LLM 관련 주요 논문 - 2025-09-05

1. ArcMemo: Abstract Reasoning Composition with Lifelong LLM Memory

Authors: Matthew Ho , Chen Si , Zhaoxiang Feng , Fangxu Yu , Zhijian Liu , Zhiting Hu , Lianhui Qin
URL: https://arxiv.org/abs/2509.04439
Abstract:

While inference-time scaling enables LLMs to carry out increasingly long and capable reasoning traces, the patterns and insights uncovered during these traces are immediately discarded once the context window is reset for a new query. External memory is a natural way to persist these discoveries, and recent work has shown clear benefits for reasoning-intensive tasks. We see an opportunity to make such memories more broadly reusable and scalable by moving beyond instance-based memory entries (e.g. exact query/response pairs, or summaries tightly coupled with the original problem context) toward concept-level memory: reusable, modular abstractions distilled from solution traces and stored in natural language. For future queries, relevant concepts are selectively retrieved and integrated into the prompt, enabling test-time continual learning without weight updates. Our design introduces new strategies for abstracting takeaways from rollouts and retrieving entries for new queries, promoting reuse and allowing memory to expand with additional experiences. On the challenging ARC-AGI benchmark, our method yields a 7.5% relative gain over a strong no-memory baseline with performance continuing to scale with inference compute. We find abstract concepts to be the most consistent memory design, outscoring the baseline at all tested inference compute scales. Moreover, we confirm that dynamically updating memory during test-time outperforms an otherwise identical fixed memory setting with additional attempts, supporting the hypothesis that solving more problems and abstracting more patterns to memory enables further solutions in a form of self-improvement. Code available at this https URL .

2. Psychologically Enhanced AI Agents

Authors: Maciej Besta , Shriram Chandran , Robert Gerstenberger , Mathis Lindner , Marcin Chrapek , Sebastian Hermann Martschat , Taraneh Ghandi , Patrick Iff , Hubert Niewiadomski , Piotr Nyczyk , Jürgen Müller , Torsten Hoefler
URL: https://arxiv.org/abs/2509.04343
Abstract:

We introduce MBTI-in-Thoughts, a framework for enhancing the effectiveness of Large Language Model (LLM) agents through psychologically grounded personality conditioning. Drawing on the Myers-Briggs Type Indicator (MBTI), our method primes agents with distinct personality archetypes via prompt engineering, enabling control over behavior along two foundational axes of human psychology, cognition and affect. We show that such personality priming yields consistent, interpretable behavioral biases across diverse tasks: emotionally expressive agents excel in narrative generation, while analytically primed agents adopt more stable strategies in game-theoretic settings. Our framework supports experimenting with structured multi-agent communication protocols and reveals that self-reflection prior to interaction improves cooperation and reasoning quality. To ensure trait persistence, we integrate the official 16Personalities test for automated verification. While our focus is on MBTI, we show that our approach generalizes seamlessly to other psychological frameworks such as Big Five, HEXACO, or Enneagram. By bridging psychological theory and LLM behavior design, we establish a foundation for psychologically enhanced AI agents without any fine-tuning.

3. EvoEmo: Towards Evolved Emotional Policies for LLM Agents in Multi-Turn Negotiation

Authors: Yunbo Long , Liming Xu , Lukas Beckenbauer , Yuhan Liu , Alexandra Brintrup
URL: https://arxiv.org/abs/2509.04310
Abstract:

Recent research on Chain-of-Thought (CoT) reasoning in Large Language Models (LLMs) has demonstrated that agents can engage in \textit{complex}, \textit{multi-turn} negotiations, opening new avenues for agentic AI. However, existing LLM agents largely overlook the functional role of emotions in such negotiations, instead generating passive, preference-driven emotional responses that make them vulnerable to manipulation and strategic exploitation by adversarial counterparts. To address this gap, we present EvoEmo, an evolutionary reinforcement learning framework that optimizes dynamic emotional expression in negotiations. EvoEmo models emotional state transitions as a Markov Decision Process and employs population-based genetic optimization to evolve high-reward emotion policies across diverse negotiation scenarios. We further propose an evaluation framework with two baselines – vanilla strategies and fixed-emotion strategies – for benchmarking emotion-aware negotiation. Extensive experiments and ablation studies show that EvoEmo consistently outperforms both baselines, achieving higher success rates, higher efficiency, and increased buyer savings. This findings highlight the importance of adaptive emotional expression in enabling more effective LLM agents for multi-turn negotiation.

4. Intermediate Languages Matter: Formal Languages and LLMs affect Neurosymbolic Reasoning

Authors: Alexander Beiser , David Penz , Nysret Musliu
URL: https://arxiv.org/abs/2509.04083
Abstract:

Large language models (LLMs) achieve astonishing results on a wide range of tasks. However, their formal reasoning ability still lags behind. A promising approach is Neurosymbolic LLM reasoning. It works by using LLMs as translators from natural to formal languages and symbolic solvers for deriving correct results. Still, the contributing factors to the success of Neurosymbolic LLM reasoning remain unclear. This paper demonstrates that one previously overlooked factor is the choice of the formal language. We introduce the intermediate language challenge: selecting a suitable formal language for neurosymbolic reasoning. By comparing four formal languages across three datasets and seven LLMs, we show that the choice of formal language affects both syntactic and semantic reasoning capabilities. We also discuss the varying effects across different LLMs.

5. CoT-Space: A Theoretical Framework for Internal Slow-Thinking via Reinforcement Learning

Authors: Zeyu Gan , Hao Yi , Yong Liu
URL: https://arxiv.org/abs/2509.04027
Abstract:

Reinforcement Learning (RL) has become a pivotal approach for enhancing the reasoning capabilities of Large Language Models (LLMs). However, a significant theoretical gap persists, as traditional token-level RL frameworks fail to align with the reasoning-level nature of complex, multi-step thought processes like Chain-of-Thought (CoT). To address this challenge, we introduce CoT-Space, a novel theoretical framework that recasts LLM reasoning from a discrete token-prediction task to an optimization process within a continuous, reasoning-level semantic space. By analyzing this process from both a noise perspective and a risk perspective, we demonstrate that the convergence to an optimal CoT length is a natural consequence of the fundamental trade-off between underfitting and overfitting. Furthermore, extensive experiments provide strong empirical validation for our theoretical findings. Our framework not only provides a coherent explanation for empirical phenomena such as overthinking but also offers a solid theoretical foundation to guide the future development of more effective and generalizable reasoning agents.

6. AutoPBO: LLM-powered Optimization for Local Search PBO Solvers

Authors: Jinyuan Li , Yi Chu , Yiwen Sun , Mengchuan Zou , Shaowei Cai
URL: https://arxiv.org/abs/2509.04007
Abstract:

Pseudo-Boolean Optimization (PBO) provides a powerful framework for modeling combinatorial problems through pseudo-Boolean (PB) constraints. Local search solvers have shown excellent performance in PBO solving, and their efficiency is highly dependent on their internal heuristics to guide the search. Still, their design often requires significant expert effort and manual tuning in practice. While Large Language Models (LLMs) have demonstrated potential in automating algorithm design, their application to optimizing PBO solvers remains unexplored. In this work, we introduce AutoPBO, a novel LLM-powered framework to automatically enhance PBO local search solvers. We conduct experiments on a broad range of four public benchmarks, including one real-world benchmark, a benchmark from PB competition, an integer linear programming optimization benchmark, and a crafted combinatorial benchmark, to evaluate the performance improvement achieved by AutoPBO and compare it with six state-of-the-art competitors, including two local search PBO solvers NuPBO and OraSLS, two complete PB solvers PBO-IHS and RoundingSat, and two mixed integer programming (MIP) solvers Gurobi and SCIP. AutoPBO demonstrates significant improvements over previous local search approaches, while maintaining competitive performance compared to state-of-the-art competitors. The results suggest that AutoPBO offers a promising approach to automating local search solver design.

7. Meta-Policy Reflexion: Reusable Reflective Memory and Rule Admissibility for Resource-Efficient LLM Agent

Authors: Chunlong Wu , Zhibo Qu
URL: https://arxiv.org/abs/2509.03990
Abstract:

Large language model (LLM) agents achieve impressive single-task performance but commonly exhibit repeated failures, inefficient exploration, and limited cross-task adaptability. Existing reflective strategies (e.g., Reflexion, ReAct) improve per-episode behavior but typically produce ephemeral, task-specific traces that are not reused across tasks. Reinforcement-learning based alternatives can produce transferable policies but require substantial parameter updates and compute. In this work we introduce Meta-Policy Reflexion (MPR): a hybrid framework that consolidates LLM-generated reflections into a structured, predicate-like Meta-Policy Memory (MPM) and applies that memory at inference time through two complementary mechanisms soft memory-guided decoding and hard rule admissibility checks(HAC). MPR (i) externalizes reusable corrective knowledge without model weight updates, (ii) enforces domain constraints to reduce unsafe or invalid actions, and (iii) retains the adaptability of language-based reflection. We formalize the MPM representation, present algorithms for update and decoding, and validate the approach in a text-based agent environment following the experimental protocol described in the provided implementation (AlfWorld-based). Empirical results reported in the supplied material indicate consistent gains in execution accuracy and robustness when compared to Reflexion baselines; rule admissibility further improves stability. We analyze mechanisms that explain these gains, discuss scalability and failure modes, and outline future directions for multimodal and multi?agent extensions.

8. World Model Implanting for Test-time Adaptation of Embodied Agents

Authors: Minjong Yoo , Jinwoo Jang , Sihyung Yoon , Honguk Woo
URL: https://arxiv.org/abs/2509.03956
Abstract:

In embodied AI, a persistent challenge is enabling agents to robustly adapt to novel domains without requiring extensive data collection or retraining. To address this, we present a world model implanting framework (WorMI) that combines the reasoning capabilities of large language models (LLMs) with independently learned, domain-specific world models through test-time composition. By allowing seamless implantation and removal of the world models, the embodied agent’s policy achieves and maintains cross-domain adaptability. In the WorMI framework, we employ a prototype-based world model retrieval approach, utilizing efficient trajectory-based abstract representation matching, to incorporate relevant models into test-time composition. We also develop a world-wise compound attention method that not only integrates the knowledge from the retrieved world models but also aligns their intermediate representations with the reasoning model’s representation within the agent’s policy. This framework design effectively fuses domain-specific knowledge from multiple world models, ensuring robust adaptation to unseen domains. We evaluate our WorMI on the VirtualHome and ALFWorld benchmarks, demonstrating superior zero-shot and few-shot performance compared to several LLM-based approaches across a range of unseen domains. These results highlight the frameworks potential for scalable, real-world deployment in embodied agent scenarios where adaptability and data efficiency are essential.

9. A Foundation Model for Chest X-ray Interpretation with Grounded Reasoning via Online Reinforcement Learning

Authors: Qika Lin , Yifan Zhu , Bin Pu , Ling Huang , Haoran Luo , Jingying Ma , Zhen Peng , Tianzhe Zhao , Fangzhi Xu , Jian Zhang , Kai He , Zhonghong Ou , Swapnil Mishra , Mengling Feng
URL: https://arxiv.org/abs/2509.03906
Abstract:

Medical foundation models (FMs) have shown tremendous promise amid the rapid advancements in artificial intelligence (AI) technologies. However, current medical FMs typically generate answers in a black-box manner, lacking transparent reasoning processes and locally grounded interpretability, which hinders their practical clinical deployments. To this end, we introduce DeepMedix-R1, a holistic medical FM for chest X-ray (CXR) interpretation. It leverages a sequential training pipeline: initially fine-tuned on curated CXR instruction data to equip with fundamental CXR interpretation capabilities, then exposed to high-quality synthetic reasoning samples to enable cold-start reasoning, and finally refined via online reinforcement learning to enhance both grounded reasoning quality and generation performance. Thus, the model produces both an answer and reasoning steps tied to the image’s local regions for each query. Quantitative evaluation demonstrates substantial improvements in report generation (e.g., 14.54% and 31.32% over LLaVA-Rad and MedGemma) and visual question answering (e.g., 57.75% and 23.06% over MedGemma and CheXagent) tasks. To facilitate robust assessment, we propose Report Arena, a benchmarking framework using advanced language models to evaluate answer quality, further highlighting the superiority of DeepMedix-R1. Expert review of generated reasoning steps reveals greater interpretability and clinical plausibility compared to the established Qwen2.5-VL-7B model (0.7416 vs. 0.2584 overall preference). Collectively, our work advances medical FM development toward holistic, transparent, and clinically actionable modeling for CXR interpretation.

10. FaMA: LLM-Empowered Agentic Assistant for Consumer-to-Consumer Marketplace

Authors: Yineng Yan , Xidong Wang , Jin Seng Cheng , Ran Hu , Wentao Guan , Nahid Farahmand , Hengte Lin , Yue Li
URL: https://arxiv.org/abs/2509.03890
Abstract:

The emergence of agentic AI, powered by Large Language Models (LLMs), marks a paradigm shift from reactive generative systems to proactive, goal-oriented autonomous agents capable of sophisticated planning, memory, and tool use. This evolution presents a novel opportunity to address long-standing challenges in complex digital environments. Core tasks on Consumer-to-Consumer (C2C) e-commerce platforms often require users to navigate complex Graphical User Interfaces (GUIs), making the experience time-consuming for both buyers and sellers. This paper introduces a novel approach to simplify these interactions through an LLM-powered agentic assistant. This agent functions as a new, conversational entry point to the marketplace, shifting the primary interaction model from a complex GUI to an intuitive AI agent. By interpreting natural language commands, the agent automates key high-friction workflows. For sellers, this includes simplified updating and renewal of listings, and the ability to send bulk messages. For buyers, the agent facilitates a more efficient product discovery process through conversational search. We present the architecture for Facebook Marketplace Assistant (FaMA), arguing that this agentic, conversational paradigm provides a lightweight and more accessible alternative to traditional app interfaces, allowing users to manage their marketplace activities with greater efficiency. Experiments show FaMA achieves a 98% task success rate on solving complex tasks on the marketplace and enables up to a 2x speedup on interaction time.

11. Expedition & Expansion: Leveraging Semantic Representations for Goal-Directed Exploration in Continuous Cellular Automata

Authors: Sina Khajehabdollahi , Gautier Hamon , Marko Cvjetko , Pierre-Yves Oudeyer , Clément Moulin-Frier , Cédric Colas
URL: https://arxiv.org/abs/2509.03863
Abstract:

Discovering diverse visual patterns in continuous cellular automata (CA) is challenging due to the vastness and redundancy of high-dimensional behavioral spaces. Traditional exploration methods like Novelty Search (NS) expand locally by mutating known novel solutions but often plateau when local novelty is exhausted, failing to reach distant, unexplored regions. We introduce Expedition and Expansion (E&E), a hybrid strategy where exploration alternates between local novelty-driven expansions and goal-directed expeditions. During expeditions, E&E leverages a Vision-Language Model (VLM) to generate linguistic goals–descriptions of interesting but hypothetical patterns that drive exploration toward uncharted regions. By operating in semantic spaces that align with human perception, E&E both evaluates novelty and generates goals in conceptually meaningful ways, enhancing the interpretability and relevance of discovered behaviors. Tested on Flow Lenia, a continuous CA known for its rich, emergent behaviors, E&E consistently uncovers more diverse solutions than existing exploration methods. A genealogical analysis further reveals that solutions originating from expeditions disproportionately influence long-term exploration, unlocking new behavioral niches that serve as stepping stones for subsequent search. These findings highlight E&E’s capacity to break through local novelty boundaries and explore behavioral landscapes in human-aligned, interpretable ways, offering a promising template for open-ended exploration in artificial life and beyond.

12. Continuous Monitoring of Large-Scale Generative AI via Deterministic Knowledge Graph Structures

Authors: Kishor Datta Gupta , Mohd Ariful Haque , Hasmot Ali , Marufa Kamal , Syed Bahauddin Alam , Mohammad Ashiqur Rahman
URL: https://arxiv.org/abs/2509.03857
Abstract:

Generative AI (GEN AI) models have revolutionized diverse application domains but present substantial challenges due to reliability concerns, including hallucinations, semantic drift, and inherent biases. These models typically operate as black-boxes, complicating transparent and objective evaluation. Current evaluation methods primarily depend on subjective human assessment, limiting scalability, transparency, and effectiveness. This research proposes a systematic methodology using deterministic and Large Language Model (LLM)-generated Knowledge Graphs (KGs) to continuously monitor and evaluate GEN AI reliability. We construct two parallel KGs: (i) a deterministic KG built using explicit rule-based methods, predefined ontologies, domain-specific dictionaries, and structured entity-relation extraction rules, and (ii) an LLM-generated KG dynamically derived from real-time textual data streams such as live news articles. Utilizing real-time news streams ensures authenticity, mitigates biases from repetitive training, and prevents adaptive LLMs from bypassing predefined benchmarks through feedback memorization. To quantify structural deviations and semantic discrepancies, we employ several established KG metrics, including Instantiated Class Ratio (ICR), Instantiated Property Ratio (IPR), and Class Instantiation (CI). An automated real-time monitoring framework continuously computes deviations between deterministic and LLM-generated KGs. By establishing dynamic anomaly thresholds based on historical structural metric distributions, our method proactively identifies and flags significant deviations, thus promptly detecting semantic anomalies or hallucinations. This structured, metric-driven comparison between deterministic and dynamically generated KGs delivers a robust and scalable evaluation framework.

13. An Agentic Model Context Protocol Framework for Medical Concept Standardization

Authors: Jaerong Ahn , Andrew Wen , Nan Wang , Heling Jia , Zhiyi Yue , Sunyang Fu , Hongfang Liu
URL: https://arxiv.org/abs/2509.03828
Abstract:

The Observational Medical Outcomes Partnership (OMOP) common data model (CDM) provides a standardized representation of heterogeneous health data to support large-scale, multi-institutional research. One critical step in data standardization using OMOP CDM is the mapping of source medical terms to OMOP standard concepts, a procedure that is resource-intensive and error-prone. While large language models (LLMs) have the potential to facilitate this process, their tendency toward hallucination makes them unsuitable for clinical deployment without training and expert validation. Here, we developed a zero-training, hallucination-preventive mapping system based on the Model Context Protocol (MCP), a standardized and secure framework allowing LLMs to interact with external resources and tools. The system enables explainable mapping and significantly improves efficiency and accuracy with minimal effort. It provides real-time vocabulary lookups and structured reasoning outputs suitable for immediate use in both exploratory and production environments.

14. What Would an LLM Do? Evaluating Policymaking Capabilities of Large Language Models

Authors: Pierre Le Coz , Jia An Liu , Debarun Bhattacharjya , Georgina Curto , Serge Stinckwich
URL: https://arxiv.org/abs/2509.03827
Abstract:

Large language models (LLMs) are increasingly being adopted in high-stakes domains. Their capacity to process vast amounts of unstructured data, explore flexible scenarios, and handle a diversity of contextual factors can make them uniquely suited to provide new insights for the complexity of social policymaking. This article evaluates whether LLMs’ are aligned with domain experts (and among themselves) to inform social policymaking on the subject of homelessness alleviation - a challenge affecting over 150 million people worldwide. We develop a novel benchmark comprised of decision scenarios with policy choices across four geographies (South Bend, USA; Barcelona, Spain; Johannesburg, South Africa; Macau SAR, China). The policies in scope are grounded in the conceptual framework of the Capability Approach for human development. We also present an automated pipeline that connects the benchmarked policies to an agent-based model, and we explore the social impact of the recommended policies through simulated social scenarios. The paper results reveal promising potential to leverage LLMs for social policy making. If responsible guardrails and contextual calibrations are introduced in collaboration with local domain experts, LLMs can provide humans with valuable insights, in the form of alternative policies at scale.

15. Learning to Deliberate: Meta-policy Collaboration for Agentic LLMs with Multi-agent Reinforcement Learning

Authors: Wei Yang , Jesse Thomason
URL: https://arxiv.org/abs/2509.03817
Abstract:

Multi-agent systems of large language models (LLMs) show promise for complex reasoning, but their effectiveness is often limited by fixed collaboration protocols. These frameworks typically focus on macro-level orchestration while overlooking agents’ internal deliberative capabilities. This critical meta-cognitive blindspot treats agents as passive executors unable to adapt their strategy based on internal cognitive states like uncertainty or confidence. We introduce the Meta-Policy Deliberation Framework (MPDF), where agents learn a decentralized policy over a set of high-level meta-cognitive actions: Persist, Refine, and Concede. To overcome the instability of traditional policy gradients in this setting, we develop SoftRankPO, a novel reinforcement learning algorithm. SoftRankPO stabilizes training by shaping advantages based on the rank of rewards mapped through smooth normal quantiles, making the learning process robust to reward variance. Experiments show that MPDF with SoftRankPO achieves a a 4-5% absolute gain in average accuracy across five mathematical and general reasoning benchmarks compared to six state-of-the-art heuristic and learning-based multi-agent reasoning algorithms. Our work presents a paradigm for learning adaptive, meta-cognitive policies for multi-agent LLM systems, shifting the focus from designing fixed protocols to learning dynamic, deliberative strategies.

16. Leveraging LLM-Based Agents for Intelligent Supply Chain Planning

Authors: Yongzhi Qi , Jiaheng Yin , Jianshen Zhang , Dongyang Geng , Zhengyu Chen , Hao Hu , Wei Qi , Zuo-Jun Max Shen
URL: https://arxiv.org/abs/2509.03811
Abstract:

In supply chain management, planning is a critical concept. The movement of physical products across different categories, from suppliers to warehouse management, to sales, and logistics transporting them to customers, entails the involvement of many entities. It covers various aspects such as demand forecasting, inventory management, sales operations, and replenishment. How to collect relevant data from an e-commerce platform’s perspective, formulate long-term plans, and dynamically adjust them based on environmental changes, while ensuring interpretability, efficiency, and reliability, is a practical and challenging problem. In recent years, the development of AI technologies, especially the rapid progress of large language models, has provided new tools to address real-world issues. In this work, we construct a Supply Chain Planning Agent (SCPA) framework that can understand domain knowledge, comprehend the operator’s needs, decompose tasks, leverage or create new tools, and return evidence-based planning reports. We deploy this framework in this http URL ‘s real-world scenario, demonstrating the feasibility of LLM-agent applications in the supply chain. It effectively reduced labor and improved accuracy, stock availability, and other key metrics.

17. RAGuard: A Novel Approach for in-context Safe Retrieval Augmented Generation for LLMs

Authors: Connor Walker , Koorosh Aslansefat , Mohammad Naveed Akram , Yiannis Papadopoulos
URL: https://arxiv.org/abs/2509.03768
Abstract:

Accuracy and safety are paramount in Offshore Wind (OSW) maintenance, yet conventional Large Language Models (LLMs) often fail when confronted with highly specialised or unexpected scenarios. We introduce RAGuard, an enhanced Retrieval-Augmented Generation (RAG) framework that explicitly integrates safety-critical documents alongside technical this http URL issuing parallel queries to two indices and allocating separate retrieval budgets for knowledge and safety, RAGuard guarantees both technical depth and safety coverage. We further develop a SafetyClamp extension that fetches a larger candidate pool, “hard-clamping” exact slot guarantees to safety. We evaluate across sparse (BM25), dense (Dense Passage Retrieval) and hybrid retrieval paradigms, measuring Technical Recall@K and Safety Recall@K. Both proposed extensions of RAG show an increase in Safety Recall@K from almost 0\% in RAG to more than 50\% in RAGuard, while maintaining Technical Recall above 60\%. These results demonstrate that RAGuard and SafetyClamp have the potential to establish a new standard for integrating safety assurance into LLM-powered decision support in critical maintenance contexts.

Authors: James Mooney , Josef Woldense , Zheng Robert Jia , Shirley Anugrah Hayati , My Ha Nguyen , Vipul Raheja , Dongyeop Kang
URL: https://arxiv.org/abs/2509.03736
Abstract:

The impressive capabilities of Large Language Models (LLMs) have fueled the notion that synthetic agents can serve as substitutes for real participants in human-subject research. In an effort to evaluate the merits of this claim, social science researchers have largely focused on whether LLM-generated survey data corresponds to that of a human counterpart whom the LLM is prompted to represent. In contrast, we address a more fundamental question: Do agents maintain internal consistency, retaining similar behaviors when examined under different experimental settings? To this end, we develop a study designed to (a) reveal the agent’s internal state and (b) examine agent behavior in a basic dialogue setting. This design enables us to explore a set of behavioral hypotheses to assess whether an agent’s conversation behavior is consistent with what we would expect from their revealed internal state. Our findings on these hypotheses show significant internal inconsistencies in LLMs across model families and at differing model sizes. Most importantly, we find that, although agents may generate responses matching those of their human counterparts, they fail to be internally consistent, representing a critical gap in their capabilities to accurately substitute for real participants in human-subject research. Our simulation code and data are publicly accessible.

19. The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior in LLMs

Authors: Pengrui Han , Rafal Kocielnik , Peiyang Song , Ramit Debnath , Dean Mobbs , Anima Anandkumar , R. Michael Alvarez
URL: https://arxiv.org/abs/2509.03730
Abstract:

Personality traits have long been studied as predictors of human this http URL advances in Large Language Models (LLMs) suggest similar patterns may emerge in artificial systems, with advanced LLMs displaying consistent behavioral tendencies resembling human traits like agreeableness and self-regulation. Understanding these patterns is crucial, yet prior work primarily relied on simplified self-reports and heuristic prompting, with little behavioral validation. In this study, we systematically characterize LLM personality across three dimensions: (1) the dynamic emergence and evolution of trait profiles throughout training stages; (2) the predictive validity of self-reported traits in behavioral tasks; and (3) the impact of targeted interventions, such as persona injection, on both self-reports and behavior. Our findings reveal that instructional alignment (e.g., RLHF, instruction tuning) significantly stabilizes trait expression and strengthens trait correlations in ways that mirror human data. However, these self-reported traits do not reliably predict behavior, and observed associations often diverge from human patterns. While persona injection successfully steers self-reports in the intended direction, it exerts little or inconsistent effect on actual behavior. By distinguishing surface-level trait expression from behavioral consistency, our findings challenge assumptions about LLM personality and underscore the need for deeper evaluation in alignment and interpretability.

20. Emergent Hierarchical Reasoning in LLMs through Reinforcement Learning

Authors: Haozhe Wang , Qixin Xu , Che Liu , Junhong Wu , Fangzhen Lin , Wenhu Chen
URL: https://arxiv.org/abs/2509.03646
Abstract:

Reinforcement Learning (RL) has proven highly effective at enhancing the complex reasoning abilities of Large Language Models (LLMs), yet underlying mechanisms driving this success remain largely opaque. Our analysis reveals that puzzling phenomena like aha moments",length-scaling’’ and entropy dynamics are not disparate occurrences but hallmarks of an emergent reasoning hierarchy, akin to the separation of high-level strategic planning from low-level procedural execution in human cognition. We uncover a compelling two-phase dynamic: initially, a model is constrained by procedural correctness and must improve its low-level skills. The learning bottleneck then decisively shifts, with performance gains being driven by the exploration and mastery of high-level strategic planning. This insight exposes a core inefficiency in prevailing RL algorithms like GRPO, which apply optimization pressure agnostically and dilute the learning signal across all tokens. To address this, we propose HIerarchy-Aware Credit Assignment (HICRA), an algorithm that concentrates optimization efforts on high-impact planning tokens. HICRA significantly outperforms strong baselines, demonstrating that focusing on this strategic bottleneck is key to unlocking advanced reasoning. Furthermore, we validate semantic entropy as a superior compass for measuring strategic exploration over misleading metrics such as token-level entropy.

21. Towards a Neurosymbolic Reasoning System Grounded in Schematic Representations

Authors: François Olivier , Zied Bouraoui
URL: https://arxiv.org/abs/2509.03644
Abstract:

Despite significant progress in natural language understanding, Large Language Models (LLMs) remain error-prone when performing logical reasoning, often lacking the robust mental representations that enable human-like comprehension. We introduce a prototype neurosymbolic system, Embodied-LM, that grounds understanding and logical reasoning in schematic representations based on image schemas-recurring patterns derived from sensorimotor experience that structure human cognition. Our system operationalizes the spatial foundations of these cognitive structures using declarative spatial reasoning within Answer Set Programming. Through evaluation on logical deduction problems, we demonstrate that LLMs can be guided to interpret scenarios through embodied cognitive structures, that these structures can be formalized as executable programs, and that the resulting representations support effective logical reasoning with enhanced interpretability. While our current implementation focuses on spatial primitives, it establishes the computational foundation for incorporating more complex and dynamic representations.

22. CausalARC: Abstract Reasoning with Causal World Models

Authors: Jacqueline Maasch , John Kalantari , Kia Khezeli
URL: https://arxiv.org/abs/2509.03636
Abstract:

Reasoning requires adaptation to novel problem settings under limited data and distribution shift. This work introduces CausalARC: an experimental testbed for AI reasoning in low-data and out-of-distribution regimes, modeled after the Abstraction and Reasoning Corpus (ARC). Each CausalARC reasoning task is sampled from a fully specified causal world model, formally expressed as a structural causal model. Principled data augmentations provide observational, interventional, and counterfactual feedback about the world model in the form of few-shot, in-context learning demonstrations. As a proof-of-concept, we illustrate the use of CausalARC for four language model evaluation settings: (1) abstract reasoning with test-time training, (2) counterfactual reasoning with in-context learning, (3) program synthesis, and (4) causal discovery with logical reasoning.

23. Explainable Knowledge Graph Retrieval-Augmented Generation (KG-RAG) with KG-SMILE

Authors: Zahra Zehtabi Sabeti Moghaddam , Zeinab Dehghani , Maneeha Rani , Koorosh Aslansefat , Bhupesh Kumar Mishra , Rameez Raja Kureshi , Dhavalkumar Thakker
URL: https://arxiv.org/abs/2509.03626
Abstract:

Generative AI, such as Large Language Models (LLMs), has achieved impressive progress but still produces hallucinations and unverifiable claims, limiting reliability in sensitive domains. Retrieval-Augmented Generation (RAG) improves accuracy by grounding outputs in external knowledge, especially in domains like healthcare, where precision is vital. However, RAG remains opaque and essentially a black box, heavily dependent on data quality. We developed a method-agnostic, perturbation-based framework that provides token and component-level interoperability for Graph RAG using SMILE and named it as Knowledge-Graph (KG)-SMILE. By applying controlled perturbations, computing similarities, and training weighted linear surrogates, KG-SMILE identifies the graph entities and relations most influential to generated outputs, thereby making RAG more transparent. We evaluate KG-SMILE using comprehensive attribution metrics, including fidelity, faithfulness, consistency, stability, and accuracy. Our findings show that KG-SMILE produces stable, human-aligned explanations, demonstrating its capacity to balance model effectiveness with interpretability and thereby fostering greater transparency and trust in machine learning technologies.

24. Learning When to Plan: Efficiently Allocating Test-Time Compute for LLM Agents

Authors: Davide Paglieri , Bartłomiej Cupiał , Jonathan Cook , Ulyana Piterbarg , Jens Tuyls , Edward Grefenstette , Jakob Nicolaus Foerster , Jack Parker-Holder , Tim Rocktäschel
URL: https://arxiv.org/abs/2509.03581
Abstract:

Training large language models (LLMs) to reason via reinforcement learning (RL) significantly improves their problem-solving capabilities. In agentic settings, existing methods like ReAct prompt LLMs to explicitly plan before every action; however, we demonstrate that always planning is computationally expensive and degrades performance on long-horizon tasks, while never planning further limits performance. To address this, we introduce a conceptual framework formalizing dynamic planning for LLM agents, enabling them to flexibly decide when to allocate test-time compute for planning. We propose a simple two-stage training pipeline: (1) supervised fine-tuning on diverse synthetic data to prime models for dynamic planning, and (2) RL to refine this capability in long-horizon environments. Experiments on the Crafter environment show that dynamic planning agents trained with this approach are more sample-efficient and consistently achieve more complex objectives. Additionally, we demonstrate that these agents can be effectively steered by human-written plans, surpassing their independent capabilities. To our knowledge, this work is the first to explore training LLM agents for dynamic test-time compute allocation in sequential decision-making tasks, paving the way for more efficient, adaptive, and controllable agentic systems.

25. PG-Agent: An Agent Powered by Page Graph

Authors: Weizhi Chen , Ziwei Wang , Leyang Yang , Sheng Zhou , Xiaoxuan Tang , Jiajun Bu , Yong Li , Wei Jiang
URL: https://arxiv.org/abs/2509.03536
Abstract:

Graphical User Interface (GUI) agents possess significant commercial and social value, and GUI agents powered by advanced multimodal large language models (MLLMs) have demonstrated remarkable potential. Currently, existing GUI agents usually utilize sequential episodes of multi-step operations across pages as the prior GUI knowledge, which fails to capture the complex transition relationship between pages, making it challenging for the agents to deeply perceive the GUI environment and generalize to new scenarios. Therefore, we design an automated pipeline to transform the sequential episodes into page graphs, which explicitly model the graph structure of the pages that are naturally connected by actions. To fully utilize the page graphs, we further introduce Retrieval-Augmented Generation (RAG) technology to effectively retrieve reliable perception guidelines of GUI from them, and a tailored multi-agent framework PG-Agent with task decomposition strategy is proposed to be injected with the guidelines so that it can generalize to unseen scenarios. Extensive experiments on various benchmarks demonstrate the effectiveness of PG-Agent, even with limited episodes for page graph construction.

26. Delta Activations: A Representation for Finetuned Large Language Models

Authors: Zhiqiu Xu , Amish Sethi , Mayur Naik , Ser-Nam Lim
URL: https://arxiv.org/abs/2509.04442
Abstract:

The success of powerful open source Large Language Models (LLMs) has enabled the community to create a vast collection of post-trained models adapted to specific tasks and domains. However, navigating and understanding these models remains challenging due to inconsistent metadata and unstructured repositories. We introduce Delta Activations, a method to represent finetuned models as vector embeddings by measuring shifts in their internal activations relative to a base model. This representation allows for effective clustering by domain and task, revealing structure in the model landscape. Delta Activations also demonstrate desirable properties: it is robust across finetuning settings and exhibits an additive property when finetuning datasets are mixed. In addition, we show that Delta Activations can embed tasks via few-shot finetuning, and further explore its use for model selection and merging. We hope Delta Activations can facilitate the practice of reusing publicly available models. Code is available at this https URL .

27. Towards a Unified View of Large Language Model Post-Training

Authors: Xingtai Lv , Yuxin Zuo , Youbang Sun , Hongyi Liu , Yuntian Wei , Zhekai Chen , Lixuan He , Xuekai Zhu , Kaiyan Zhang , Bingning Wang , Ning Ding , Bowen Zhou
URL: https://arxiv.org/abs/2509.04419
Abstract:

Two major sources of training data exist for post-training modern language models: online (model-generated rollouts) data, and offline (human or other-model demonstrations) data. These two types of data are typically used by approaches like Reinforcement Learning (RL) and Supervised Fine-Tuning (SFT), respectively. In this paper, we show that these approaches are not in contradiction, but are instances of a single optimization process. We derive a Unified Policy Gradient Estimator, and present the calculations of a wide spectrum of post-training approaches as the gradient of a common objective under different data distribution assumptions and various bias-variance tradeoffs. The gradient estimator is constructed with four interchangeable parts: stabilization mask, reference policy denominator, advantage estimate, and likelihood gradient. Motivated by our theoretical findings, we propose Hybrid Post-Training (HPT), an algorithm that dynamically selects different training signals. HPT is designed to yield both effective exploitation of demonstration and stable exploration without sacrificing learned reasoning patterns. We provide extensive experiments and ablation studies to verify the effectiveness of our unified theoretical framework and HPT. Across six mathematical reasoning benchmarks and two out-of-distribution suites, HPT consistently surpasses strong baselines across models of varying scales and families.

28. No Thoughts Just AI: Biased LLM Recommendations Limit Human Agency in Resume Screening

Authors: Kyra Wilson , Mattea Sim , Anna-Maria Gueorguieva , Aylin Caliskan
URL: https://arxiv.org/abs/2509.04404
Abstract:

In this study, we conduct a resume-screening experiment (N=528) where people collaborate with simulated AI models exhibiting race-based preferences (bias) to evaluate candidates for 16 high and low status occupations. Simulated AI bias approximates factual and counterfactual estimates of racial bias in real-world AI systems. We investigate people’s preferences for White, Black, Hispanic, and Asian candidates (represented through names and affinity groups on quality-controlled resumes) across 1,526 scenarios and measure their unconscious associations between race and status using implicit association tests (IATs), which predict discriminatory hiring decisions but have not been investigated in human-AI collaboration. When making decisions without AI or with AI that exhibits no race-based preferences, people select all candidates at equal rates. However, when interacting with AI favoring a particular group, people also favor those candidates up to 90% of the time, indicating a significant behavioral shift. The likelihood of selecting candidates whose identities do not align with common race-status stereotypes can increase by 13% if people complete an IAT before conducting resume screening. Finally, even if people think AI recommendations are low quality or not important, their decisions are still vulnerable to AI bias under certain circumstances. This work has implications for people’s autonomy in AI-HITL scenarios, AI and work, design and evaluation of AI hiring systems, and strategies for mitigating bias in collaborative decision-making tasks. In particular, organizational and regulatory policy should acknowledge the complex nature of AI-HITL decision making when implementing these systems, educating people who use them, and determining which are subject to oversight.

29. Facts Fade Fast: Evaluating Memorization of Outdated Medical Knowledge in Large Language Models

Authors: Juraj Vladika , Mahdi Dhaini , Florian Matthes
URL: https://arxiv.org/abs/2509.04304
Abstract:

The growing capabilities of Large Language Models (LLMs) show significant potential to enhance healthcare by assisting medical researchers and physicians. However, their reliance on static training data is a major risk when medical recommendations evolve with new research and developments. When LLMs memorize outdated medical knowledge, they can provide harmful advice or fail at clinical reasoning tasks. To investigate this problem, we introduce two novel question-answering (QA) datasets derived from systematic reviews: MedRevQA (16,501 QA pairs covering general biomedical knowledge) and MedChangeQA (a subset of 512 QA pairs where medical consensus has changed over time). Our evaluation of eight prominent LLMs on the datasets reveals consistent reliance on outdated knowledge across all models. We additionally analyze the influence of obsolete pre-training data and training strategies to explain this phenomenon and propose future directions for mitigation, laying the groundwork for developing more current and reliable medical AI systems.

30. An Empirical Study of Vulnerabilities in Python Packages and Their Detection

Authors: Haowei Quan , Junjie Wang , Xinzhe Li , Terry Yue Zhuo , Xiao Chen , Xiaoning Du
URL: https://arxiv.org/abs/2509.04260
Abstract:

In the rapidly evolving software development landscape, Python stands out for its simplicity, versatility, and extensive ecosystem. Python packages, as units of organization, reusability, and distribution, have become a pressing concern, highlighted by the considerable number of vulnerability reports. As a scripting language, Python often cooperates with other languages for performance or interoperability. This adds complexity to the vulnerabilities inherent to Python packages, and the effectiveness of current vulnerability detection tools remains underexplored. This paper addresses these gaps by introducing PyVul, the first comprehensive benchmark suite of Python-package vulnerabilities. PyVul includes 1,157 publicly reported, developer-verified vulnerabilities, each linked to its affected packages. To accommodate diverse detection techniques, it provides annotations at both commit and function levels. An LLM-assisted data cleansing method is incorporated to improve label accuracy, achieving 100% commit-level and 94% function-level accuracy, establishing PyVul as the most precise large-scale Python vulnerability benchmark. We further carry out a distribution analysis of PyVul, which demonstrates that vulnerabilities in Python packages involve multiple programming languages and exhibit a wide variety of types. Moreover, our analysis reveals that multi-lingual Python packages are potentially more susceptible to vulnerabilities. Evaluation of state-of-the-art detectors using this benchmark reveals a significant discrepancy between the capabilities of existing tools and the demands of effectively identifying real-world security issues in Python packages. Additionally, we conduct an empirical review of the top-ranked CWEs observed in Python packages, to diagnose the fine-grained limitations of current detection tools and highlight the necessity for future advancements in the field.

31. How many patients could we save with LLM priors?

Authors: Shota Arai , David Selby , Andrew Vargo , Sebastian Vollmer
URL: https://arxiv.org/abs/2509.04250
Abstract:

Imagine a world where clinical trials need far fewer patients to achieve the same statistical power, thanks to the knowledge encoded in large language models (LLMs). We present a novel framework for hierarchical Bayesian modeling of adverse events in multi-center clinical trials, leveraging LLM-informed prior distributions. Unlike data augmentation approaches that generate synthetic data points, our methodology directly obtains parametric priors from the model. Our approach systematically elicits informative priors for hyperparameters in hierarchical Bayesian models using a pre-trained LLM, enabling the incorporation of external clinical expertise directly into Bayesian safety modeling. Through comprehensive temperature sensitivity analysis and rigorous cross-validation on real-world clinical trial data, we demonstrate that LLM-derived priors consistently improve predictive performance compared to traditional meta-analytical approaches. This methodology paves the way for more efficient and expert-informed clinical trial design, enabling substantial reductions in the number of patients required to achieve robust safety assessment and with the potential to transform drug safety monitoring and regulatory decision making.

32. Learning Active Perception via Self-Evolving Preference Optimization for GUI Grounding

Authors: Wanfu Wang , Qipeng Huang , Guangquan Xue , Xiaobo Liang , Juntao Li
URL: https://arxiv.org/abs/2509.04243
Abstract:

Vision Language Models (VLMs) have recently achieved significant progress in bridging visual perception and linguistic reasoning. Recently, OpenAI o3 model introduced a zoom-in search strategy that effectively elicits active perception capabilities in VLMs, improving downstream task performance. However, enabling VLMs to reason effectively over appropriate image regions remains a core challenge in GUI grounding, particularly under high-resolution inputs and complex multi-element visual interactions. In this work, we propose LASER, a self-evolving framework that progressively endows VLMs with multi-step perception capabilities, enabling precise coordinate prediction. Specifically, our approach integrate Monte Carlo quality estimation with Intersection-over-Union (IoU)-based region quality evaluation to jointly encourage both accuracy and diversity in constructing high-quality preference data. This combination explicitly guides the model to focus on instruction-relevant key regions while adaptively allocating reasoning steps based on task complexity. Comprehensive experiments on the ScreenSpot Pro and ScreenSpot-v2 benchmarks demonstrate consistent performance gains, validating the effectiveness of our method. Furthermore, when fine-tuned on GTA1-7B, LASER achieves a score of 55.7 on the ScreenSpot-Pro benchmark, establishing a new state-of-the-art (SoTA) among 7B-scale models.

33. MAGneT: Coordinated Multi-Agent Generation of Synthetic Multi-Turn Mental Health Counseling Sessions

Authors: Aishik Mandal , Tanmoy Chakraborty , Iryna Gurevych
URL: https://arxiv.org/abs/2509.04183
Abstract:

The growing demand for scalable psychological counseling highlights the need for fine-tuning open-source Large Language Models (LLMs) with high-quality, privacy-compliant data, yet such data remains scarce. Here we introduce MAGneT, a novel multi-agent framework for synthetic psychological counseling session generation that decomposes counselor response generation into coordinated sub-tasks handled by specialized LLM agents, each modeling a key psychological technique. Unlike prior single-agent approaches, MAGneT better captures the structure and nuance of real counseling. In addition, we address inconsistencies in prior evaluation protocols by proposing a unified evaluation framework integrating diverse automatic and expert metrics. Furthermore, we expand the expert evaluations from four aspects of counseling in previous works to nine aspects, enabling a more thorough and robust assessment of data quality. Empirical results show that MAGneT significantly outperforms existing methods in quality, diversity, and therapeutic alignment of the generated counseling sessions, improving general counseling skills by 3.2% and CBT-specific skills by 4.3% on average on cognitive therapy rating scale (CTRS). Crucially, experts prefer MAGneT-generated sessions in 77.2% of cases on average across all aspects. Moreover, fine-tuning an open-source model on MAGneT-generated sessions shows better performance, with improvements of 6.3% on general counseling skills and 7.3% on CBT-specific skills on average on CTRS over those fine-tuned with sessions generated by baseline methods. We also make our code and data public.

34. TAGAL: Tabular Data Generation using Agentic LLM Methods

Authors: Benoît Ronval , Pierre Dupont , Siegfried Nijssen
URL: https://arxiv.org/abs/2509.04152
Abstract:

The generation of data is a common approach to improve the performance of machine learning tasks, among which is the training of models for classification. In this paper, we present TAGAL, a collection of methods able to generate synthetic tabular data using an agentic workflow. The methods leverage Large Language Models (LLMs) for an automatic and iterative process that uses feedback to improve the generated data without any further LLM training. The use of LLMs also allows for the addition of external knowledge in the generation process. We evaluate TAGAL across diverse datasets and different aspects of quality for the generated data. We look at the utility of downstream ML models, both by training classifiers on synthetic data only and by combining real and synthetic data. Moreover, we compare the similarities between the real and the generated data. We show that TAGAL is able to perform on par with state-of-the-art approaches that require LLM training and generally outperforms other training-free approaches. These findings highlight the potential of agentic workflow and open new directions for LLM-based data generation methods.

35. Enhancing Technical Documents Retrieval for RAG

Authors: Songjiang Lai , Tsun-Hin Cheung , Ka-Chun Fung , Kaiwen Xue , Kwan-Ho Lin , Yan-Ming Choi , Vincent Ng , Kin-Man Lam
URL: https://arxiv.org/abs/2509.04139
Abstract:

In this paper, we introduce Technical-Embeddings, a novel framework designed to optimize semantic retrieval in technical documentation, with applications in both hardware and software development. Our approach addresses the challenges of understanding and retrieving complex technical content by leveraging the capabilities of Large Language Models (LLMs). First, we enhance user queries by generating expanded representations that better capture user intent and improve dataset diversity, thereby enriching the fine-tuning process for embedding models. Second, we apply summary extraction techniques to encode essential contextual information, refining the representation of technical documents. To further enhance retrieval performance, we fine-tune a bi-encoder BERT model using soft prompting, incorporating separate learning parameters for queries and document context to capture fine-grained semantic nuances. We evaluate our approach on two public datasets, RAG-EDA and Rust-Docs-QA, demonstrating that Technical-Embeddings significantly outperforms baseline models in both precision and recall. Our findings highlight the effectiveness of integrating query expansion and contextual summarization to enhance information access and comprehension in technical domains. This work advances the state of Retrieval-Augmented Generation (RAG) systems, offering new avenues for efficient and accurate technical document retrieval in engineering and product development workflows.

36. MEPG:Multi-Expert Planning and Generation for Compositionally-Rich Image Generation

Authors: Yuan Zhao , Liu Lin
URL: https://arxiv.org/abs/2509.04126
Abstract:

Text-to-image diffusion models have achieved remarkable image quality, but they still struggle with complex, multiele ment prompts, and limited stylistic diversity. To address these limitations, we propose a Multi-Expert Planning and Gen eration Framework (MEPG) that synergistically integrates position- and style-aware large language models (LLMs) with spatial-semantic expert modules. The framework comprises two core components: (1) a Position-Style-Aware (PSA) module that utilizes a supervised fine-tuned LLM to decom pose input prompts into precise spatial coordinates and style encoded semantic instructions; and (2) a Multi-Expert Dif fusion (MED) module that implements cross-region genera tion through dynamic expert routing across both local regions and global areas. During the generation process for each lo cal region, specialized models (e.g., realism experts, styliza tion specialists) are selectively activated for each spatial par tition via attention-based gating mechanisms. The architec ture supports lightweight integration and replacement of ex pert models, providing strong extensibility. Additionally, an interactive interface enables real-time spatial layout editing and per-region style selection from a portfolio of experts. Ex periments show that MEPG significantly outperforms base line models with the same backbone in both image quality and style diversity.

37. RepoDebug: Repository-Level Multi-Task and Multi-Language Debugging Evaluation of Large Language Models

Authors: Jingjing Liu , Zeming Liu , Zihao Cheng , Mengliang He , Xiaoming Shi , Yuhang Guo , Xiangrong Zhu , Yuanfang Guo , Yunhong Wang , Haifeng Wang
URL: https://arxiv.org/abs/2509.04078
Abstract:

Large Language Models (LLMs) have exhibited significant proficiency in code debugging, especially in automatic program repair, which may substantially reduce the time consumption of developers and enhance their efficiency. Significant advancements in debugging datasets have been made to promote the development of code debugging. However, these datasets primarily focus on assessing the LLM’s function-level code repair capabilities, neglecting the more complex and realistic repository-level scenarios, which leads to an incomplete understanding of the LLM’s challenges in repository-level debugging. While several repository-level datasets have been proposed, they often suffer from limitations such as limited diversity of tasks, languages, and error types. To mitigate this challenge, this paper introduces RepoDebug, a multi-task and multi-language repository-level code debugging dataset with 22 subtypes of errors that supports 8 commonly used programming languages and 3 debugging tasks. Furthermore, we conduct evaluation experiments on 10 LLMs, where Claude 3.5 Sonnect, the best-performing model, still cannot perform well in repository-level debugging.

38. On Robustness and Reliability of Benchmark-Based Evaluation of LLMs

Authors: Riccardo Lunardi , Vincenzo Della Mea , Stefano Mizzaro , Kevin Roitero
URL: https://arxiv.org/abs/2509.04013
Abstract:

Large Language Models (LLMs) effectiveness is usually evaluated by means of benchmarks such as MMLU, ARC-C, or HellaSwag, where questions are presented in their original wording, thus in a fixed, standardized format. However, real-world applications involve linguistic variability, requiring models to maintain their effectiveness across diverse rewordings of the same question or query. In this study, we systematically assess the robustness of LLMs to paraphrased benchmark questions and investigate whether benchmark-based evaluations provide a reliable measure of model capabilities. We systematically generate various paraphrases of all the questions across six different common benchmarks, and measure the resulting variations in effectiveness of 34 state-of-the-art LLMs, of different size and effectiveness. Our findings reveal that while LLM rankings remain relatively stable across paraphrased inputs, absolute effectiveness scores change, and decline significantly. This suggests that LLMs struggle with linguistic variability, raising concerns about their generalization abilities and evaluation methodologies. Furthermore, the observed performance drop challenges the reliability of benchmark-based evaluations, indicating that high benchmark scores may not fully capture a model’s robustness to real-world input variations. We discuss the implications of these findings for LLM evaluation methodologies, emphasizing the need for robustness-aware benchmarks that better reflect practical deployment scenarios.

39. NER Retriever: Zero-Shot Named Entity Retrieval with Type-Aware Embeddings

Authors: Or Shachar , Uri Katz , Yoav Goldberg , Oren Glickman
URL: https://arxiv.org/abs/2509.04011
Abstract:

We present NER Retriever, a zero-shot retrieval framework for ad-hoc Named Entity Retrieval, a variant of Named Entity Recognition (NER), where the types of interest are not provided in advance, and a user-defined type description is used to retrieve documents mentioning entities of that type. Instead of relying on fixed schemas or fine-tuned models, our method builds on internal representations of large language models (LLMs) to embed both entity mentions and user-provided open-ended type descriptions into a shared semantic space. We show that internal representations, specifically the value vectors from mid-layer transformer blocks, encode fine-grained type information more effectively than commonly used top-layer embeddings. To refine these representations, we train a lightweight contrastive projection network that aligns type-compatible entities while separating unrelated types. The resulting entity embeddings are compact, type-aware, and well-suited for nearest-neighbor search. Evaluated on three benchmarks, NER Retriever significantly outperforms both lexical and dense sentence-level retrieval baselines. Our findings provide empirical support for representation selection within LLMs and demonstrate a practical solution for scalable, schema-free entity retrieval. The NER Retriever Codebase is publicly available at this https URL

40. RTQA : Recursive Thinking for Complex Temporal Knowledge Graph Question Answering with Large Language Models

Authors: Zhaoyan Gong , Juan Li , Zhiqiang Liu , Lei Liang , Huajun Chen , Wen Zhang
URL: https://arxiv.org/abs/2509.03995
Abstract:

Current temporal knowledge graph question answering (TKGQA) methods primarily focus on implicit temporal constraints, lacking the capability of handling more complex temporal queries, and struggle with limited reasoning abilities and error propagation in decomposition frameworks. We propose RTQA, a novel framework to address these challenges by enhancing reasoning over TKGs without requiring training. Following recursive thinking, RTQA recursively decomposes questions into sub-problems, solves them bottom-up using LLMs and TKG knowledge, and employs multi-path answer aggregation to improve fault tolerance. RTQA consists of three core components: the Temporal Question Decomposer, the Recursive Solver, and the Answer Aggregator. Experiments on MultiTQ and TimelineKGQA benchmarks demonstrate significant Hits@1 improvements in “Multiple” and “Complex” categories, outperforming state-of-the-art methods. Our code and data are available at this https URL .

41. NeuroBreak: Unveil Internal Jailbreak Mechanisms in Large Language Models

Authors: Chuhan Zhang , Ye Zhang , Bowen Shi , Yuyou Gan , Tianyu Du , Shouling Ji , Dazhan Deng , Yingcai Wu
URL: https://arxiv.org/abs/2509.03985
Abstract:

In deployment and application, large language models (LLMs) typically undergo safety alignment to prevent illegal and unethical outputs. However, the continuous advancement of jailbreak attack techniques, designed to bypass safety mechanisms with adversarial prompts, has placed increasing pressure on the security defenses of LLMs. Strengthening resistance to jailbreak attacks requires an in-depth understanding of the security mechanisms and vulnerabilities of LLMs. However, the vast number of parameters and complex structure of LLMs make analyzing security weaknesses from an internal perspective a challenging task. This paper presents NeuroBreak, a top-down jailbreak analysis system designed to analyze neuron-level safety mechanisms and mitigate vulnerabilities. We carefully design system requirements through collaboration with three experts in the field of AI security. The system provides a comprehensive analysis of various jailbreak attack methods. By incorporating layer-wise representation probing analysis, NeuroBreak offers a novel perspective on the model’s decision-making process throughout its generation steps. Furthermore, the system supports the analysis of critical neurons from both semantic and functional perspectives, facilitating a deeper exploration of security mechanisms. We conduct quantitative evaluations and case studies to verify the effectiveness of our system, offering mechanistic insights for developing next-generation defense strategies against evolving jailbreak attacks.

42. Expanding Foundational Language Capabilities in Open-Source LLMs through a Korean Case Study

Authors: Junghwan Lim , Gangwon Jo , Sungmin Lee , Jiyoung Park , Dongseok Kim , Jihwan Kim , Junhyeok Lee , Wai Ting Cheung , Dahye Choi , Kibong Choi , Jaeyeon Huh , Beomgyu Kim , Jangwoong Kim , Taehyun Kim , Haesol Lee , Jeesoo Lee , Dongpin Oh , Changseok Song , Daewon Suh
URL: https://arxiv.org/abs/2509.03972
Abstract:

We introduce Llama-3-Motif, a language model consisting of 102 billion parameters, specifically designed to enhance Korean capabilities while retaining strong performance in English. Developed on the Llama 3 architecture, Llama-3-Motif employs advanced training techniques, including LlamaPro and Masked Structure Growth, to effectively scale the model without altering its core Transformer architecture. Using the MoAI platform for efficient training across hyperscale GPU clusters, we optimized Llama-3-Motif using a carefully curated dataset that maintains a balanced ratio of Korean and English data. Llama-3-Motif shows decent performance on Korean-specific benchmarks, outperforming existing models and achieving results comparable to GPT-4.

43. Multimodal Feature Fusion Network with Text Difference Enhancement for Remote Sensing Change Detection

Authors: Yijun Zhou , Yikui Zhai , Zilu Ying , Tingfeng Xian , Wenlve Zhou , Zhiheng Zhou , Xiaolin Tian , Xudong Jia , Hongsheng Zhang , C. L. Philip Chen
URL: https://arxiv.org/abs/2509.03961
Abstract:

Although deep learning has advanced remote sensing change detection (RSCD), most methods rely solely on image modality, limiting feature representation, change pattern modeling, and generalization especially under illumination and noise disturbances. To address this, we propose MMChange, a multimodal RSCD method that combines image and text modalities to enhance accuracy and robustness. An Image Feature Refinement (IFR) module is introduced to highlight key regions and suppress environmental noise. To overcome the semantic limitations of image features, we employ a vision language model (VLM) to generate semantic descriptions of bitemporal images. A Textual Difference Enhancement (TDE) module then captures fine grained semantic shifts, guiding the model toward meaningful changes. To bridge the heterogeneity between modalities, we design an Image Text Feature Fusion (ITFF) module that enables deep cross modal integration. Extensive experiments on LEVIRCD, WHUCD, and SYSUCD demonstrate that MMChange consistently surpasses state of the art methods across multiple metrics, validating its effectiveness for multimodal RSCD. Code is available at: this https URL .

44. CANDY: Benchmarking LLMs’ Limitations and Assistive Potential in Chinese Misinformation Fact-Checking

Authors: Ruiling Guo , Xinwei Yang , Chen Huang , Tong Zhang , Yong Hu
URL: https://arxiv.org/abs/2509.03957
Abstract:

The effectiveness of large language models (LLMs) to fact-check misinformation remains uncertain, despite their growing use. To this end, we present CANDY, a benchmark designed to systematically evaluate the capabilities and limitations of LLMs in fact-checking Chinese misinformation. Specifically, we curate a carefully annotated dataset of ~20k instances. Our analysis shows that current LLMs exhibit limitations in generating accurate fact-checking conclusions, even when enhanced with chain-of-thought reasoning and few-shot prompting. To understand these limitations, we develop a taxonomy to categorize flawed LLM-generated explanations for their conclusions and identify factual fabrication as the most common failure mode. Although LLMs alone are unreliable for fact-checking, our findings indicate their considerable potential to augment human performance when deployed as assistive tools in scenarios. Our dataset and code can be accessed at this https URL

45. VoxRole: A Comprehensive Benchmark for Evaluating Speech-Based Role-Playing Agents

Authors: Weihao Wu , Liang Cao , Xinyu Wu , Zhiwei Lin , Rui Niu , Jingbei Li , Zhiyong Wu
URL: https://arxiv.org/abs/2509.03940
Abstract:

Recent significant advancements in Large Language Models (LLMs) have greatly propelled the development of Role-Playing Conversational Agents (RPCAs). These systems aim to create immersive user experiences through consistent persona adoption. However, current RPCA research faces dual limitations. First, existing work predominantly focuses on the textual modality, entirely overlooking critical paralinguistic features including intonation, prosody, and rhythm in speech, which are essential for conveying character emotions and shaping vivid identities. Second, the speech-based role-playing domain suffers from a long-standing lack of standardized evaluation benchmarks. Most current spoken dialogue datasets target only fundamental capability assessments, featuring thinly sketched or ill-defined character profiles. Consequently, they fail to effectively quantify model performance on core competencies like long-term persona consistency. To address this critical gap, we introduce VoxRole, the first comprehensive benchmark specifically designed for the evaluation of speech-based RPCAs. The benchmark comprises 13335 multi-turn dialogues, totaling 65.6 hours of speech from 1228 unique characters across 261 movies. To construct this resource, we propose a novel two-stage automated pipeline that first aligns movie audio with scripts and subsequently employs an LLM to systematically build multi-dimensional profiles for each character. Leveraging VoxRole, we conduct a multi-dimensional evaluation of contemporary spoken dialogue models, revealing crucial insights into their respective strengths and limitations in maintaining persona consistency.

46. SPFT-SQL: Enhancing Large Language Model for Text-to-SQL Parsing by Self-Play Fine-Tuning

Authors: Yuhao Zhang , Shaoming Duan , Jinhang Su , Chuanyi Liu , Peiyi Han
URL: https://arxiv.org/abs/2509.03937
Abstract:

Despite the significant advancements of self-play fine-tuning (SPIN), which can transform a weak large language model (LLM) into a strong one through competitive interactions between models of varying capabilities, it still faces challenges in the Text-to-SQL task. SPIN does not generate new information, and the large number of correct SQL queries produced by the opponent model during self-play reduces the main model’s ability to generate accurate SQL queries. To address this challenge, we propose a new self-play fine-tuning method tailored for the Text-to-SQL task, called SPFT-SQL. Prior to self-play, we introduce a verification-based iterative fine-tuning approach, which synthesizes high-quality fine-tuning data iteratively based on the database schema and validation feedback to enhance model performance, while building a model base with varying capabilities. During the self-play fine-tuning phase, we propose an error-driven loss method that incentivizes incorrect outputs from the opponent model, enabling the main model to distinguish between correct SQL and erroneous SQL generated by the opponent model, thereby improving its ability to generate correct SQL. Extensive experiments and in-depth analyses on six open-source LLMs and five widely used benchmarks demonstrate that our approach outperforms existing state-of-the-art (SOTA) methods.

47. SelfAug: Mitigating Catastrophic Forgetting in Retrieval-Augmented Generation via Distribution Self-Alignment

Authors: Yuqing Huang , Rongyang Zhang , Qimeng Wang , Chengqiang Lu , Yan Gao , Yi Wu , Yao Hu , Xuyang Zhi , Guiquan Liu , Xin Li , Hao Wang , Enhong Chen
URL: https://arxiv.org/abs/2509.03934
Abstract:

Recent advancements in large language models (LLMs) have revolutionized natural language processing through their remarkable capabilities in understanding and executing diverse tasks. While supervised fine-tuning, particularly in Retrieval-Augmented Generation (RAG) scenarios, effectively enhances task-specific performance, it often leads to catastrophic forgetting, where models lose their previously acquired knowledge and general capabilities. Existing solutions either require access to general instruction data or face limitations in preserving the model’s original distribution. To overcome these limitations, we propose SelfAug, a self-distribution alignment method that aligns input sequence logits to preserve the model’s semantic distribution, thereby mitigating catastrophic forgetting and improving downstream performance. Extensive experiments demonstrate that SelfAug achieves a superior balance between downstream learning and general capability retention. Our comprehensive empirical analysis reveals a direct correlation between distribution shifts and the severity of catastrophic forgetting in RAG scenarios, highlighting how the absence of RAG capabilities in general instruction tuning leads to significant distribution shifts during fine-tuning. Our findings not only advance the understanding of catastrophic forgetting in RAG contexts but also provide a practical solution applicable across diverse fine-tuning scenarios. Our code is publicly available at this https URL .

48. MTQA:Matrix of Thought for Enhanced Reasoning in Complex Question Answering

Authors: Fengxiao Tang , Yufeng Li , Zongzong Wu , Ming Zhao
URL: https://arxiv.org/abs/2509.03918
Abstract:

Complex Question Answering (QA) is a fundamental and challenging task in NLP. While large language models (LLMs) exhibit impressive performance in QA, they suffer from significant performance degradation when facing complex and abstract QA tasks due to insufficient reasoning capabilities. Works such as Chain-of-Thought (CoT) and Tree-of-Thought (ToT) aim to enhance LLMs’ reasoning abilities, but they face issues such as in-layer redundancy in tree structures and single paths in chain structures. Although some studies utilize Retrieval-Augmented Generation (RAG) methods to assist LLMs in reasoning, the challenge of effectively utilizing large amounts of information involving multiple entities and hops remains critical. To address this, we propose the Matrix of Thought (MoT), a novel and efficient LLM thought structure. MoT explores the problem in both horizontal and vertical dimensions through the “column-cell communication” mechanism, enabling LLMs to actively engage in multi-strategy and deep-level thinking, reducing redundancy within the column cells and enhancing reasoning capabilities. Furthermore, we develop a fact-correction mechanism by constructing knowledge units from retrieved knowledge graph triples and raw text to enhance the initial knowledge for LLM reasoning and correct erroneous answers. This leads to the development of an efficient and accurate QA framework (MTQA). Experimental results show that our framework outperforms state-of-the-art methods on four widely-used datasets in terms of F1 and EM scores, with reasoning time only 14.4\% of the baseline methods, demonstrating both its efficiency and accuracy. The code for this framework is available at this https URL .

49. A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models

Authors: Yanbo Wang , Yongcan Yu , Jian Liang , Ran He
URL: https://arxiv.org/abs/2509.03871
Abstract:

The development of Long-CoT reasoning has advanced LLM performance across various tasks, including language understanding, complex problem solving, and code generation. This paradigm enables models to generate intermediate reasoning steps, thereby improving both accuracy and interpretability. However, despite these advancements, a comprehensive understanding of how CoT-based reasoning affects the trustworthiness of language models remains underdeveloped. In this paper, we survey recent work on reasoning models and CoT techniques, focusing on five core dimensions of trustworthy reasoning: truthfulness, safety, robustness, fairness, and privacy. For each aspect, we provide a clear and structured overview of recent studies in chronological order, along with detailed analyses of their methodologies, findings, and limitations. Future research directions are also appended at the end for reference and discussion. Overall, while reasoning techniques hold promise for enhancing model trustworthiness through hallucination mitigation, harmful content detection, and robustness improvement, cutting-edge reasoning models themselves often suffer from comparable or even greater vulnerabilities in safety, robustness, and privacy. By synthesizing these insights, we hope this work serves as a valuable and timely resource for the AI safety community to stay informed on the latest progress in reasoning trustworthiness. A full list of related papers can be found at \href{ this https URL }{ this https URL }.

50. INGRID: Intelligent Generative Robotic Design Using Large Language Models

Authors: Guanglu Jia , Ceng Zhang , Gregory S. Chirikjian
URL: https://arxiv.org/abs/2509.03842
Abstract:

The integration of large language models (LLMs) into robotic systems has accelerated progress in embodied artificial intelligence, yet current approaches remain constrained by existing robotic architectures, particularly serial mechanisms. This hardware dependency fundamentally limits the scope of robotic intelligence. Here, we present INGRID (Intelligent Generative Robotic Design), a framework that enables the automated design of parallel robotic mechanisms through deep integration with reciprocal screw theory and kinematic synthesis methods. We decompose the design challenge into four progressive tasks: constraint analysis, kinematic joint generation, chain construction, and complete mechanism design. INGRID demonstrates the ability to generate novel parallel mechanisms with both fixed and variable mobility, discovering kinematic configurations not previously documented in the literature. We validate our approach through three case studies demonstrating how INGRID assists users in designing task-specific parallel robots based on desired mobility requirements. By bridging the gap between mechanism theory and machine learning, INGRID enables researchers without specialized robotics training to create custom parallel mechanisms, thereby decoupling advances in robotic intelligence from hardware constraints. This work establishes a foundation for mechanism intelligence, where AI systems actively design robotic hardware, potentially transforming the development of embodied AI systems.

51. Gravity Well Echo Chamber Modeling With An LLM-Based Confirmation Bias Model

Authors: Joseph Jackson , Georgiy Lapin , Jeremy E. Thompson
URL: https://arxiv.org/abs/2509.03832
Abstract:

Social media echo chambers play a central role in the spread of misinformation, yet existing models often overlook the influence of individual confirmation bias. An existing model of echo chambers is the “gravity well” model, which creates an analog between echo chambers and spatial gravity wells. We extend this established model by introducing a dynamic confirmation bias variable that adjusts the strength of pull based on a user’s susceptibility to belief-reinforcing content. This variable is calculated for each user through comparisons between their posting history and their responses to posts of a wide range of viewpoints. Incorporating this factor produces a confirmation-bias-integrated gravity well model that more accurately identifies echo chambers and reveals community-level markers of information health. We validated the approach on nineteen Reddit communities, demonstrating improved detection of echo chambers. Our contribution is a framework for systematically capturing the role of confirmation bias in online group dynamics, enabling more effective identification of echo chambers. By flagging these high-risk environments, the model supports efforts to curb the spread of misinformation at its most common points of amplification.

52. Align-then-Slide: A complete evaluation framework for Ultra-Long Document-Level Machine Translation

Authors: Jiaxin Guo , Daimeng Wei , Yuanchang Luo , Xiaoyu Chen , Zhanglin Wu , Huan Yang , Hengchao Shang , Zongyao Li , Zhiqiang Rao , Jinlong Yang , Hao Yang
URL: https://arxiv.org/abs/2509.03809
Abstract:

Large language models (LLMs) have ushered in a new era for document-level machine translation (\textit{doc}-mt), yet their whole-document outputs challenge existing evaluation methods that assume sentence-by-sentence alignment. We introduce \textit{\textbf{Align-then-Slide}}, a complete evaluation framework for ultra-long doc-mt. In the Align stage, we automatically infer sentence-level source-target correspondences and rebuild the target to match the source sentence number, resolving omissions and many-to-one/one-to-many mappings. In the n-Chunk Sliding Evaluate stage, we calculate averaged metric scores under 1-, 2-, 3- and 4-chunk for multi-granularity assessment. Experiments on the WMT benchmark show a Pearson correlation of 0.929 between our method with expert MQM rankings. On a newly curated real-world test set, our method again aligns closely with human judgments. Furthermore, preference data produced by Align-then-Slide enables effective CPO training and its direct use as a reward model for GRPO, both yielding translations preferred over a vanilla SFT baseline. The results validate our framework as an accurate, robust, and actionable evaluation tool for doc-mt systems.

53. Measuring How (Not Just Whether) VLMs Build Common Ground

Authors: Saki Imai , Mert İnan , Anthony Sicilia , Malihe Alikhani
URL: https://arxiv.org/abs/2509.03805
Abstract:

Large vision language models (VLMs) increasingly claim reasoning skills, yet current benchmarks evaluate them in single-turn or question answering settings. However, grounding is an interactive process in which people gradually develop shared understanding through ongoing communication. We introduce a four-metric suite (grounding efficiency, content alignment, lexical adaptation, and human-likeness) to systematically evaluate VLM performance in interactive grounding contexts. We deploy the suite on 150 self-play sessions of interactive referential games between three proprietary VLMs and compare them with human dyads. All three models diverge from human patterns on at least three metrics, while GPT4o-mini is the closest overall. We find that (i) task success scores do not indicate successful grounding and (ii) high image-utterance alignment does not necessarily predict task success. Our metric suite and findings offer a framework for future research on VLM grounding.

54. SAMVAD: A Multi-Agent System for Simulating Judicial Deliberation Dynamics in India

Authors: Prathamesh Devadiga , Omkaar Jayadev Shetty , Pooja Agarwal
URL: https://arxiv.org/abs/2509.03793
Abstract:

Understanding the complexities of judicial deliberation is crucial for assessing the efficacy and fairness of a justice system. However, empirical studies of judicial panels are constrained by significant ethical and practical barriers. This paper introduces SAMVAD, an innovative Multi-Agent System (MAS) designed to simulate the deliberation process within the framework of the Indian justice system. Our system comprises agents representing key judicial roles: a Judge, a Prosecution Counsel, a Defense Counsel, and multiple Adjudicators (simulating a judicial bench), all powered by large language models (LLMs). A primary contribution of this work is the integration of Retrieval-Augmented Generation (RAG), grounded in a domain-specific knowledge base of landmark Indian legal documents, including the Indian Penal Code and the Constitution of India. This RAG functionality enables the Judge and Counsel agents to generate legally sound instructions and arguments, complete with source citations, thereby enhancing both the fidelity and transparency of the simulation. The Adjudicator agents engage in iterative deliberation rounds, processing case facts, legal instructions, and arguments to reach a consensus-based verdict. We detail the system architecture, agent communication protocols, the RAG pipeline, the simulation workflow, and a comprehensive evaluation plan designed to assess performance, deliberation quality, and outcome consistency. This work provides a configurable and explainable MAS platform for exploring legal reasoning and group decision-making dynamics in judicial simulations, specifically tailored to the Indian legal context and augmented with verifiable legal grounding via RAG.

55. Designing Gaze Analytics for ELA Instruction: A User-Centered Dashboard with Conversational AI Support

Authors: Eduardo Davalos , Yike Zhang , Shruti Jain , Namrata Srivastava , Trieu Truong , Nafees-ul Haque , Tristan Van , Jorge Salas , Sara McFadden , Sun-Joo Cho , Gautam Biswas , Amanda Goodwin
URL: https://arxiv.org/abs/2509.03741
Abstract:

Eye-tracking offers rich insights into student cognition and engagement, but remains underutilized in classroom-facing educational technology due to challenges in data interpretation and accessibility. In this paper, we present the iterative design and evaluation of a gaze-based learning analytics dashboard for English Language Arts (ELA), developed through five studies involving teachers and students. Guided by user-centered design and data storytelling principles, we explored how gaze data can support reflection, formative assessment, and instructional decision-making. Our findings demonstrate that gaze analytics can be approachable and pedagogically valuable when supported by familiar visualizations, layered explanations, and narrative scaffolds. We further show how a conversational agent, powered by a large language model (LLM), can lower cognitive barriers to interpreting gaze data by enabling natural language interactions with multimodal learning analytics. We conclude with design implications for future EdTech systems that aim to integrate novel data modalities in classroom contexts.

56. Breaking the Mirror: Activation-Based Mitigation of Self-Preference in LLM Evaluators

Authors: Dani Roytburg , Matthew Bozoukov , Matthew Nguyen , Jou Barzdukas , Simon Fu , Narmeen Oozeer
URL: https://arxiv.org/abs/2509.03647
Abstract:

Large language models (LLMs) increasingly serve as automated evaluators, yet they suffer from “self-preference bias”: a tendency to favor their own outputs over those of other models. This bias undermines fairness and reliability in evaluation pipelines, particularly for tasks like preference tuning and model routing. We investigate whether lightweight steering vectors can mitigate this problem at inference time without retraining. We introduce a curated dataset that distinguishes self-preference bias into justified examples of self-preference and unjustified examples of self-preference, and we construct steering vectors using two methods: Contrastive Activation Addition (CAA) and an optimization-based approach. Our results show that steering vectors can reduce unjustified self-preference bias by up to 97\%, substantially outperforming prompting and direct preference optimization baselines. Yet steering vectors are unstable on legitimate self-preference and unbiased agreement, implying self-preference spans multiple or nonlinear directions. This underscores both their promise and limits as safeguards for LLM-as-judges and motivates more robust interventions.

57. E-ARMOR: Edge case Assessment and Review of Multilingual Optical Character Recognition

Authors: Aryan Gupta , Anupam Purwar
URL: https://arxiv.org/abs/2509.03615
Abstract:

Optical Character Recognition (OCR) in multilingual, noisy, and diverse real-world images remains a significant challenge for optical character recognition systems. With the rise of Large Vision-Language Models (LVLMs), there is growing interest in their ability to generalize and reason beyond fixed OCR pipelines. In this work, we introduce Sprinklr-Edge-OCR, a novel OCR system built specifically optimized for edge deployment in resource-constrained environments. We present a large-scale comparative evaluation of five state-of-the-art LVLMs (InternVL, Qwen, GOT OCR, LLaMA, MiniCPM) and two traditional OCR systems (Sprinklr-Edge-OCR, SuryaOCR) on a proprietary, doubly hand annotated dataset of multilingual (54 languages) images. Our benchmark covers a broad range of metrics including accuracy, semantic consistency, language coverage, computational efficiency (latency, memory, GPU usage), and deployment cost. To better reflect real-world applicability, we also conducted edge case deployment analysis, evaluating model performance on CPU only environments. Among the results, Qwen achieved the highest precision (0.54), while Sprinklr-Edge-OCR delivered the best overall F1 score (0.46) and outperformed others in efficiency, processing images 35 faster (0.17 seconds per image on average) and at less than 0.01 of the cost (0.006 USD per 1,000 images) compared to LVLM. Our findings demonstrate that the most optimal OCR systems for edge deployment are the traditional ones even in the era of LLMs due to their low compute requirements, low latency, and very high affordability.

58. Improving Factuality in LLMs via Inference-Time Knowledge Graph Construction

Authors: Shanglin Wu , Lihui Liu , Jinho D. Choi , Kai Shu
URL: https://arxiv.org/abs/2509.03540
Abstract:

Large Language Models (LLMs) often struggle with producing factually consistent answers due to limitations in their parametric memory. Retrieval-Augmented Generation (RAG) methods address this issue by incorporating external knowledge from trusted sources at inference time. However, such methods typically treat knowledge as unstructured text, which limits their ability to support compositional reasoning and identify factual inconsistencies. To overcome these limitations, we propose a novel framework that dynamically constructs and expands knowledge graphs (KGs) during inference, integrating both internal knowledge extracted from LLMs and external information retrieved from external sources. Our method begins by extracting a seed KG from the question via prompting, followed by iterative expansion using the LLM’s latent knowledge. The graph is then selectively refined through external retrieval, enhancing factual coverage and correcting inaccuracies. We evaluate our approach on three diverse factual QA benchmarks, demonstrating consistent improvements in factual accuracy, answer precision, and interpretability over baseline prompting and static KG-augmented methods. Our findings suggest that inference-time KG construction is a promising direction for enhancing LLM factuality in a structured, interpretable, and scalable manner.

59. AR$^2$: Adversarial Reinforcement Learning for Abstract Reasoning in Large Language Models

Authors: Cheng-Kai Yeh , Hsing-Wang Lee , Chung-Hung Kuo , Hen-Hsen Huang
URL: https://arxiv.org/abs/2509.03537
Abstract:

Abstraction–the ability to recognize and distill essential computational patterns from complex problem statements–is a foundational skill in computer science, critical both for human problem-solvers and coding-oriented large language models (LLMs). Despite recent advances in training LLMs for code generation using reinforcement learning (RL), most existing approaches focus primarily on superficial pattern recognition, overlooking explicit training for abstraction. In this study, we propose AR$^2$ (Adversarial Reinforcement Learning for Abstract Reasoning), a novel framework explicitly designed to enhance the abstraction abilities of LLMs. AR$^2$ employs a teacher model to transform kernel problems into narrative-rich, challenging descriptions without changing their fundamental logic. Simultaneously, a student coding model is trained to solve these complex narrative problems by extracting their underlying computational kernels. Experimental results demonstrate that AR$^2$ substantially improves the student model’s accuracy on previously unseen, challenging programming tasks, underscoring abstraction as a key skill for enhancing LLM generalization.

60. Real-Time Detection of Hallucinated Entities in Long-Form Generation

Authors: Oscar Obeso , Andy Arditi , Javier Ferrando , Joshua Freeman , Cameron Holmes , Neel Nanda
URL: https://arxiv.org/abs/2509.03531
Abstract:

Large language models are now routinely used in high-stakes applications where hallucinations can cause serious harm, such as medical consultations or legal advice. Existing hallucination detection methods, however, are impractical for real-world use, as they are either limited to short factual queries or require costly external verification. We present a cheap, scalable method for real-time identification of hallucinated tokens in long-form generations, and scale it effectively to 70B parameter models. Our approach targets \emph{entity-level hallucinations} – e.g., fabricated names, dates, citations – rather than claim-level, thereby naturally mapping to token-level labels and enabling streaming detection. We develop an annotation methodology that leverages web search to annotate model responses with grounded labels indicating which tokens correspond to fabricated entities. This dataset enables us to train effective hallucination classifiers with simple and efficient methods such as linear probes. Evaluating across four model families, our classifiers consistently outperform baselines on long-form responses, including more expensive methods such as semantic entropy (e.g., AUC 0.90 vs 0.71 for Llama-3.3-70B), and are also an improvement in short-form question-answering settings. Moreover, despite being trained only with entity-level labels, our probes effectively detect incorrect answers in mathematical reasoning tasks, indicating generalization beyond entities. While our annotation methodology is expensive, we find that annotated responses from one model can be used to train effective classifiers on other models; accordingly, we publicly release our datasets to facilitate reuse. Overall, our work suggests a promising new approach for scalable, real-world hallucination detection.

61. Multilevel Analysis of Cryptocurrency News using RAG Approach with Fine-Tuned Mistral Large Language Model

Authors: Bohdan M. Pavlyshenko
URL: https://arxiv.org/abs/2509.03527
Abstract:

In the paper, we consider multilevel multitask analysis of cryptocurrency news using a fine-tuned Mistral 7B large language model with retrieval-augmented generation (RAG). On the first level of analytics, the fine-tuned model generates graph and text summaries with sentiment scores as well as JSON representations of summaries. Higher levels perform hierarchical stacking that consolidates sets of graph-based and text-based summaries as well as summaries of summaries into comprehensive reports. The combination of graph and text summaries provides complementary views of cryptocurrency news. The model is fine-tuned with 4-bit quantization using the PEFT/LoRA approach. The representation of cryptocurrency news as knowledge graph can essentially eliminate problems with large language model hallucinations. The obtained results demonstrate that the use of fine-tuned Mistral 7B LLM models for multilevel cryptocurrency news analysis can conduct informative qualitative and quantitative analytics, providing important insights.

62. Speech-Based Cognitive Screening: A Systematic Evaluation of LLM Adaptation Strategies

Authors: Fatemeh Taherinezhad , Mohamad Javad Momeni Nezhad , Sepehr Karimi , Sina Rashidi , Ali Zolnour , Maryam Dadkhah , Yasaman Haghbin , Hossein AzadMaleki , Maryam Zolnoori
URL: https://arxiv.org/abs/2509.03525
Abstract:

Over half of US adults with Alzheimer disease and related dementias remain undiagnosed, and speech-based screening offers a scalable detection approach. We compared large language model adaptation strategies for dementia detection using the DementiaBank speech corpus, evaluating nine text-only models and three multimodal audio-text models on recordings from DementiaBank speech corpus. Adaptations included in-context learning with different demonstration selection policies, reasoning-augmented prompting, parameter-efficient fine-tuning, and multimodal integration. Results showed that class-centroid demonstrations achieved the highest in-context learning performance, reasoning improved smaller models, and token-level fine-tuning generally produced the best scores. Adding a classification head substantially improved underperforming models. Among multimodal models, fine-tuned audio-text systems performed well but did not surpass the top text-only models. These findings highlight that model adaptation strategies, including demonstration selection, reasoning design, and tuning method, critically influence speech-based dementia detection, and that properly adapted open-weight models can match or exceed commercial systems.

LLM 관련 주요 논문 - 2025-09-05

1. ArcMemo: Abstract Reasoning Composition with Lifelong LLM Memory

2. Psychologically Enhanced AI Agents

3. EvoEmo: Towards Evolved Emotional Policies for LLM Agents in Multi-Turn Negotiation

4. Intermediate Languages Matter: Formal Languages and LLMs affect Neurosymbolic Reasoning

5. CoT-Space: A Theoretical Framework for Internal Slow-Thinking via Reinforcement Learning

6. AutoPBO: LLM-powered Optimization for Local Search PBO Solvers

7. Meta-Policy Reflexion: Reusable Reflective Memory and Rule Admissibility for Resource-Efficient LLM Agent

8. World Model Implanting for Test-time Adaptation of Embodied Agents

9. A Foundation Model for Chest X-ray Interpretation with Grounded Reasoning via Online Reinforcement Learning

10. FaMA: LLM-Empowered Agentic Assistant for Consumer-to-Consumer Marketplace

11. Expedition & Expansion: Leveraging Semantic Representations for Goal-Directed Exploration in Continuous Cellular Automata

12. Continuous Monitoring of Large-Scale Generative AI via Deterministic Knowledge Graph Structures

13. An Agentic Model Context Protocol Framework for Medical Concept Standardization

14. What Would an LLM Do? Evaluating Policymaking Capabilities of Large Language Models

15. Learning to Deliberate: Meta-policy Collaboration for Agentic LLMs with Multi-agent Reinforcement Learning

16. Leveraging LLM-Based Agents for Intelligent Supply Chain Planning

17. RAGuard: A Novel Approach for in-context Safe Retrieval Augmented Generation for LLMs

18. Are LLM Agents Behaviorally Coherent? Latent Profiles for Social Simulation

19. The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior in LLMs

20. Emergent Hierarchical Reasoning in LLMs through Reinforcement Learning

21. Towards a Neurosymbolic Reasoning System Grounded in Schematic Representations

22. CausalARC: Abstract Reasoning with Causal World Models

23. Explainable Knowledge Graph Retrieval-Augmented Generation (KG-RAG) with KG-SMILE

24. Learning When to Plan: Efficiently Allocating Test-Time Compute for LLM Agents

25. PG-Agent: An Agent Powered by Page Graph

26. Delta Activations: A Representation for Finetuned Large Language Models

27. Towards a Unified View of Large Language Model Post-Training

28. No Thoughts Just AI: Biased LLM Recommendations Limit Human Agency in Resume Screening

29. Facts Fade Fast: Evaluating Memorization of Outdated Medical Knowledge in Large Language Models

30. An Empirical Study of Vulnerabilities in Python Packages and Their Detection

31. How many patients could we save with LLM priors?

32. Learning Active Perception via Self-Evolving Preference Optimization for GUI Grounding

33. MAGneT: Coordinated Multi-Agent Generation of Synthetic Multi-Turn Mental Health Counseling Sessions

34. TAGAL: Tabular Data Generation using Agentic LLM Methods

35. Enhancing Technical Documents Retrieval for RAG

36. MEPG:Multi-Expert Planning and Generation for Compositionally-Rich Image Generation

37. RepoDebug: Repository-Level Multi-Task and Multi-Language Debugging Evaluation of Large Language Models

38. On Robustness and Reliability of Benchmark-Based Evaluation of LLMs

39. NER Retriever: Zero-Shot Named Entity Retrieval with Type-Aware Embeddings

40. RTQA : Recursive Thinking for Complex Temporal Knowledge Graph Question Answering with Large Language Models

41. NeuroBreak: Unveil Internal Jailbreak Mechanisms in Large Language Models

42. Expanding Foundational Language Capabilities in Open-Source LLMs through a Korean Case Study

43. Multimodal Feature Fusion Network with Text Difference Enhancement for Remote Sensing Change Detection

44. CANDY: Benchmarking LLMs’ Limitations and Assistive Potential in Chinese Misinformation Fact-Checking

45. VoxRole: A Comprehensive Benchmark for Evaluating Speech-Based Role-Playing Agents

46. SPFT-SQL: Enhancing Large Language Model for Text-to-SQL Parsing by Self-Play Fine-Tuning

47. SelfAug: Mitigating Catastrophic Forgetting in Retrieval-Augmented Generation via Distribution Self-Alignment

48. MTQA:Matrix of Thought for Enhanced Reasoning in Complex Question Answering

49. A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models

50. INGRID: Intelligent Generative Robotic Design Using Large Language Models

51. Gravity Well Echo Chamber Modeling With An LLM-Based Confirmation Bias Model

52. Align-then-Slide: A complete evaluation framework for Ultra-Long Document-Level Machine Translation

53. Measuring How (Not Just Whether) VLMs Build Common Ground

54. SAMVAD: A Multi-Agent System for Simulating Judicial Deliberation Dynamics in India

55. Designing Gaze Analytics for ELA Instruction: A User-Centered Dashboard with Conversational AI Support

56. Breaking the Mirror: Activation-Based Mitigation of Self-Preference in LLM Evaluators

57. E-ARMOR: Edge case Assessment and Review of Multilingual Optical Character Recognition

58. Improving Factuality in LLMs via Inference-Time Knowledge Graph Construction

59. AR$^2$: Adversarial Reinforcement Learning for Abstract Reasoning in Large Language Models

60. Real-Time Detection of Hallucinated Entities in Long-Form Generation

61. Multilevel Analysis of Cryptocurrency News using RAG Approach with Fine-Tuned Mistral Large Language Model

62. Speech-Based Cognitive Screening: A Systematic Evaluation of LLM Adaptation Strategies