves competitive results against baseline methods, providing principled and visualization-friendly explanations. This work enhances the trustworthy deployment of opaque models by offering explanations with stronger theoretical grounding.
Authors: Yaowen Hu, Wenxuan Tu, Yue Liu, Xinhang Wan, Junyi Yan, Taichun Zhou, Xinwang Liu
Abstract: Deep graph clustering (DGC), which aims to unsupervisedly separate the nodes in an attribute graph into different clusters, has seen substantial potential in various industrial scenarios like community detection and recommendation. However, the real-world attribute graphs, e.g., social networks interactions, are usually large-scale and attribute-missing. To solve these two problems, we propose a novel DGC method termed \underline{\textbf{C}}omplementary \underline{\textbf{M}}ulti-\underline{\textbf{V}}iew \underline{\textbf{N}}eighborhood \underline{\textbf{D}}ifferentiation (\textit{CMV-ND}), which preprocesses graph structural information into multiple views in a complete but non-redundant manner. First, to ensure completeness of the structural information, we propose a recursive neighborhood search that recursively explores the local structure of the graph by completely expanding node neighborhoods across different hop distances. Second, to eliminate the redundancy between neighborhoods at different hops, we introduce a neighborhood differential strategy that ensures no overlapping nodes between the differential hop representations. Then, we construct $K+1$ complementary views from the $K$ differential hop representations and the features of the target node. Last, we apply existing multi-view clustering or DGC methods to the views. Experimental results on six widely used graph datasets demonstrate that CMV-ND significantly improves the performance of various methods.
Authors: Goeric Huybrechts, Srikanth Ronanki, Sai Muralidhar Jayanthi, Jack Fitzgerald, Srinivasan Veeravanallur
Abstract: The proliferation of multimodal Large Language Models has significantly advanced the ability to analyze and understand complex data inputs from different modalities. However, the processing of long documents remains under-explored, largely due to a lack of suitable benchmarks. To address this, we introduce Document Haystack, a comprehensive benchmark designed to evaluate the performance of Vision Language Models (VLMs) on long, visually complex documents. Document Haystack features documents ranging from 5 to 200 pages and strategically inserts pure text or multimodal text+image "needles" at various depths within the documents to challenge VLMs' retrieval capabilities. Comprising 400 document variants and a total of 8,250 questions, it is supported by an objective, automated evaluation framework. We detail the construction and characteristics of the Document Haystack dataset, present results from prominent VLMs and discuss potential research avenues in this area.
Authors: Wei Fan, JinYi Yoon, Xiaochang Li, Huajie Shao, Bo Ji
Abstract: Split Learning (SL) is an emerging privacy-preserving machine learning technique that enables resource constrained edge devices to participate in model training by partitioning a model into client-side and server-side sub-models. While SL reduces computational overhead on edge devices, it encounters significant challenges in heterogeneous environments where devices vary in computing resources, communication capabilities, environmental conditions, and privacy requirements. Although recent studies have explored heterogeneous SL frameworks that optimize split points for devices with varying resource constraints, they often neglect personalized privacy requirements and local model customization under varying environmental conditions. To address these limitations, we propose P3SL, a Personalized Privacy-Preserving Split Learning framework designed for heterogeneous, resource-constrained edge device systems. The key contributions of this work are twofold. First, we design a personalized sequential split learning pipeline that allows each client to achieve customized privacy protection and maintain personalized local models tailored to their computational resources, environmental conditions, and privacy needs. Second, we adopt a bi-level optimization technique that empowers clients to determine their own optimal personalized split points without sharing private sensitive information (i.e., computational resources, environmental conditions, privacy requirements) with the server. This approach balances energy consumption and privacy leakage risks while maintaining high model accuracy. We implement and evaluate P3SL on a testbed consisting of 7 devices including 4 Jetson Nano P3450 devices, 2 Raspberry Pis, and 1 laptop, using diverse model architectures and datasets under varying environmental conditions.
Authors: Zhuokun Chen, Zeren Chen, Jiahao He, Mingkui Tan, Jianfei Cai, Bohan Zhuang
Abstract: Chain-of-thought (CoT) reasoning enhances the problem-solving capabilities of large language models by encouraging step-by-step intermediate reasoning during inference. While effective, CoT introduces substantial computational overhead due to its reliance on autoregressive decoding over long token sequences. Existing acceleration strategies either reduce sequence length through early stopping or compressive reward designs, or improve decoding speed via speculative decoding with smaller models. However, speculative decoding suffers from limited speedup when the agreement between small and large models is low, and fails to exploit the potential advantages of small models in producing concise intermediate reasoning. In this paper, we present R-Stitch, a token-level, confidence-based hybrid decoding framework that accelerates CoT inference by switching between a small language model (SLM) and a large language model (LLM) along the reasoning trajectory. R-Stitch uses the SLM to generate tokens by default and delegates to the LLM only when the SLM's confidence falls below a threshold. This design avoids full-sequence rollback and selectively invokes the LLM on uncertain steps, preserving both efficiency and answer quality. R-Stitch is model-agnostic, training-free, and compatible with standard decoding pipelines. Experiments on math reasoning benchmarks demonstrate that R-Stitch achieves up to 85\% reduction in inference latency with negligible accuracy drop, highlighting its practical effectiveness in accelerating CoT reasoning.
Authors: Melih Barsbey, Lucas Prieto, Stefanos Zafeiriou, Tolga Birdal
Abstract: Robustness and resource-efficiency are two highly desirable properties for modern machine learning models. However, achieving them jointly remains a challenge. In this paper, we identify high learning rates as a facilitator for simultaneously achieving robustness to spurious correlations and network compressibility. We demonstrate that large learning rates also produce desirable representation properties such as invariant feature utilization, class separation, and activation sparsity. Our findings indicate that large learning rates compare favorably to other hyperparameters and regularization methods, in consistently satisfying these properties in tandem. In addition to demonstrating the positive effect of large learning rates across diverse spurious correlation datasets, models, and optimizers, we also present strong evidence that the previously documented success of large learning rates in standard classification tasks is related to addressing hidden/rare spurious correlations in the training dataset. Our investigation of the mechanisms underlying this phenomenon reveals the importance of confident mispredictions of bias-conflicting samples under large learning rates.
Authors: Zhongzheng Yuan, Lianshuai Guo, Xunkai Li, Yinlin Zhu, Wenyu Wang, Meixia Qu
Abstract: Federated Graph Learning (FGL) is a distributed learning paradigm that enables collaborative training over large-scale subgraphs located on multiple local systems. However, most existing FGL approaches rely on synchronous communication, which leads to inefficiencies and is often impractical in real-world deployments. Meanwhile, current asynchronous federated learning (AFL) methods are primarily designed for conventional tasks such as image classification and natural language processing, without accounting for the unique topological properties of graph data. Directly applying these methods to graph learning can possibly result in semantic drift and representational inconsistency in the global model. To address these challenges, we propose FedSA-GCL, a semi-asynchronous federated framework that leverages both inter-client label distribution divergence and graph topological characteristics through a novel ClusterCast mechanism for efficient training. We evaluate FedSA-GCL on multiple real-world graph datasets using the Louvain and Metis split algorithms, and compare it against 9 baselines. Extensive experiments demonstrate that our method achieves strong robustness and outstanding efficiency, outperforming the baselines by an average of 2.92% with the Louvain and by 3.4% with the Metis.
Authors: Minje Park, Jeonghwa Lim, Taehyung Yu, Sunghoon Joo
Abstract: Electrocardiogram (ECG) delineation, the segmentation of meaningful waveform features, is critical for clinical diagnosis. Despite recent advances using deep learning, progress has been limited by the scarcity of publicly available annotated datasets. Semi-supervised learning presents a promising solution by leveraging abundant unlabeled ECG data. In this study, we present SemiSegECG, the first systematic benchmark for semi-supervised semantic segmentation (SemiSeg) in ECG delineation. We curated and unified multiple public datasets, including previously underused sources, to support robust and diverse evaluation. We adopted five representative SemiSeg algorithms from computer vision, implemented them on two different architectures: the convolutional network and the transformer, and evaluated them in two different settings: in-domain and cross-domain. Additionally, we propose ECG-specific training configurations and augmentation strategies and introduce a standardized evaluation framework. Our results show that the transformer outperforms the convolutional network in semi-supervised ECG delineation. We anticipate that SemiSegECG will serve as a foundation for advancing semi-supervised ECG delineation methods and will facilitate further research in this domain.
Authors: Pinhao Song, Yutong Hu, Pengteng Li, Renaud Detry
Abstract: We propose a new volumetric grasp model that is equivariant to rotations around the vertical axis, leading to a significant improvement in sample efficiency. Our model employs a tri-plane volumetric feature representation -- i.e., the projection of 3D features onto three canonical planes. We introduce a novel tri-plane feature design in which features on the horizontal plane are equivariant to 90{\deg} rotations, while the sum of features from the other two planes remains invariant to the same transformations. This design is enabled by a new deformable steerable convolution, which combines the adaptability of deformable convolutions with the rotational equivariance of steerable ones. This allows the receptive field to adapt to local object geometry while preserving equivariance properties. We further develop equivariant adaptations of two state-of-the-art volumetric grasp planners, GIGA and IGD. Specifically, we derive a new equivariant formulation of IGD's deformable attention mechanism and propose an equivariant generative model of grasp orientations based on flow matching. We provide a detailed analytical justification of the proposed equivariance properties and validate our approach through extensive simulated and real-world experiments. Our results demonstrate that the proposed projection-based design significantly reduces both computational and memory costs. Moreover, the equivariant grasp models built on top of our tri-plane features consistently outperform their non-equivariant counterparts, achieving higher performance with only a modest computational overhead. Video and code can be viewed in: https://mousecpn.github.io/evg-page/
Authors: Ran Tong, Songtao Wei, Jiaqi Liu, Lanruo Wang
Abstract: Hateful memes aimed at LGBTQ\,+ communities often evade detection by tweaking either the caption, the image, or both. We build the first robustness benchmark for this setting, pairing four realistic caption attacks with three canonical image corruptions and testing all combinations on the PrideMM dataset. Two state-of-the-art detectors, MemeCLIP and MemeBLIP2, serve as case studies, and we introduce a lightweight \textbf{Text Denoising Adapter (TDA)} to enhance the latter's resilience. Across the grid, MemeCLIP degrades more gently, while MemeBLIP2 is particularly sensitive to the caption edits that disrupt its language processing. However, the addition of the TDA not only remedies this weakness but makes MemeBLIP2 the most robust model overall. Ablations reveal that all systems lean heavily on text, but architectural choices and pre-training data significantly impact robustness. Our benchmark exposes where current multimodal safety models crack and demonstrates that targeted, lightweight modules like the TDA offer a powerful path towards stronger defences.
Authors: Xiang Fei, Siqi Wang, Shu Wei, Yuxiang Nie, Wei Shi, Hao Feng, Chao Feng, Can Huang
Abstract: Current language model training paradigms typically terminate learning upon reaching the end-of-sequence (
Authors: Murray Shanahan, Tara Das, Robert Thurman
Abstract: This paper presents a case study in the use of a large language model to generate a fictional Buddhist "sutra"', and offers a detailed analysis of the resulting text from a philosophical and literary point of view. The conceptual subtlety, rich imagery, and density of allusion found in the text make it hard to causally dismiss on account of its mechanistic origin. This raises questions about how we, as a society, should come to terms with the potentially unsettling possibility of a technology that encroaches on human meaning-making. We suggest that Buddhist philosophy, by its very nature, is well placed to adapt.
Authors: Rongyao Cai, Ming Jin, Qingsong Wen, Kexin Zhang
Abstract: Domain shift poses a fundamental challenge in time series analysis, where models trained on source domain often fail dramatically when applied in target domain with different yet similar distributions. While current unsupervised domain adaptation (UDA) methods attempt to align cross-domain feature distributions, they typically treat features as indivisible entities, ignoring their intrinsic compositions that govern domain adaptation. We introduce DARSD, a novel UDA framework with theoretical explainability that explicitly realizes UDA tasks from the perspective of representation space decomposition. Our core insight is that effective domain adaptation requires not just alignment, but principled disentanglement of transferable knowledge from mixed representations. DARSD consists of three synergistic components: (I) An adversarial learnable common invariant basis that projects original features into a domain-invariant subspace while preserving semantic content; (II) A prototypical pseudo-labeling mechanism that dynamically separates target features based on confidence, hindering error accumulation; (III) A hybrid contrastive optimization strategy that simultaneously enforces feature clustering and consistency while mitigating emerging distribution gaps. Comprehensive experiments conducted on four benchmarks (WISDM, HAR, HHAR, and MFD) demonstrate DARSD's superiority against 12 UDA algorithms, achieving optimal performance in 35 out of 53 scenarios and ranking first across all benchmarks.
Authors: Danil Savine
Abstract: This study investigates the mechanisms and factors influencing memorization in fine-tuned large language models (LLMs), with a focus on the medical domain due to its privacy-sensitive nature. We examine how different aspects of the fine-tuning process affect a model's propensity to memorize training data, using the PHEE dataset of pharmacovigilance events. Our research employs two main approaches: a membership inference attack to detect memorized data, and a generation task with prompted prefixes to assess verbatim reproduction. We analyze the impact of adapting different weight matrices in the transformer architecture, the relationship between perplexity and memorization, and the effect of increasing the rank in low-rank adaptation (LoRA) fine-tuning. Key findings include: (1) Value and Output matrices contribute more significantly to memorization compared to Query and Key matrices; (2) Lower perplexity in the fine-tuned model correlates with increased memorization; (3) Higher LoRA ranks lead to increased memorization, but with diminishing returns at higher ranks. These results provide insights into the trade-offs between model performance and privacy risks in fine-tuned LLMs. Our findings have implications for developing more effective and responsible strategies for adapting large language models while managing data privacy concerns.
Authors: Sheng-Feng Yu, Jia-Jiun Yao, Wei-Chen Chiu
Abstract: Although larger datasets are crucial for training large deep models, the rapid growth of dataset size has brought a significant challenge in terms of considerable training costs, which even results in prohibitive computational expenses. Dataset Distillation becomes a popular technique recently to reduce the dataset size via learning a highly compact set of representative exemplars, where the model trained with these exemplars ideally should have comparable performance with respect to the one trained with the full dataset. While most of existing works upon dataset distillation focus on supervised datasets, we instead aim to distill images and their self-supervisedly trained representations into a distilled set. This procedure, named as Self-Supervised Dataset Distillation, effectively extracts rich information from real datasets, yielding the distilled sets with enhanced cross-architecture generalizability. Particularly, in order to preserve the key characteristics of original dataset more faithfully and compactly, several novel techniques are proposed: 1) we introduce an innovative parameterization upon images and representations via distinct low-dimensional bases, where the base selection for parameterization is experimentally shown to play a crucial role; 2) we tackle the instability induced by the randomness of data augmentation -- a key component in self-supervised learning but being underestimated in the prior work of self-supervised dataset distillation -- by utilizing predetermined augmentations; 3) we further leverage a lightweight network to model the connections among the representations of augmented views from the same image, leading to more compact pairs of distillation. Extensive experiments conducted on various datasets validate the superiority of our approach in terms of distillation efficiency, cross-architecture generalization, and transfer learning performance.
Authors: Yuqi Pang, Bowen Yang, Yun Cao, Rong Fan, Xiaoyu Li, Chen He
Abstract: Vision large language models (VLLMs) are focusing primarily on handling complex and fine-grained visual information by incorporating advanced vision encoders and scaling up visual models. However, these approaches face high training and inference costs, as well as challenges in extracting visual details, effectively bridging across modalities. In this work, we propose a novel visual framework, MoCHA, to address these issues. Our framework integrates four vision backbones (i.e., CLIP, SigLIP, DINOv2 and ConvNeXt) to extract complementary visual features and is equipped with a sparse Mixture of Experts Connectors (MoECs) module to dynamically select experts tailored to different visual dimensions. To mitigate redundant or insufficient use of the visual information encoded by the MoECs module, we further design a Hierarchical Group Attention (HGA) with intra- and inter-group operations and an adaptive gating strategy for encoded visual features. We train MoCHA on two mainstream LLMs (e.g., Phi2-2.7B and Vicuna-7B) and evaluate their performance across various benchmarks. Notably, MoCHA outperforms state-of-the-art open-weight models on various tasks. For example, compared to CuMo (Mistral-7B), our MoCHA (Phi2-2.7B) presents outstanding abilities to mitigate hallucination by showing improvements of 3.25% in POPE and to follow visual instructions by raising 153 points on MME. Finally, ablation studies further confirm the effectiveness and robustness of the proposed MoECs and HGA in improving the overall performance of MoCHA.
Authors: Santosh Patapati, Trisanth Srinivasan, Murari Ambati
Abstract: Autonomous cars need geometric accuracy and semantic understanding to navigate complex environments, yet most stacks handle them separately. We present XYZ-Drive, a single vision-language model that reads a front-camera frame, a 25m $\times$ 25m overhead map, and the next waypoint, then outputs steering and speed. A lightweight goal-centered cross-attention layer lets waypoint tokens highlight relevant image and map patches, supporting both action and textual explanations, before the fused tokens enter a partially fine-tuned LLaMA-3.2 11B model. On the MD-NEX Outdoor-Driving benchmark XYZ-Drive attains 95% success and 0.80 Success weighted by Path Length (SPL), surpassing PhysNav-DG by 15%. and halving collisions, all while significantly improving efficiency by using only a single branch. Sixteen ablations explain the gains. Removing any modality (vision, waypoint, map) drops success by up to 11%, confirming their complementary roles and rich connections. Replacing goal-centered attention with simple concatenation cuts 3% in performance, showing query-based fusion injects map knowledge more effectively. Keeping the transformer frozen loses 5%, showing the importance of fine-tuning when applying VLMs for specific tasks such as autonomous driving. Coarsening map resolution from 10 cm to 40 cm blurs lane edges and raises crash rate. Overall, these results demonstrate that early, token-level fusion of intent and map layout enables accurate, transparent, real-time driving.
Authors: Ishani Mondal, Meera Bharadwaj, Ayush Roy, Aparna Garimella, Jordan Lee Boyd-Graber
Abstract: We present SMART-Editor, a framework for compositional layout and content editing across structured (posters, websites) and unstructured (natural images) domains. Unlike prior models that perform local edits, SMART-Editor preserves global coherence through two strategies: Reward-Refine, an inference-time rewardguided refinement method, and RewardDPO, a training-time preference optimization approach using reward-aligned layout pairs. To evaluate model performance, we introduce SMARTEdit-Bench, a benchmark covering multi-domain, cascading edit scenarios. SMART-Editor outperforms strong baselines like InstructPix2Pix and HIVE, with RewardDPO achieving up to 15% gains in structured settings and Reward-Refine showing advantages on natural images. Automatic and human evaluations confirm the value of reward-guided planning in producing semantically consistent and visually aligned edits.
Authors: Chende Zheng, Ruiqi suo, Chenhao Lin, Zhengyu Zhao, Le Yang, Shuai Liu, Minghui Yang, Cong Wang, Chao Shen
Abstract: The evolution of video generation techniques, such as Sora, has made it increasingly easy to produce high-fidelity AI-generated videos, raising public concern over the dissemination of synthetic content. However, existing detection methodologies remain limited by their insufficient exploration of temporal artifacts in synthetic videos. To bridge this gap, we establish a theoretical framework through second-order dynamical analysis under Newtonian mechanics, subsequently extending the Second-order Central Difference features tailored for temporal artifact detection. Building on this theoretical foundation, we reveal a fundamental divergence in second-order feature distributions between real and AI-generated videos. Concretely, we propose Detection by Difference of Differences (D3), a novel training-free detection method that leverages the above second-order temporal discrepancies. We validate the superiority of our D3 on 4 open-source datasets (Gen-Video, VideoPhy, EvalCrafter, VidProM), 40 subsets in total. For example, on GenVideo, D3 outperforms the previous best method by 10.39% (absolute) mean Average Precision. Additional experiments on time cost and post-processing operations demonstrate D3's exceptional computational efficiency and strong robust performance. Our code is available at https://github.com/Zig-HS/D3.
Authors: Hudson de Martim
Abstract: Structuring legal norms for machine readability is a critical prerequisite for building advanced AI and information retrieval systems, such as Legal Knowledge Graphs (LKGs). Grounded in the Functional Requirements for Bibliographic Records (FRBR) model, this paper proposes a foundational mapping for the abstract legal Work - which is materialized as the Norm node in our legal Graph RAG framework - to the interoperable schema.org/Legislation vocabulary. Using the Normas.leg.br portal as a practical case study, we demonstrate how to describe this Work entity via JSON-LD, considering stable URN identifiers, inter-norm relationships, and lifecycle properties. This structured, formal approach provides the essential first step toward creating a deterministic and verifiable knowledge graph, which can serve as a formalized "ground truth" for Legal AI applications, overcoming the limitations of purely probabilistic models.
Authors: Xuan Liu, Siru Ouyang, Xianrui Zhong, Jiawei Han, Huimin Zhao
Abstract: Large language models (LLMs) have gained significant attention in chemistry. However, most existing datasets center on molecular-level property prediction and overlook the role of fine-grained functional group (FG) information. Incorporating FG-level data can provide valuable prior knowledge that links molecular structures with textual descriptions, which can be used to build more interpretable, structure-aware LLMs for reasoning on molecule-related tasks. Moreover, LLMs can learn from such fine-grained information to uncover hidden relationships between specific functional groups and molecular properties, thereby advancing molecular design and drug discovery. Here, we introduce FGBench, a dataset comprising 625K molecular property reasoning problems with functional group information. Functional groups are precisely annotated and localized within the molecule, which ensures the dataset's interoperability thereby facilitating further multimodal applications. FGBench includes both regression and classification tasks on 245 different functional groups across three categories for molecular property reasoning: (1) single functional group impacts, (2) multiple functional group interactions, and (3) direct molecular comparisons. In the benchmark of state-of-the-art LLMs on 7K curated data, the results indicate that current LLMs struggle with FG-level property reasoning, highlighting the need to enhance reasoning capabilities in LLMs for chemistry tasks. We anticipate that the methodology employed in FGBench to construct datasets with functional group-level information will serve as a foundational framework for generating new question-answer pairs, enabling LLMs to better understand fine-grained molecular structure-property relationships. The dataset and evaluation code are available at https://github.com/xuanliugit/FGBench.
Authors: Sebastian Elbaum, Jonathan Panter
Abstract: U.S. national security customers have begun to utilize large language models, including enterprise versions of ``off-the-shelf'' models (e.g., ChatGPT) familiar to the public. This uptake will likely accelerate. However, recent studies suggest that off-the-shelf large language models frequently suggest escalatory actions when prompted with geopolitical or strategic scenarios. We demonstrate two simple, non-technical interventions to control these tendencies. Introducing these interventions into the experimental wargame design of a recent study, we substantially reduce escalation throughout the game. Calls to restrict the use of large language models in national security applications are thus premature. The U.S. government is already, and will continue, employing large language models for scenario planning and suggesting courses of action. Rather than warning against such applications, this study acknowledges the imminent adoption of large language models, and provides actionable measures to align them with national security goals, including escalation management.
Authors: Zhuo Yang, Jiaqing Xie, Shuaike Shen, Daolang Wang, Yeyun Chen, Ben Gao, Shuzhou Sun, Biqing Qi, Dongzhan Zhou, Lei Bai, Linjiang Chen, Shufei Zhang, Jun Jiang, Tianfan Fu, Yuqiang Li
Abstract: Deep learning holds immense promise for spectroscopy, yet research and evaluation in this emerging field often lack standardized formulations. To address this issue, we introduce SpectrumLab, a pioneering unified platform designed to systematize and accelerate deep learning research in spectroscopy. SpectrumLab integrates three core components: a comprehensive Python library featuring essential data processing and evaluation tools, along with leaderboards; an innovative SpectrumAnnotator module that generates high-quality benchmarks from limited seed data; and SpectrumBench, a multi-layered benchmark suite covering 14 spectroscopic tasks and over 10 spectrum types, featuring spectra curated from over 1.2 million distinct chemical substances. Thorough empirical studies on SpectrumBench with 18 cutting-edge multimodal LLMs reveal critical limitations of current approaches. We hope SpectrumLab will serve as a crucial foundation for future advancements in deep learning-driven spectroscopy.
Authors: Zhenhua Zou, Zhuotao Liu, Lepeng Zhao, Qiuyang Zhan
Abstract: The rapid adoption of agentic AI, powered by large language models (LLMs), is transforming enterprise ecosystems with autonomous agents that execute complex workflows. Yet we observe several key security vulnerabilities in LLM-driven multi-agent systems (MASes): fragmented identity frameworks, insecure communication channels, and inadequate defenses against Byzantine agents or adversarial prompts. In this paper, we present the first systematic analysis of these emerging multi-agent risks and explain why the legacy security strategies cannot effectively address these risks. Afterwards, we propose BlockA2A, the first unified multi-agent trust framework that enables secure and verifiable and agent-to-agent interoperability. At a high level, BlockA2A adopts decentralized identifiers (DIDs) to enable fine-grained cross-domain agent authentication, blockchain-anchored ledgers to enable immutable auditability, and smart contracts to dynamically enforce context-aware access control policies. BlockA2A eliminates centralized trust bottlenecks, ensures message authenticity and execution integrity, and guarantees accountability across agent interactions. Furthermore, we propose a Defense Orchestration Engine (DOE) that actively neutralizes attacks through real-time mechanisms, including Byzantine agent flagging, reactive execution halting, and instant permission revocation. Empirical evaluations demonstrate BlockA2A's effectiveness in neutralizing prompt-based, communication-based, behavioral and systemic MAS attacks. We formalize its integration into existing MAS and showcase a practical implementation for Google's A2A protocol. Experiments confirm that BlockA2A and DOE operate with sub-second overhead, enabling scalable deployment in production LLM-based MAS environments.
Authors: Yi Jiang, Sendong Zhao, Jianbo Li, Haochun Wang, Lizhe Zhang, Yan Liu, Bing Qin
Abstract: Retrieval-Augmented Generation (RAG) has emerged as a promising framework for enhancing the capabilities of Large Language Models (LLMs), especially in knowledge-intensive tasks. Despite its advantages, current RAG methods often struggle to *fully exploit knowledge during generation*. In particular, the synergy between the model's internal parametric knowledge and external retrieved knowledge remains limited. Retrieved contents may sometimes mislead generation, while certain generated content can guide the model toward more accurate outputs. In this work, we propose Collaborative Chain-of-Agents, a framework designed to enhance explicitly synergy over both parametric and retrieved knowledge. Specifically, we first introduce CoCoA-zero, a multi-agent RAG framework that first performs conditional knowledge induction and then reasons answers. Building on this, we develop CoCoA, a long-chain training strategy that synthesizes extended multi-agent reasoning trajectories from CoCoA-zero to fine-tune the LLM. This strategy enhances the model's capability to explicitly integrate and jointly leverage parametric and retrieved knowledge. Experiments results show that CoCoA-zero and CoCoA achieve superior performance on open-domain and multi-hop QA tasks.
Authors: Yebo Peng, Zixiang Liu, Yaoming Li, Zhizhuo Yang, Xinye Xu, Bowen Ye, Weijun Yuan, Zihan Wang, Tong Yang
Abstract: Evaluating the mathematical capability of Large Language Models (LLMs) is a critical yet challenging frontier. Existing benchmarks fall short, particularly for proof-centric problems, as manual creation is unscalable and costly, leaving the true mathematical abilities of LLMs largely unassessed. To overcome these barriers, we propose Proof2Hybrid, the first fully automated framework that synthesizes high-quality, proof-centric benchmarks from natural language mathematical corpora. The key novelty of our solution is Proof2X, a roadmap of converting mathematical proofs into various kinds of questions that are easy to verify. Instructed by this roadmap, we propose a new type of hybrid-formatted questions, named ``$m$-out-of-$n$ multiple judge questions'', specifically designed to enable robust, automatic evaluation while being resilient to guessing and superficial pattern matching inherent in traditional formats. As a demonstration of our framework, we introduce AlgGeoTest, a benchmark for algebraic geometry--a frontier domain of modern mathematics--comprising 456 challenging items. Our extensive evaluations on state-of-the-art LLMs using AlgGeoTest reveal profound deficits in their comprehension of algebraic geometry, providing a more precise measure of their true mathematical capabilities. Our framework and benchmark pave the way for a new wave of in-depth research into the mathematical intelligence of AI systems.
Authors: Xiaoliu Guan, Lielin Jiang, Hanqi Chen, Xu Zhang, Jiaxing Yan, Guanzhong Wang, Yi Liu, Zetao Zhang, Yu Wu
Abstract: Diffusion Transformers (DiTs) have demonstrated remarkable performance in visual generation tasks. However, their low inference speed limits their deployment in low-resource applications. Recent training-free approaches exploit the redundancy of features across timesteps by caching and reusing past representations to accelerate inference. Building on this idea, TaylorSeer instead uses cached features to predict future ones via Taylor expansion. However, its module-level prediction across all transformer blocks (e.g., attention or feedforward modules) requires storing fine-grained intermediate features, leading to notable memory and computation overhead. Moreover, it adopts a fixed caching schedule without considering the varying accuracy of predictions across timesteps, which can lead to degraded outputs when prediction fails. To address these limitations, we propose a novel approach to better leverage Taylor-based acceleration. First, we shift the Taylor prediction target from the module level to the last block level, significantly reducing the number of cached features. Furthermore, observing strong sequential dependencies among Transformer blocks, we propose to use the error between the Taylor-estimated and actual outputs of the first block as an indicator of prediction reliability. If the error is small, we trust the Taylor prediction for the last block; otherwise, we fall back to full computation, thereby enabling a dynamic caching mechanism. Empirical results show that our method achieves a better balance between speed and quality, achieving a 3.17x acceleration on FLUX, 2.36x on DiT, and 4.14x on Wan Video with negligible quality drop. The Project Page is \href{https://cg-taylor-acce.github.io/CG-Taylor/}{here.}
Authors: Kenneth Enevoldsen, Kristian N{\o}rgaard Jensen, Jan Kostkan, Bal\'azs Szab\'o, M\'arton Kardos, Kirten Vad, Johan Heinsen, Andrea Blasi N\'u\~nez, Gianluca Barmina, Jacob Nielsen, Rasmus Larsen, Peter Vahlstrup, Per M{\o}ldrup Dalum, Desmond Elliott, Lukas Galke, Peter Schneider-Kamp, Kristoffer Nielbo
Abstract: Large-scale datasets are foundational for research and development in natural language processing. However, current approaches face three key challenges: (1) reliance on ambiguously licensed sources restricting use, sharing, and derivative works; (2) static dataset releases that prevent community contributions and diminish longevity; and (3) quality assurance processes restricted to publishing teams rather than leveraging community expertise. To address these limitations, we introduce two contributions: the Dynaword approach and Danish Dynaword. The Dynaword approach is a framework for creating large-scale, open datasets that can be continuously updated through community collaboration. Danish Dynaword is a concrete implementation that validates this approach and demonstrates its potential. Danish Dynaword contains over four times as many tokens as comparable releases, is exclusively openly licensed, and has received multiple contributions across industry and research. The repository includes light-weight tests to ensure data formatting, quality, and documentation, establishing a sustainable framework for ongoing community contributions and dataset evolution.
Authors: Qianli Ma, Yaowei Zheng, Zhelun Shi, Zhongkai Zhao, Bin Jia, Ziyue Huang, Zhiqi Lin, Youjie Li, Jiacheng Yang, Yanghua Peng, Zhi Zhang, Xin Liu
Abstract: Recent advances in large language models (LLMs) have driven impressive progress in omni-modal understanding and generation. However, training omni-modal LLMs remains a significant challenge due to the heterogeneous model architectures required to process diverse modalities, necessitating sophisticated system design for efficient large-scale training. Existing frameworks typically entangle model definition with parallel logic, incurring limited scalability and substantial engineering overhead for end-to-end omni-modal training. We present VeOmni, a modular and efficient training framework to accelerate the development of omni-modal LLMs. VeOmni introduces model-centric distributed recipes that decouples communication from computation, enabling efficient 3D parallelism on omni-modal LLMs. VeOmni also features a flexible configuration interface supporting seamless integration of new modalities with minimal code change. Using VeOmni, a omni-modal mixture-of-experts (MoE) model with 30B parameters can be trained with over 2,800 tokens/sec/GPU throughput and scale to 160K context lengths via 3D parallelism on 128 GPUs, showcasing its superior efficiency and scalability for training large omni-modal LLMs.