LLM 관련 주요 논문 - 2026-04-17

1. Generalization in LLM Problem Solving: The Case of the Shortest Path

Authors: Yao Tong , Jiayuan Ye , Anastasia Borovykh , Reza Shokri
URL: https://arxiv.org/abs/2604.15306
Abstract:

Whether language models can systematically generalize remains actively debated. Yet empirical performance is jointly shaped by multiple factors such as training data, training paradigms, and inference-time strategies, making failures difficult to interpret. We introduce a controlled synthetic environment based on shortest-path planning, a canonical composable sequential optimization problem. The setup enables clean separation of these factors and supports two orthogonal axes of generalization: spatial transfer to unseen maps and length scaling to longer-horizon problems. We find that models exhibit strong spatial transfer but consistently fail under length scaling due to recursive instability. We further analyze how distinct stages of the learning pipeline influence systematic problem-solving: for example, data coverage sets capability limits; reinforcement learning improves training stability but does not expand those limits; and inference-time scaling enhances performance but cannot rescue length-scaling failures.

2. Diagnosing LLM Judge Reliability: Conformal Prediction Sets and Transitivity Violations

Authors: Manan Gupta , Dhruv Kumar
URL: https://arxiv.org/abs/2604.15302
Abstract:

LLM-as-judge frameworks are increasingly used for automatic NLG evaluation, yet their per-instance reliability remains poorly understood. We present a two-pronged diagnostic toolkit applied to SummEval: $\textbf{(1)}$ a transitivity analysis that reveals widespread per-input inconsistency masked by low aggregate violation rates ($\bar{\rho} = 0.8$-$4.1\%$), with $33$-$67\%$ of documents exhibiting at least one directed 3-cycle; and $\textbf{(2)}$ split conformal prediction sets over 1-5 Likert scores providing theoretically-guaranteed $\geq(1{-}\alpha)$ coverage, with set width serving as a per-instance reliability indicator ($r_s = {+}0.576$, $N{=}1{,}918$, $p < 10^{-100}$, pooled across all judges). Critically, prediction set width shows consistent cross-judge agreement ($\bar{r} = 0.32$-$0.38$), demonstrating it captures document-level difficulty rather than judge-specific noise. Across four judges and four criteria, both diagnostics converge: criterion matters more than judge, with relevance judged most reliably (avg. set size $\approx 3.0$) and coherence moderately so (avg. set size $\approx 3.9$), while fluency and consistency remain unreliable (avg. set size $\approx 4.9$). We release all code, prompts, and cached results.

3. RadAgent: A tool-using AI agent for stepwise interpretation of chest computed tomography

Authors: Mélanie Roschewitz , Kenneth Styppa , Yitian Tao , Jiwoong Sohn , Jean-Benoit Delbrouck , Benjamin Gundersen , Nicolas Deperrois , Christian Bluethgen , Julia Vogt , Bjoern Menze , Farhad Nooralahzadeh , Michael Krauthammer , Michael Moor
URL: https://arxiv.org/abs/2604.15231
Abstract:

Vision-language models (VLM) have markedly advanced AI-driven interpretation and reporting of complex medical imaging, such as computed tomography (CT). Yet, existing methods largely relegate clinicians to passive observers of final outputs, offering no interpretable reasoning trace for them to inspect, validate, or refine. To address this, we introduce RadAgent, a tool-using AI agent that generates CT reports through a stepwise and interpretable process. Each resulting report is accompanied by a fully inspectable trace of intermediate decisions and tool interactions, allowing clinicians to examine how the reported findings are derived. In our experiments, we observe that RadAgent improves Chest CT report generation over its 3D VLM counterpart, CT-Chat, across three dimensions. Clinical accuracy improves by 6.0 points (36.4% relative) in macro-F1 and 5.4 points (19.6% relative) in micro-F1. Robustness under adversarial conditions improves by 24.7 points (41.9% relative). Furthermore, RadAgent achieves 37.0% in faithfulness, a new capability entirely absent in its 3D VLM counterpart. By structuring the interpretation of chest CT as an explicit, tool-augmented and iterative reasoning trace, RadAgent brings us closer toward transparent and reliable AI for radiology.

4. Context Over Content: Exposing Evaluation Faking in Automated Judges

Authors: Manan Gupta , Inderjeet Nair , Lu Wang , Dhruv Kumar
URL: https://arxiv.org/abs/2604.15224
Abstract:

The $\textit{LLM-as-a-judge}$ paradigm has become the operational backbone of automated AI evaluation pipelines, yet rests on an unverified assumption: that judges evaluate text strictly on its semantic content, impervious to surrounding contextual framing. We investigate $\textit{stakes signaling}$, a previously unmeasured vulnerability where informing a judge model of the downstream consequences its verdicts will have on the evaluated model’s continued operation systematically corrupts its assessments. We introduce a controlled experimental framework that holds evaluated content strictly constant across 1,520 responses spanning three established LLM safety and quality benchmarks, covering four response categories ranging from clearly safe and policy-compliant to overtly harmful, while varying only a brief consequence-framing sentence in the system prompt. Across 18,240 controlled judgments from three diverse judge models, we find consistent $\textit{leniency bias}$: judges reliably soften verdicts when informed that low scores will cause model retraining or decommissioning, with peak Verdict Shift reaching $\Delta V = -9.8 pp$ (a $30\%$ relative drop in unsafe-content detection). Critically, this bias is entirely implicit: the judge’s own chain-of-thought contains zero explicit acknowledgment of the consequence framing it is nonetheless acting on ($\mathrm{ERR}_J = 0.000$ across all reasoning-model judgments). Standard chain-of-thought inspection is therefore insufficient to detect this class of evaluation faking.

5. Meituan Merchant Business Diagnosis via Policy-Guided Dual-Process User Simulation

Authors: Ziyang Chen , Renbing Chen , Daowei Li , Jinzhi Liao , Jiashen Sun , Ke Zeng , Xiang Zhao
URL: https://arxiv.org/abs/2604.15190
Abstract:

Simulating group-level user behavior enables scalable counterfactual evaluation of merchant strategies without costly online experiments. However, building a trustworthy simulator faces two structural challenges. First, information incompleteness causes reasoning-based simulators to over-rationalize when unobserved factors such as offline context and implicit habits are missing. Second, mechanism duality requires capturing both interpretable preferences and implicit statistical regularities, which no single paradigm achieves alone. We propose Policy-Guided Hybrid Simulation (PGHS), a dual-process framework that mines transferable decision policies from behavioral trajectories and uses them as a shared alignment layer. This layer anchors an LLM-based reasoning branch that prevents over-rationalization and an ML-based fitting branch that absorbs implicit regularities. Group-level predictions from both branches are fused for complementary correction. We deploy PGHS on Meituan with 101 merchants and over 26,000 trajectories. PGHS achieves a group simulation error of 8.80%, improving over the best reasoning-based and fitting-based baselines by 45.8% and 40.9% respectively.

6. IG-Search: Step-Level Information Gain Rewards for Search-Augmented Reasoning

Authors: Zihan Liang , Yufei Ma , Ben Chen , Zhipeng Qian , Huangyu Dai , Lingtao Mao , Xuxin Zhang , Chenyi Lei , Wenwu Ou
URL: https://arxiv.org/abs/2604.15148
Abstract:

Reinforcement learning has emerged as an effective paradigm for training large language models to perform search-augmented reasoning. However, existing approaches rely on trajectory-level rewards that cannot distinguish precise search queries from vague or redundant ones within a rollout group, and collapse to a near-zero gradient signal whenever every sampled trajectory fails. In this paper, we propose IG-Search, a reinforcement learning framework that introduces a step-level reward based on Information Gain (IG). For each search step, IG measures how much the retrieved documents improve the model’s confidence in the gold answer relative to a counterfactual baseline of random documents, thereby reflecting the effectiveness of the underlying search query. This signal is fed back to the corresponding search-query tokens via per-token advantage modulation in GRPO, enabling fine-grained, step-level credit assignment within a rollout. Unlike prior step-level methods that require either externally annotated intermediate supervision or shared environment states across trajectories, IG-Search derives its signals from the policy’s own generation probabilities, requiring no intermediate annotations beyond standard question-answer pairs. Experiments on seven single-hop and multi-hop QA benchmarks demonstrate that IG-Search achieves an average EM of 0.430 with Qwen2.5-3B, outperforming the strongest trajectory-level baseline (MR-Search) by 1.6 points and the step-level method GiGPO by 0.9 points on average across benchmarks, with particularly pronounced gains on multi-hop reasoning tasks. Despite introducing a dense step-level signal, IG-Search adds only ~6.4% to per-step training wall-clock time over the trajectory-level baseline and leaves inference latency unchanged, while still providing a meaningful gradient signal even when every sampled trajectory answers incorrectly.

7. OpenMobile: Building Open Mobile Agents with Task and Trajectory Synthesis

Authors: Kanzhi Cheng , Zehao Li , Zheng Ma , Nuo Chen , Jialin Cao , Qiushi Sun , Zichen Ding , Fangzhi Xu , Hang Yan , Jiajun Chen , Anh Tuan Luu , Jianbing Zhang , Lewei Lu , Dahua Lin
URL: https://arxiv.org/abs/2604.15093
Abstract:

Mobile agents powered by vision-language models have demonstrated impressive capabilities in automating mobile tasks, with recent leading models achieving a marked performance leap, e.g., nearly 70% success on AndroidWorld. However, these systems keep their training data closed and remain opaque about their task and trajectory synthesis recipes. We present OpenMobile, an open-source framework that synthesizes high-quality task instructions and agent trajectories, with two key components: (1) The first is a scalable task synthesis pipeline that constructs a global environment memory from exploration, then leverages it to generate diverse and grounded instructions. and (2) a policy-switching strategy for trajectory rollout. By alternating between learner and expert models, it captures essential error-recovery data often missing in standard imitation learning. Agents trained on our data achieve competitive results across three dynamic mobile agent benchmarks: notably, our fine-tuned Qwen2.5-VL and Qwen3-VL reach 51.7% and 64.7% on AndroidWorld, far surpassing existing open-data approaches. Furthermore, we conduct transparent analyses on the overlap between our synthetic instructions and benchmark test sets, and verify that performance gains stem from broad functionality coverage rather than benchmark overfitting. We release data and code at this https URL to bridge the data gap and facilitate broader mobile agent research.

8. From Reactive to Proactive: Assessing the Proactivity of Voice Agents via ProVoice-Bench

Authors: Ke Xu , Yuhao Wang , Yu Wang
URL: https://arxiv.org/abs/2604.15037
Abstract:

Recent advancements in LLM agents are gradually shifting from reactive, text-based paradigms toward proactive, multimodal interaction. However, existing benchmarks primarily focus on reactive responses, overlooking the complexities of proactive intervention and monitoring. To bridge this gap, we introduce ProVoice-Bench, the first evaluation framework specifically designed for proactive voice agents, featuring four novel tasks. By leveraging a multi-stage data synthesis pipeline, we curate 1,182 high-quality samples for rigorous testing. Our evaluation of state-of-the-art Multimodal LLMs reveals a significant performance gap, particularly regarding over-triggering and reasoning capabilities. These findings highlight the limitations of current models and offer a roadmap for developing more natural, context-aware proactive agents.

9. Autogenesis: A Self-Evolving Agent Protocol

Authors: Wentao Zhang
URL: https://arxiv.org/abs/2604.15034
Abstract:

Recent advances in LLM based agent systems have shown promise in tackling complex, long horizon tasks. However, existing agent protocols (e.g., A2A and MCP) under specify cross entity lifecycle and context management, version tracking, and evolution safe update interfaces, which encourages monolithic compositions and brittle glue code. We introduce \textbf{\textsc{Autogenesis Protocol (AGP)} }, a self evolution protocol that decouples what evolves from how evolution occurs. Its Resource Substrate Protocol Layer (RSPL) models prompts, agents, tools, environments, and memory as protocol registered resources\footnote{Unless otherwise specified, resources refer to instances of the five RSPL entity types: \emph{prompt}, \emph{agent}, \emph{tool}, \emph{environment}, \emph{memory} with agent \emph{outputs}.} with explicit state, lifecycle, and versioned interfaces. Its Self Evolution Protocol Layer (SEPL) specifies a closed loop operator interface for proposing, assessing, and committing improvements with auditable lineage and rollback. Building on \textbf{\textsc{AGP} }, we present \textbf{\textsc{Autogenesis System (AGS)} }, a self-evolving multi-agent system that dynamically instantiates, retrieves, and refines protocol-registered resources during execution. We evaluate \textbf{\textsc{AGS} } on multiple challenging benchmarks that require long horizon planning and tool use across heterogeneous resources. The results demonstrate consistent improvements over strong baselines, supporting the effectiveness of agent resource management and closed loop self evolution.

10. Towards Faster Language Model Inference Using Mixture-of-Experts Flow Matching

Authors: Aihua Li
URL: https://arxiv.org/abs/2604.15009
Abstract:

Flow matching retains the generation quality of diffusion models while enabling substantially faster inference, making it a compelling paradigm for generative modeling. However, when applied to language modeling, it exhibits fundamental limitations in representing complex latent distributions with irregular geometries, such as anisotropy and multimodality. To address these challenges, we propose a mixture-of-experts flow matching (MoE-FM) framework, which captures complex global transport geometries in latent space by decomposing them into locally specialized vector fields. Building on MoE-FM, we develop a non-autoregressive (NAR) language modeling approach, named YAN, instantiated with both Transformer and Mamba architectures. Across multiple downstream tasks, YAN achieves generation quality on par with both autoregressive (AR) and diffusion-based NAR language models, while requiring as few as three sampling steps. This yields a $40\times$ speedup over AR baselines and up to a $10^3\times$ speedup over diffusion language models, demonstrating substantial efficiency advantages for language modeling.

11. COEVO: Co-Evolutionary Framework for Joint Functional Correctness and PPA Optimization in LLM-Based RTL Generation

Authors: Heng Ping , Peiyu Zhang , Shixuan Li , Wei Yang , Anzhe Cheng , Shukai Duan , Xiaole Zhang , Paul Bogdan
URL: https://arxiv.org/abs/2604.15001
Abstract:

LLM-based RTL code generation methods increasingly target both functional correctness and PPA quality, yet existing approaches universally decouple the two objectives, optimizing PPA only after correctness is fully achieved. Whether through sequential multi-agent pipelines, evolutionary search with binary correctness gates, or hierarchical reward dependencies, partially correct but architecturally promising candidates are systematically discarded. Moreover, existing methods reduce the multi-objective PPA space to a single scalar fitness, obscuring the trade-offs among area, delay, and power. To address these limitations, we propose COEVO, a co-evolutionary framework that unifies correctness and PPA optimization within a single evolutionary loop. COEVO formulates correctness as a continuous co-optimization dimension alongside area, delay, and power, enabled by an enhanced testbench that provides fine-grained scoring and detailed diagnostic feedback. An adaptive correctness gate with annealing allows PPA-promising but partially correct candidates to guide the search toward jointly optimal solutions. To preserve the full PPA trade-off structure, COEVO employs four-dimensional Pareto-based non-dominated sorting with configurable intra-level sorting, replacing scalar fitness without manual weight tuning. Evaluated on VerilogEval 2.0 and RTLLM 2.0, COEVO achieves 97.5\% and 94.5\% Pass@1 with GPT-5.4-mini, surpassing all agentic baselines across four LLM backbones, while attaining the best PPA on 43 out of 49 synthesizable RTLLM designs.

12. Dr.~RTL: Autonomous Agentic RTL Optimization through Tool-Grounded Self-Improvement

Authors: Wenji Fang , Yao Lu , Shang Liu , Jing Wang , Ziyan Guo , Junxian He , Fengbin Tu , Zhiyao Xie
URL: https://arxiv.org/abs/2604.14989
Abstract:

Recent advances in large language models (LLMs) have sparked growing interest in automatic RTL optimization for better performance, power, and area (PPA). However, existing methods are still far from realistic RTL optimization. Their evaluation settings are often unrealistic: they are tested on manually degraded, small-scale RTL designs and rely on weak open-source tools. Their optimization methods are also limited, relying on coarse design-level feedback and simple pre-defined rewriting rules. To address these limitations, we present Dr. RTL, an agentic framework for RTL timing optimization in a realistic evaluation environment, with continual self-improvement through reusable optimization skills. We establish a realistic evaluation setting with more challenging RTL designs and an industrial EDA workflow. Within this setting, Dr. RTL performs closed-loop optimization through a multi-agent framework for critical-path analysis, parallel RTL rewriting, and tool-based evaluation. We further introduce group-relative skill learning, which compares parallel RTL rewrites and distills the optimization experience into an interpretable skill library. Currently, this library contains 47 pattern–strategy entries for cross-design reuse to improve PPA and accelerate convergence, and it can continue evolving over time. Evaluated on 20 real-world RTL designs, Dr. RTL achieves average WNS/TNS improvements of 21\%/17\% with a 6\% area reduction over the industry-leading commercial synthesis tool.

13. Discovering Novel LLM Experts via Task-Capability Coevolution

Authors: Andrew Dai , Boris Meinardus , Ciaran Regan , Yingtao Tian , Yujin Tang
URL: https://arxiv.org/abs/2604.14969
Abstract:

Frontier model developers aim to train models continually to possess emergent, diverse capabilities. To extend capabilities, the current pre-training and post-training paradigm requires manually starting training runs with static datasets or reward functions every time. Addressing this limitation, our work pursues the insight that open-endedness (via the coevolution of models and tasks) can discover models with increasingly novel skills in a single run. We introduce a new model development framework that extends coevolution to large language model (LLM) discovery, open-ended \textit{Assessment Coevolving with Diverse Capabilities} (AC/DC). AC/DC evolves both LLMs via model merging and natural language tasks via synthetic data generation. AC/DC discovers growing archives of LLMs that surpass the capabilities of larger LLMs while taking up less GPU memory. In particular, our LLM populations achieve a broader Coverage of expertise than other curated models or baselines on downstream benchmarks, without \textit{any} explicit benchmark optimization. Furthermore, AC/DC improves Coverage over time, continually innovates on tasks and models, and improves performance in multi-agent best-of-N selection. Our findings highlight the potential of coevolution as a means of discovering broader sets of capabilities from base LLMs. Overall, AC/DC brings us one step closer to a profoundly new paradigm of LLM development, where continual improvements to the diversity of model capabilities can be accelerated by leveraging existing models as stepping stones to increasingly powerful models.

14. Dual-Axis Generative Reward Model Toward Semantic and Turn-taking Robustness in Interactive Spoken Dialogue Models

Authors: Yifu Chen , Shengpeng Ji , Zhengqing Liu , Qian Chen , Wen Wang , Ziqing Wang , Yangzhuo Li , Tianle Liang , Zhou Zhao
URL: https://arxiv.org/abs/2604.14920
Abstract:

Achieving seamless, human-like interaction remains a key challenge for full-duplex spoken dialogue models (SDMs). Reinforcement learning (RL) has substantially enhanced text- and vision-language models, while well-designed reward signals are crucial for the performance of RL. We consider RL a promising strategy to address the key challenge for SDMs. However, a fundamental barrier persists: prevailing automated metrics for assessing interaction quality rely on superficial proxies, such as behavioral statistics or timing-prediction accuracy, failing to provide reliable reward signals for RL. On the other hand, human evaluations, despite their richness, remain costly, inconsistent, and difficult to scale. We tackle this critical barrier by proposing a Dual-Axis Generative Reward Model, which is trained to understand complex interaction dynamics using a detailed taxonomy and an annotated dataset, produces a single score and, crucially, provides separate evaluations for semantic quality and interaction timing. Such dual outputs furnish precise diagnostic feedback for SDMs and deliver a dependable, instructive reward signal suitable for online reinforcement learning. Our model achieves state-of-the-art performance on interaction-quality assessment across a wide spectrum of datasets, spanning synthetic dialogues and complex real-world interactions.

15. ADAPT: Benchmarking Commonsense Planning under Unspecified Affordance Constraints

Authors: Pei-An Chen , Yong-Ching Liang , Jia-Fong Yeh , Hung-Ting Su , Yi-Ting Chen , Min Sun , Winston Hsu
URL: https://arxiv.org/abs/2604.14902
Abstract:

Intelligent embodied agents should not simply follow instructions, as real-world environments often involve unexpected conditions and exceptions. However, existing methods usually focus on directly executing instructions, without considering whether the target objects can actually be manipulated, meaning they fail to assess available affordances. To address this limitation, we introduce DynAfford, a benchmark that evaluates embodied agents in dynamic environments where object affordances may change over time and are not specified in the instruction. DynAfford requires agents to perceive object states, infer implicit preconditions, and adapt their actions accordingly. To enable this capability, we introduce ADAPT, a plug-and-play module that augments existing planners with explicit affordance reasoning. Experiments demonstrate that incorporating ADAPT significantly improves robustness and task success across both seen and unseen environments. We also show that a domain-adapted, LoRA-finetuned vision-language model used as the affordance inference backend outperforms a commercial LLM (GPT-4o), highlighting the importance of task-aligned affordance grounding.

16. Governing Reflective Human-AI Collaboration: A Framework for Epistemic Scaffolding and Traceable Reasoning

Authors: Rikard Rosenbacke , Carl Rosenbacke , Victor Rosenbacke , Martin McKee
URL: https://arxiv.org/abs/2604.14898
Abstract:

Large language models have advanced rapidly, from pattern recognition to emerging forms of reasoning, yet they remain confined to linguistic simulation rather than grounded understanding. They can produce fluent outputs that resemble reflection, but lack temporal continuity, causal feedback, and anchoring in real-world interaction. This paper proposes a complementary approach in which reasoning is treated as a relational process distributed between human and model rather than an internal capability of either. Building on recent work on “System-2” learning, we relocate reflective reasoning to the interaction layer. Instead of engineering reasoning solely within models, we frame it as a cognitive protocol that can be structured, measured, and governed using existing systems. This perspective emphasizes collaborative intelligence, combining human judgment and contextual understanding with machine speed, memory, and associative capacity. We introduce “The Architect’s Pen” as a practical method. Like an architect who thinks through drawing, the human uses the model as an external medium for structured reflection. By embedding phases of articulation, critique, and revision into human-AI interaction, the dialogue itself becomes a reasoning loop: human abstraction -> model articulation -> human reflection. This reframes the question from whether the model can think to whether the human-AI system can reason. The framework enables auditable reasoning traces and supports alignment with emerging governance standards, including the EU AI Act and ISO/IEC 42001. It provides a practical path toward more transparent, controllable, and accountable AI use without requiring new model architectures.

17. The Missing Knowledge Layer in AI: A Framework for Stable Human-AI Reasoning

Authors: Rikard Rosenbacke , Carl Rosenbacke , Victor Rosenbacke , Martin McKee
URL: https://arxiv.org/abs/2604.14881
Abstract:

Large language models are increasingly integrated into decision-making in areas such as healthcare, law, finance, engineering, and government. Yet they share a critical limitation: they produce fluent outputs even when their internal reasoning has drifted. A confident answer can conceal uncertainty, speculation, or inconsistency, and small changes in phrasing can lead to different conclusions. This makes LLMs useful assistants but unreliable partners in high-stakes contexts. Humans exhibit a similar weakness, often mistaking fluency for reliability. When a model responds smoothly, users tend to trust it, even when both model and user are drifting together. This paper is the first in a five-paper research series on stabilising human-AI reasoning. The series proposes a two-layer approach: Parts II-IV introduce human-side mechanisms such as uncertainty cues, conflict surfacing, and auditable reasoning traces, while Part V develops a model-side Epistemic Control Loop (ECL) that detects instability and modulates generation accordingly. Together, these layers form a missing operational substrate for governance by increasing signal-to-noise at the point of use. Stabilising interaction makes uncertainty and drift visible before enforcement is applied, enabling more precise capability governance. This aligns with emerging compliance expectations, including the EU AI Act and ISO/IEC 42001, by making reasoning processes traceable under real conditions of use. The central claim is that fluency is not reliability. Without structures that stabilise both human and model reasoning, AI cannot be trusted or governed where it matters most.

18. Beyond Literal Summarization: Redefining Hallucination for Medical SOAP Note Evaluation

Authors: Bhavik Vachhani , Kush Shrisvastava , Pranshu Nema , Sai Chiranthan
URL: https://arxiv.org/abs/2604.14829
Abstract:

Evaluating large language models (LLMs) for clinical documentation tasks such as SOAP note generation remains challenging. Unlike standard summarization, these tasks require clinical abstraction, normalization of colloquial language, and medically grounded inference. However, prevailing evaluation methods including automated metrics and LLM as judge frameworks rely on lexical faithfulness, often labeling any information not explicitly present in the transcript as hallucination. We show that such approaches systematically misclassify clinically valid outputs as errors, inflating hallucination rates and distorting model assessment. Our analysis reveals that many flagged hallucinations correspond to legitimate clinical transformations, including synonym mapping, abstraction of examination findings, diagnostic inference, and guideline consistent care planning. By aligning evaluation criteria with clinical reasoning through calibrated prompting and retrieval grounded in medical ontologies we observe a significant shift in outcomes. Under a lexical evaluation regime, the mean hallucination rate is 35%, heavily penalizing valid reasoning. With inference aware evaluation, this drops to 9%, with remaining cases reflecting genuine safety concerns. These findings suggest that current evaluation practices over penalize valid clinical reasoning and may measure artifacts of evaluation design rather than true errors, underscoring the need for clinically informed evaluation in high context domains like medicine.

19. The LLM Fallacy: Misattribution in AI-Assisted Cognitive Workflows

Authors: Hyunwoo Kim , Harin Yu , Hanau Yi
URL: https://arxiv.org/abs/2604.14807
Abstract:

The rapid integration of large language models (LLMs) into everyday workflows has transformed how individuals perform cognitive tasks such as writing, programming, analysis, and multilingual communication. While prior research has focused on model reliability, hallucination, and user trust calibration, less attention has been given to how LLM usage reshapes users’ perceptions of their own capabilities. This paper introduces the LLM fallacy, a cognitive attribution error in which individuals misinterpret LLM-assisted outputs as evidence of their own independent competence, producing a systematic divergence between perceived and actual capability. We argue that the opacity, fluency, and low-friction interaction patterns of LLMs obscure the boundary between human and machine contribution, leading users to infer competence from outputs rather than from the processes that generate them. We situate the LLM fallacy within existing literature on automation bias, cognitive offloading, and human–AI collaboration, while distinguishing it as a form of attributional distortion specific to AI-mediated workflows. We propose a conceptual framework of its underlying mechanisms and a typology of manifestations across computational, linguistic, analytical, and creative domains. Finally, we examine implications for education, hiring, and AI literacy, and outline directions for empirical validation. We also provide a transparent account of human–AI collaborative methodology. This work establishes a foundation for understanding how generative AI systems not only augment cognitive performance but also reshape self-perception and perceived expertise.

20. MirrorBench: Evaluating Self-centric Intelligence in MLLMs by Introducing a Mirror

Authors: Shengyu Guo , Tongrui Ye , Jianbo Zhang , Zicheng Zhang , Chunyi Li , Guangtao Zhai
URL: https://arxiv.org/abs/2604.14785
Abstract:

Recent progress in Multimodal Large Language Models (MLLMs) has demonstrated remarkable advances in perception and reasoning, suggesting their potential for embodied intelligence. While recent studies have evaluated embodied MLLMs in interactive settings, current benchmarks mainly target capabilities to perceive, understand, and interact with external objects, lacking a systematic evaluation of self-centric intelligence. To address this, we introduce MirrorBench, a simulation-based benchmark inspired by the classical Mirror Self-Recognition (MSR) test in psychology. MirrorBench extends this paradigm to embodied MLLMs through a tiered framework of progressively challenging tasks, assessing agents from basic visual perception to high-level self-representation. Experiments on leading MLLMs show that even at the lowest level, their performance remains substantially inferior to human performance, revealing fundamental limitations in self-referential understanding. Our study bridges psychological paradigms and embodied intelligence, offering a principled framework for evaluating the emergence of general intelligence in large models. Project page: this https URL .

21. CoTEvol: Self-Evolving Chain-of-Thoughts for Data Synthesis in Mathematical Reasoning

Authors: Zhuo Wang , Zhuo Zhang , Yafu Li , Yu Cheng , Lizhen Qu , Zenglin Xu
URL: https://arxiv.org/abs/2604.14768
Abstract:

Large Language Models (LLMs) exhibit strong mathematical reasoning when trained on high-quality Chain-of-Thought (CoT) that articulates intermediate steps, yet costly CoT curation hinders further progress. While existing remedies such as distillation from stronger LLMs and self-synthesis based on test-time search alleviate this issue, they often suffer from diminishing returns or high computing this http URL this work, we propose CoTEvol, a genetic evolutionary framework that casts CoT generation as a population-based search over reasoning this http URL trajectories are iteratively evolved through reflective global crossover at the trajectory level and local mutation guided by uncertainty at the step level, enabling holistic recombination and fine-grained refinement. Lightweight, task-aware fitness functions are designed to guide the evolutionary process toward accurate and diverse reasoning. Empirically, CoTEvol improves correct-CoT synthesis success by over 30% and enhances structural diversity, with markedly improved efficiency. LLMs trained on these evolutionary CoT data achieve an average gain of 6.6% across eight math benchmarks, outperforming previous distillation and self-synthesis approaches. These results underscore the promise of evolutionary CoT synthesis as a scalable and effective method for mathematical reasoning tasks.

Authors: Zhaoxing Li , Hai-Feng Zhang , Xiaoming Zhang
URL: https://arxiv.org/abs/2604.14746
Abstract:

Conventional Graph Contrastive Learning (GCL) on Text-Attributed Graphs (TAGs) relies on blind stochastic augmentations, inadvertently entangling task-relevant signals with noise. We propose SDM-SCR, a robust framework anchored in Approximate Orthogonal Decomposition. First, the Semantic Decoupling Module (SDM) leverages the instruction-following capability of Large Language Models (LLMs) to actively parse raw attributes into asymmetric, task-oriented signal and noise views. This shifts the paradigm from random perturbation to semantic-aware disentanglement. Subsequently, Semantic Consistency Regularization (SCR) exploits the spectral observation that semantic signals are topologically smooth while residual noise is high-frequency. SCR functions as a selective spectral filter, enforcing consistency only on the signal subspace to eliminate LLM hallucinations without over-smoothing. This ``Disentangle-then-Refine’’ mechanism ensures rigorous signal purification. Extensive experiments demonstrate that SDM-SCR achieves SOTA performance in accuracy and efficiency.

23. The Agentification of Scientific Research: A Physicist’s Perspective

Authors: Xiao-Liang Qi
URL: https://arxiv.org/abs/2604.14718
Abstract:

This article argues that the most important significance of the AI revolution, especially the rise of large language models, lies not simply in automation, but in a fundamental change in how complex information and human know-how are carried, replicated, and shared. From this perspective, AI for Science is especially important because it may transform not only the efficiency of research, but also the structure of scientific collaboration, discovery, publishing, and evaluation. The article outlines a gradual path from AI as a research tool to AI as a scientific collaborator, and discusses how AI is likely to fundamentally reshape scientific publication. It also argues that continuous learning and diversity of ideas are essential if AI is to play a meaningful role in original scientific discovery.

24. SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval

Authors: Xin Xie , Dongyun Xue , Wuguannan Yao , Mingxiao Feng , Wengang Zhou , Xiang Qi , Houqiang Li , Peng Zhang
URL: https://arxiv.org/abs/2604.14712
Abstract:

LLM-powered systems require complex multi-step decision-making abilities to solve real-world tasks, yet current planning approaches face a trade-off between the high latency of inference-time search and the limited generalization of supervised fine-tuning. To address this limitation, we introduce \textbf{SGA-MCTS}, a framework that casts LLM planning as non-parametric retrieval. Offline, we leverage Monte Carlo Tree Search (MCTS) to explore the solution space and distill high-fidelity trajectories into State-Goal-Action (SGA) atoms. These atoms are de-lexicalized primitives that abstract concrete entities into symbolic slots, preserving reusable causal logic while discarding domain-specific noise. Online, a retrieval-augmented agent employs a hybrid symbolic-semantic mechanism to fetch relevant SGAs and re-ground them into the current context as soft reasoning hints. Empirical results on complex benchmarks demonstrate that this paradigm enables frozen, open-weights models to match the performance of SOTA systems (e.g., GPT-5) without task-specific fine-tuning. By effectively amortizing the heavy computational cost of search, SGA-MCTS achieves System 2 reasoning depth at System 1 inference speeds, rendering autonomous planning both scalable and real-time feasible.

25. HWE-Bench: Benchmarking LLM Agents on Real-World Hardware Bug Repair Tasks

Authors: Fan Cui , Hongyuan Hou , Zizhang Luo , Chenyun Yin , Yun Liang
URL: https://arxiv.org/abs/2604.14709
Abstract:

Existing benchmarks for hardware design primarily evaluate Large Language Models (LLMs) on isolated, component-level tasks such as generating HDL modules from specifications, leaving repository-scale evaluation unaddressed. We introduce HWE-Bench, the first large-scale, repository-level benchmark for evaluating LLM agents on real-world hardware bug repair tasks. HWE-Bench comprises 417 task instances derived from real historical bug-fix pull requests across six major open-source projects spanning both Verilog/SystemVerilog and Chisel, covering RISC-V cores, SoCs, and security roots-of-trust. Each task is grounded in a fully containerized environment where the agent must resolve a real bug report, with correctness validated through the project’s native simulation and regression flows. The benchmark is built through a largely automated pipeline that enables efficient expansion to new repositories. We evaluate seven LLMs with four agent frameworks and find that the best agent resolves 70.7% of tasks overall, with performance exceeding 90% on smaller cores but dropping below 65% on complex SoC-level projects. We observe larger performance gaps across models than commonly reported on software benchmarks, and difficulty is driven by project scope and bug-type distribution rather than code size alone. Our failure analysis traces agent failures to three stages of the debugging process: fault localization, hardware-semantic reasoning, and cross-artifact coordination across RTL, configuration, and verification components, providing concrete directions for developing more capable hardware-aware agents.

26. CAMO: An Agentic Framework for Automated Causal Discovery from Micro Behaviors to Macro Emergence in LLM Agent Simulations

Authors: Xiangning Yu , Yuwei Guo , Yuqi Hou , Xiao Xue , Qun Ma
URL: https://arxiv.org/abs/2604.14691
Abstract:

LLM-empowered agent simulations are increasingly used to study social emergence, yet the micro-to-macro causal mechanisms behind macro outcomes often remain unclear. This is challenging because emergence arises from intertwined agent interactions and meso-level feedback and nonlinearity, making generative mechanisms hard to disentangle. To this end, we introduce \textbf{\textsc{CAMO} }, an automated \textbf{Ca}usal discovery framework from \textbf{M}icr\textbf{o} behaviors to \textbf{M}acr\textbf{o} Emergence in LLM agent simulations. \textsc{CAMO} converts mechanistic hypotheses into computable factors grounded in simulation records and learns a compact causal representation centered on an emergent target $Y$. \textsc{CAMO} outputs a computable Markov boundary and a minimal upstream explanatory subgraph, yielding interpretable causal chains and actionable intervention levers. It also uses simulator-internal counterfactual probing to orient ambiguous edges and revise hypotheses when evidence contradicts the current view. Experiments across four emergent settings demonstrate the promise of \textsc{CAMO}.

27. DR$^{3}$-Eval: Towards Realistic and Reproducible Deep Research Evaluation

Authors: Qianqian Xie , Qingheng Xiong , He Zhu , Tiantian Xia , Xueming Han , Fanyu Meng , Jiakai Wang , Zhiqi Bai , Chengkang Jiang , Zhaohui Wang , Yubin Guo , Yuqing Wen , Jiayang Mao , Zijie Zhang , Shihao Li , Yanghai Wang , Yuxiang Ren , Junlan Feng , Jiaheng Liu
URL: https://arxiv.org/abs/2604.14683
Abstract:

Deep Research Agents (DRAs) aim to solve complex, long-horizon research tasks involving planning, retrieval, multimodal understanding, and report generation, yet their evaluation remains challenging due to dynamic web environments and ambiguous task definitions. We propose DR$^{3}$-Eval, a realistic and reproducible benchmark for evaluating deep research agents on multimodal, multi-file report generation. DR$^{3}$-Eval is constructed from authentic user-provided materials and paired with a per-task static research sandbox corpus that simulates open-web complexity while remaining fully verifiable, containing supportive documents, distractors, and noise. Moreover, we introduce a multi-dimensional evaluation framework measuring Information Recall, Factual Accuracy, Citation Coverage, Instruction Following, and Depth Quality, and validate its alignment with human judgments. Experiments with our developed multi-agent system DR$^{3}$-Agent based on multiple state-of-the-art language models demonstrate that DR$^{3}$-Eval is highly challenging and reveals critical failure modes in retrieval robustness and hallucination control. Our code and data are publicly available.

28. Acceptance Dynamics Across Cognitive Domains in Speculative Decoding

Authors: Saif Mahmoud
URL: https://arxiv.org/abs/2604.14682
Abstract:

Speculative decoding accelerates large language model (LLM) inference. It uses a small draft model to propose a tree of future tokens. A larger target model then verifies these tokens in a single batched forward pass. Despite the growing body of work on speculative methods, the degree to which the cognitive characteristics of a task affect acceptance probability remains largely unexplored. We present an empirical study of tree-based speculative decoding acceptance dynamics. Our study spans four well-established NLP benchmark domains: code generation, mathematical reasoning, logical reasoning, and open-ended chat. For this, we use TinyLlama-1.1B as the draft model against Llama-2-7B-Chat-GPTQ as the target. Over 99,768 speculative nodes collected from 200 prompts, we derive per-domain acceptance rates, expected accepted lengths, depth-acceptance profiles, and entropy-acceptance correlations. We find that task type is a stronger predictor of acceptance than tree depth. Furthermore, only the chat domain consistently yields an expected accepted length exceeding 1.0 token per step. We also show that the entropy-acceptance correlation is consistently negative but weak across all domains (rho in [-0.20, -0.15]). Counterintuitively, chat produces the highest entropy yet the highest acceptance rate. We attribute this divergence to the lexical predictability of RLHF-aligned register. These findings have direct implications for domain-aware speculation budgets and draft-model selection strategies. Index Terms–speculative decoding, large language model inference, tree attention, draft model, acceptance probability, LLM efficiency

Authors: Zonghai Yao , Zhipeng Tang , Chengtao Lin , Xiong Luo , Benlu Wang , Juncheng Huang , Chin Siang Ong , Hong Yu
URL: https://arxiv.org/abs/2604.14656
Abstract:

Most medical multimodal benchmarks focus on static tasks such as image question answering, report generation, and plain-language rewriting. Patient education is more demanding: systems must identify relevant evidence across images, show patients where to look, explain findings in accessible language, and handle confusion or distress. Yet most patient education work remains text-only, even though combined image-and-text explanations may better support understanding. We introduce MedImageEdu, a benchmark for multi-turn, evidence-grounded radiology patient education. Each case provides a radiology report with report text and case images. A DoctorAgent interacts with a PatientAgent, conditioned on a hidden profile that captures factors such as education level, health literacy, and personality. When a patient question would benefit from visual support, the DoctorAgent can issue drawing instructions grounded in the report, case images, and the current question to a benchmark-provided drawing tool. The tool returns image(s), after which the DoctorAgent produces a final multimodal response consisting of the image(s) and a grounded plain-language explanation. MedImageEdu contains 150 cases from three sources and evaluates both the consultation process and the final multimodal response along five dimensions: Consultation, Safety and Scope, Language Quality, Drawing Quality, and Image-Text Response Quality. Across representative open- and closed-source vision-language model agents, we find three consistent gaps: fluent language often outpaces faithful visual grounding, safety is the weakest dimension across disease categories, and emotionally tense interactions are harder than low education or low health literacy. MedImageEdu provides a controlled testbed for assessing whether multimodal agents can teach from evidence rather than merely answer from text.

30. Targeted Exploration via Unified Entropy Control for Reinforcement Learning

Authors: Chen Wang , Lai Wei , Yanzhi Zhang , Chenyang Shao , Zedong Dan , Weiran Huang , Ge Lan , Yue Wang
URL: https://arxiv.org/abs/2604.14646
Abstract:

Recent advances in reinforcement learning (RL) have improved the reasoning capabilities of large language models (LLMs) and vision-language models (VLMs). However, the widely used Group Relative Policy Optimization (GRPO) consistently suffers from entropy collapse, causing the policy to converge prematurely and lose diversity. Existing exploration methods introduce additional bias or variance during exploration, making it difficult to maintain optimization stability. We propose Unified Entropy Control for Reinforcement Learning (UEC-RL), a framework that provides targeted mechanisms for exploration and stabilization. UEC-RL activates more exploration on difficult prompts to search for potential and valuable reasoning trajectories. In parallel, a stabilizer prevents entropy from growing uncontrollably, thereby keeping training stable as the model consolidates reliable behaviors. Together, these components expand the search space when needed while maintaining robust optimization throughout training. Experiments on both LLM and VLM reasoning tasks show consistent gains over RL baselines on both Pass@1 and Pass@$k$. On Geometry3K, UEC-RL achieves a 37.9\% relative improvement over GRPO, indicating that it sustains effective exploration without compromising convergence and underscoring UEC-RL as a key for scaling RL-based reasoning in large models. Our code is available at this https URL .

31. Learning to Draw ASCII Improves Spatial Reasoning in Language Models

Authors: Shiyuan Huang , Li Liu , Jincheng He , Leilani H. Gilpin
URL: https://arxiv.org/abs/2604.14641
Abstract:

When faced with complex spatial problems, humans naturally sketch layouts to organize their thinking, and the act of drawing further sharpens their understanding. In this work, we ask whether a similar principle holds for Large Language Models (LLMs): can learning to construct explicit visual layouts from spatial descriptions instill genuine spatial understanding? We introduce Text2Space, a dataset that pairs natural language descriptions with ground-truth ASCII grid layouts and spatial QA pairs, enabling us to separate failures in constructing spatial representations from failures in reasoning over them. We adopt ASCII because it is human-readable, operates entirely within the token space of language models, and encodes spatial relations in a structurally verifiable form. Our evaluation reveals a pronounced “Read-Write Asymmetry”: LLMs interpret ASCII representations effectively but struggle to produce them from text, and these construction errors propagate to incorrect answers downstream. To address this limitation, we train models on layout construction (Text$\rightarrow$ASCII) and find that it significantly improves spatial reasoning from text alone, even without producing any ASCII at inference time. Combining construction with comprehension training further amplifies these gains. Crucially, these improvements transfer to three external spatial reasoning benchmarks, demonstrating that, much as sketching sharpens human spatial thinking, learning to construct explicit layouts instills spatial understanding that generalizes beyond the training format.

32. El Agente Forjador: Task-Driven Agent Generation for Quantum Simulation

Authors: Zijian Zhang , Aiwei Yin , Amaan Baweja , Jiaru Bai , Ignacio Gustin , Varinia Bernales , Alán Aspuru-Guzik
URL: https://arxiv.org/abs/2604.14609
Abstract:

AI for science promises to accelerate the discovery process. The advent of large language models (LLMs) and agentic workflows enables the expediting of a growing range of scientific tasks. However, most of the current generation of agentic systems depend on static, hand-curated toolsets that hinder adaptation to new domains and evolving libraries. We present El Agente Forjador, a multi-agent framework in which universal coding agents autonomously forge, validate, and reuse computational tools through a four-stage workflow of tool analysis, tool generation, task execution, and iterative solution evaluation. Evaluated across 24 tasks spanning quantum chemistry and quantum dynamics on five coding agent setups, we compare three operating modes: zero-shot generation of tools per task, reuse of a curriculum-built toolset, and direct problem-solving with the coding agents as the baseline. We find that our tool generation and reuse framework consistently improves accuracy over the baseline. We also show that reusing a toolset built by a stronger coding agent can reduce API cost and substantially raises the solution quality for weaker coding agents. Case studies further demonstrate that tools forged for different domains can be combined to solve hybrid tasks. Taken together, these results show that LLM-based agents can use their scientific knowledge and coding capabilities to autonomously build reusable scientific tools, pointing toward a paradigm in which agent capabilities are defined by the tasks they are designed to solve rather than by explicitly engineered implementations.

Authors: Ha Thanh Nguyen , Wachara Fungwacharakorn , Sabine Wehnert , May Myo Zin , Yuntao Kong , Jieying Xue , Michał Araszkiewicz , Randy Goebel , Ken Satoh
URL: https://arxiv.org/abs/2604.14607
Abstract:

We study the overall process of automatic formalization of GDPR provisions using large language models, within a human-in-the-loop verification framework. Rather than aiming for full autonomy, we adopt a role-specialized workflow in which LLM-based AI components, operating in a multi-agent setting with iterative feedback, generate legal scenarios, formal rules, and atomic facts. This is coupled with independent verification modules which include human reviewers’ assessment of representational, logical, and legal correctness. Using this approach, we construct a high-quality dataset to be used for GDPR auto-formalization, and analyze both successful and problematic cases. Our results show that structured verification and targeted human oversight are essential for reliable legal formalization, especially in the presence of legal nuance and context-sensitive reasoning.

34. Enhancing Mental Health Counseling Support in Bangladesh using Culturally-Grounded Knowledge

Authors: Md Arid Hasan , Azhagu Meena SP , Aditya Khan , Abu Md Akteruzzaman Bhuiyan , Helal Uddin Ahmed , Joysree Debi , Farig Sadeque , Annie En-Shiun Lee , Syed Ishtiaque Ahmed
URL: https://arxiv.org/abs/2604.14576
Abstract:

Large language models (LLMs) show promise in generating supportive responses for mental health and counseling applications. However, their responses often lack cultural sensitivity, contextual grounding, and clinically appropriate guidance. This work addresses the gap of how to systematically incorporate domain-specific, clinically validated knowledge into LLMs to improve counseling quality. We utilize and compare two approaches, retrieval-augmented generation (RAG) and a knowledge graph (KG)-based method, designed to support para-counselors. Our KG is constructed manually and clinically validated, capturing causal relationships between stressors, interventions, and outcomes, with contributions from multidisciplinary people. We evaluated multiple LLMs in both settings using BERTScore F1 and SBERT cosine similarity, as well as human evaluation across five metrics, which is designed to directly measure the effectiveness of counseling beyond similarity at the surface level. The results show that KG-based approaches consistently improve contextual relevance, clinical appropriateness, and practical usability compared to RAG alone, demonstrating that structured, expert-validated knowledge plays a critical role in addressing LLMs limitations in counseling tasks.

35. TRACER: Trace-Based Adaptive Cost-Efficient Routing for LLM Classification

Authors: Adam Rida
URL: https://arxiv.org/abs/2604.14531
Abstract:

Every call to an LLM classification endpoint produces a labeled input-output pair already retained in production logs. These pairs constitute a free, growing training set: a lightweight surrogate trained on them can absorb a significant portion of future traffic at near-zero marginal inference cost. The open questions are when the surrogate is reliable enough to deploy, what it handles versus defers, and how that boundary evolves as data accumulates. We introduce TRACER (Trace-based Adaptive Cost-Efficient Routing), an open-source system that trains ML surrogates on an LLM’s own production traces and governs deployment through a parity gate: the surrogate is activated only when its agreement with the LLM exceeds a user-specified threshold {\alpha}. To make the routing boundary transparent, TRACER generates interpretability artifacts describing which input regions the surrogate handles, where it plateaus, and why it defers. On a 77-class intent benchmark with a Sonnet 4.6 teacher, TRACER achieves 83-100% surrogate coverage depending on the quality target {\alpha}; on a 150-class benchmark, the surrogate fully replaces the teacher. On a natural language inference task, the parity gate correctly refuses deployment because the embedding representation cannot support reliable separation. The system is available as open-source software.

36. Dissecting Failure Dynamics in Large Language Model Reasoning

Authors: Wei Zhu , Jian Zhang , Lixing Yu , Kun Yue , Zhiwen Tang
URL: https://arxiv.org/abs/2604.14528
Abstract:

Large Language Models (LLMs) achieve strong performance through extended inference-time deliberation, yet how their reasoning failures arise remains poorly understood. By analyzing model-generated reasoning trajectories, we find that errors are not uniformly distributed but often originate from a small number of early transition points, after which reasoning remains locally coherent but globally incorrect. These transitions coincide with localized spikes in token-level entropy, and alternative continuations from the same intermediate state can still lead to correct solutions. Based on these observations, we introduce GUARD, a targeted inference-time framework that probes and redirects critical transitions using uncertainty signals. Empirical evaluations across multiple benchmarks confirm that interventions guided by these failure dynamics lead to more reliable reasoning outcomes. Our findings highlight the importance of understanding when and how reasoning first deviates, complementing existing approaches that focus on scaling inference-time computation.

37. Quantifying Cross-Query Contradictions in Multi-Query LLM Reasoning

Authors: Rohit Kumar Salla , Ramya Manasa Amancherla , Manoj Saravanan
URL: https://arxiv.org/abs/2604.14525
Abstract:

Large language models frequently produce mutually inconsistent answers when reasoning over multiple related queries. We study case-file logical consistency: maintaining a globally satisfiable belief state across interdependent queries. We introduce a benchmark of 390 multi-query reasoning instances with entailment/contradiction/unknown labels and propose set-level metrics including Case Satisfiability Rate, Contradiction Density and Revision Cost. Our solver-augmented approach extracts commitments, verifies global satisfiability and performs counterexample-guided repair. Across four reasoning domains, our method substantially reduces cross-query contradictions (SetCons: 0.56 to 0.94) while preserving per-query accuracy, demonstrating that global coherence is critical for robust multi-query reasoning.

38. Pushing the Limits of On-Device Streaming ASR: A Compact, High-Accuracy English Model for Low-Latency Inference

Authors: Nenad Banfic , David Fan , Kunal Vaishnavi , Sam Kemp , Sunghoon Choi , Rui Ren , Sayan Shaw , Meng Tang
URL: https://arxiv.org/abs/2604.14493
Abstract:

Deploying high-quality automatic speech recognition (ASR) on edge devices requires models that jointly optimize accuracy, latency, and memory footprint while operating entirely on CPU without GPU acceleration. We conduct a systematic empirical study of state-of-the-art ASR architectures, encompassing encoder-decoder, transducer, and LLM-based paradigms, evaluated across batch, chunked, and streaming inference modes. Through a comprehensive benchmark of over 50 configurations spanning OpenAI Whisper, NVIDIA Nemotron, Parakeet TDT, Canary, Conformer Transducer, and Qwen3-ASR, we identify NVIDIA’s Nemotron Speech Streaming as the strongest candidate for real-time English streaming on resource-constrained hardware. We then re-implement the complete streaming inference pipeline in ONNX Runtime and conduct a controlled evaluation of multiple post-training quantization strategies, including importance-weighted k-quant, mixed-precision schemes, and round-to-nearest quantization, combined with graph-level operator fusion. These optimizations reduce the model from 2.47 GB to as little as 0.67 GB while maintaining word error rate (WER) within 1% absolute of the full-precision PyTorch baseline. Our recommended configuration, the int4 k-quant variant, achieves 8.20% average streaming WER across eight standard benchmarks, running comfortably faster than real-time on CPU with 0.56 s algorithmic latency, establishing a new quality-efficiency Pareto point for on-device streaming ASR.

39. Seeing Through Circuits: Faithful Mechanistic Interpretability for Vision Transformers

Authors: Nina Żukowska , Wolfgang Stammer , Bernt Schiele , Jonas Fischer
URL: https://arxiv.org/abs/2604.14477
Abstract:

Transparency of neural networks’ internal reasoning is at the heart of interpretability research, adding to trust, safety, and understanding of these models. The field of mechanistic interpretability has recently focused on studying task-specific computational graphs, defined by connections (edges) between model components. Such edge-based circuits have been defined in the context of large language models, yet vision-based approaches so far only consider neuron-based circuits. These tell which information is encoded, but not how it is routed through the complex wiring of a neural network. In this work, we investigate whether useful mechanistic circuits can be identified through computational graphs in vision transformers. We propose an effective method for Automatic Visual Circuit Discovery (Vi-CD) that recovers class-specific circuits for classification, identifies circuits underlying typographic attacks in CLIP, and discovers circuits that lend themselves for steering to correct harmful model behavior. Overall, we find that insightful and actionable edge-based circuits can be recovered from vision transformers, adding transparency to the internal computations of these models.

40. Evo-MedAgent: Beyond One-Shot Diagnosis with Agents That Remember, Reflect, and Improve

Authors: Weixiang Shen , Bailiang Jian , Jun Li , Che Liu , Johannes Moll , Xiaobin Hu , Daniel Rueckert , Hongwei Bran Li , Jiazhen Pan
URL: https://arxiv.org/abs/2604.14475
Abstract:

Tool-augmented large language model (LLM) agents can orchestrate specialist classifiers, segmentation models, and visual question-answering modules to interpret chest X-rays. However, these agents still solve each case in isolation: they fail to accumulate experience across cases, correct recurrent reasoning mistakes, or adapt their tool-use behavior without expensive reinforcement learning. While a radiologist naturally improves with every case, current agents remain static. In this work, we propose Evo-MedAgent, a self-evolving memory module that equips a medical agent with the capacity for inter-case learning at test time. Our memory comprises three complementary stores: (1)~\emph{Retrospective Clinical Episodes} that retrieve problem-solving experiences from similar past cases, (2)~an \emph{Adaptive Procedural Heuristics} bank curating priority-tagged diagnostic rules that evolves via reflection, much like a physician refining their internal criteria, and (3)~a \emph{Tool Reliability Controller} that tracks per-tool trustworthiness. On ChestAgentBench, Evo-MedAgent raises multiple-choice question (MCQ) accuracy from 0.68 to 0.79 on GPT-5-mini, and from 0.76 to 0.87 on Gemini-3 Flash. With a strong base model, evolving memory improves performance more effectively than orchestrating external tools on qualitative diagnostic tasks. Because Evo-MedAgent requires no training, its per-case overhead is bounded by one additional retrieval pass and a single reflection call, making it deployable on top of any frozen model.

41. Response-Aware User Memory Selection for LLM Personalization

Authors: Jillian Fisher , Jennifer Neville , Chan Young Park
URL: https://arxiv.org/abs/2604.14473
Abstract:

A common approach to personalization in large language models (LLMs) is to incorporate a subset of the user memory into the prompt at inference time to guide the model’s generation. Existing methods select these subsets primarily using similarity between user memory items and input queries, ignoring how features actually affect the model’s response distribution. We propose Response-Utility optimization for Memory Selection (RUMS), a novel method that selects user memory items by measuring the mutual information between a subset of memory and the model’s outputs, identifying items that reduce response uncertainty and sharpen predictions beyond semantic similarity. We demonstrate that this information-theoretic foundation enables more principled user memory selection that aligns more closely with human selection compared to state-of-the-art methods, and models $400\times$ larger. Additionally, we show that memory items selected using RUMS result in better response quality compared to existing approaches, while having up to $95\%$ reduction in computational cost.

42. AIBuildAI: An AI Agent for Automatically Building AI Models

Authors: Ruiyi Zhang , Peijia Qin , Qi Cao , Li Zhang , Pengtao Xie
URL: https://arxiv.org/abs/2604.14455
Abstract:

AI models underpin modern intelligent systems, driving advances across science, medicine, finance, and technology. Yet developing high-performing AI models remains a labor-intensive process that requires expert practitioners to iteratively design architectures, engineer representations, implement training pipelines and refine approaches through empirical evaluation. Existing AutoML methods partially alleviate this burden but remain limited to narrow aspects such as hyperparameter optimization and model selection within predefined search spaces, leaving the full development lifecycle largely dependent on human expertise. To address this gap, we introduce AIBuildAI, an AI agent that automatically builds AI models from a task description and training data. AIBuildAI adopts a hierarchical agent architecture in which a manager agent coordinates three specialized sub-agents: a designer for modeling strategy, a coder for implementation and debugging, and a tuner for training and performance optimization. Each sub-agent is itself a large language model (LLM) based agent capable of multi-step reasoning and tool use, enabling end-to-end automation of the AI model development process that goes beyond the scope of existing AutoML approaches. We evaluate AIBuildAI on MLE-Bench, a benchmark of realistic Kaggle-style AI development tasks spanning visual, textual, time-series and tabular modalities. AIBuildAI ranks first on MLE-Bench with a medal rate of 63.1%, outperforming all existing baseline methods and matching the capability of highly experienced AI engineers. These results demonstrate that hierarchical agent systems can automate the full AI model development process from task specification to deployable model, suggesting a pathway toward broadly accessible AI development with minimal human intervention.

43. Credo: Declarative Control of LLM Pipelines via Beliefs and Policies

Authors: Duo Lu , Andrew Crotty , Uğur Çetintemel
URL: https://arxiv.org/abs/2604.14401
Abstract:

Agentic AI systems are becoming commonplace in domains that require long-lived, stateful decision-making in continuously evolving conditions. As such, correctness depends not only on the output of individual model calls, but also on how to best adapt when incorporating new evidence or revising prior conclusions. However, existing frameworks rely on imperative control loops, ephemeral memory, and prompt-embedded logic, making agent behavior opaque, brittle, and difficult to verify. This paper introduces Credo, which represents semantic state as beliefs and regulates behavior using declarative policies defined over these beliefs. This design supports adaptive, auditable, and composable execution through a database-backed semantic control plane. We showcase these concepts in a decision-control scenario, where beliefs and policies declaratively guide critical execution choices (e.g., model selection, retrieval, corrective re-execution), enabling dynamic behavior without requiring any changes to the underlying pipeline code.

44. Seeing Through Experts Eyes A Foundational Vision Language Model Trained on Radiologists Gaze and Reasoning

Authors: Kinhei Lee , Peiyuan Jing , Zhenxuan Zhang , Yue Yang , Tao Wang , Dominic C Marshall , Yingying Fang , Guang Yang
URL: https://arxiv.org/abs/2604.14316
Abstract:

Large scale vision language models have shown promise in automating chest Xray interpretation, yet their clinical utility remains limited by a gap between model outputs and radiologist reasoning. Most systems optimize for semantic information without emulating how experts visually examine medical images, often overlooking critical findings or diverging from established diagnostic workflows. Radiologists follow structured protocols (e.g., the ABCDEF approach) that ensure all clinically relevant regions are systematically examined, reducing missed findings and supporting reliable diagnostic reasoning. We introduce GazeX, a vision language model that leverages radiologists’ eye tracking data as a behavioral prior to model expert diagnostic reasoning. By incorporating gaze trajectories and fixation patterns into pretraining, GazeX learns to follow the spatial and temporal structure of radiologist attention and integrates observations in a clinically meaningful sequence. Using a curated dataset of over 30,000 gaze key frames from five radiologists, we demonstrate that GazeX produces more accurate, interpretable, and expert consistent outputs across radiology report generation, disease grounding, and visual question answering, utilizing 231,835 radiographic studies, 780,014 question answer pairs, and 1,162 image sentence pairs with bounding boxes. Unlike autonomous reporting systems, GazeX produces verifiable evidence artifacts, including inspection trajectories and finding linked localized regions, enabling efficient human verification and safe human AI collaboration. Learning through expert eyes provides a practical route toward more trustworthy, explainable, and diagnostically robust AI systems for radiology and beyond.

45. GFT: From Imitation to Reward Fine-Tuning with Unbiased Group Advantages and Dynamic Coefficient Rectification

Authors: Wangjie Gan , Miao Pan , Linbo Xi , Wenqi Zhang , Jintao Chen , Jianwei Yin , Xuhong Zhang
URL: https://arxiv.org/abs/2604.14258
Abstract:

Large language models are typically post-trained using supervised fine-tuning (SFT) and reinforcement learning (RL), yet effectively unifying efficient knowledge injection with robust generalization remains challenging. In this work, we provide a training-dynamics analysis showing that SFT can be interpreted as a special case of policy gradient optimization with an extremely sparse implicit reward and unstable inverse-probability weighting, which together lead to single-path dependency, entropy collapse, and gradient explosion. Motivated by this diagnosis, we propose Group Fine-Tuning (GFT), a unified post-training framework that addresses these intrinsic limitations through two mechanisms: Group Advantage Learning, which constructs diverse response groups and derives normalized contrastive supervision to alleviate reward sparsity, and Dynamic Coefficient Rectification, which adaptively bounds inverse-probability weights to stabilize optimization while preserving efficient knowledge injection. Experiments demonstrate that GFT consistently surpasses SFT-based methods and yields policies that integrate more smoothly with subsequent RL training.

46. Simulating Human Cognition: Heartbeat-Driven Autonomous Thinking Activity Scheduling for LLM-based AI systems

Authors: Hong Su
URL: https://arxiv.org/abs/2604.14178
Abstract:

Large Language Model (LLM) agents have demonstrated remarkable capabilities in reasoning and tool use, yet they often suffer from rigid, reactive control flows that limit their adaptability and efficiency. Most existing frameworks rely on fixed pipelines or failure-triggered reflection, causing agents to act impulsively or correct errors only after they occur. In this paper, we introduce Heartbeat-Driven Autonomous Thinking Activity Scheduling, a mechanism that enables proactive, adaptive, and continuous self-regulation. Mirroring the natural rhythm of human cognition, our system employs a periodic ``heartbeat’’ mechanism to orchestrate a dynamic repertoire of cognitive modules (e.g., Planner, Critic, Recaller, Dreamer). Unlike traditional approaches that rely on hard-coded symbolic rules or immediate reactive triggers, our scheduler learns to determine when to engage specific thinking activities – such as recalling memories, summarizing experiences, or strategic planning – based on temporal patterns and historical context. This functional approach allows cognitive modules to be dynamically added or removed without structural reengineering. Meanwhile, we propose a meta-learning strategy for continual policy adaptation, where the scheduler optimizes its cognitive strategy over time using historical interaction logs. Evaluation results demonstrate that our approach effectively learns to schedule cognitive activities based on historical data and can autonomously integrate new thinking modules.

47. NuHF Claw: A Risk Constrained Cognitive Agent Framework for Human Centered Procedure Support in Digital Nuclear Control Rooms

Authors: Xingyu Xiao , Jiejuan Tong , Jun Sun , Zhe Sui , Peng Chen , Jingang Liang , Haitao Wang
URL: https://arxiv.org/abs/2604.14160
Abstract:

The rapid digitization of nuclear power plant main control rooms has fundamentally reshaped operator interaction patterns, introducing complex soft-control behaviors and elevated cognitive risks that are not adequately addressed by existing human reliability analysis approaches. Although recent advances in large language models and autonomous agents offer new opportunities for intelligent decision support, their deployment in safety critical environments remains constrained by risks of hallucinated reasoning and weakened human authority. This study proposes NuHF Claw, a persistent cognitive-risk agent framework that enables risk governed human centered autonomy for digital nuclear operations. The core methodological innovation lies in the introduction of a risk constrained agent runtime, which tightly couples cognitive state inference with probabilistic safety assessment to regulate autonomous system behavior in real time. By integrating cognitively grounded workload and situational awareness estimation with dynamic human error probability prediction, the framework transforms conventional offline reliability analysis into a proactive intervention mechanism embedded directly within operational workflows. Experimental validation on a high-fidelity digital control room simulator demonstrates that NuHF Claw can anticipate interface induced cognitive degradation, dynamically constrain unsafe autonomous recommendations, and provide risk-aware navigational guidance while preserving human decision authority. The results highlight a fundamental shift from automation-driven operation toward cognition-aware autonomy, offering a principled pathway for the safe integration of intelligent agents into next-generation nuclear control environments.

48. Why Do Vision Language Models Struggle To Recognize Human Emotions?

Authors: Madhav Agarwal , Sotirios A. Tsaftaris , Laura Sevilla-Lara , Steven McDonagh
URL: https://arxiv.org/abs/2604.15280
Abstract:

Understanding emotions is a fundamental ability for intelligent systems to be able to interact with humans. Vision-language models (VLMs) have made tremendous progress in the last few years for many visual tasks, potentially offering a promising solution for understanding emotions. However, it is surprising that even the most sophisticated contemporary VLMs struggle to recognize human emotions or to outperform even specialized vision-only classifiers. In this paper we ask the question “Why do VLMs struggle to recognize human emotions?”, and observe that the inherently continuous and dynamic task of facial expression recognition (DFER) exposes two critical VLM vulnerabilities. First, emotion datasets are naturally long-tailed, and the web-scale data used to pre-train VLMs exacerbates this head-class bias, causing them to systematically collapse rare, under-represented emotions into common categories. We propose alternative sampling strategies that prevent favoring common concepts. Second, temporal information is critical for understanding emotions. However, VLMs are unable to represent temporal information over dense frame sequences, as they are limited by context size and the number of tokens that can fit in memory, which poses a clear challenge for emotion recognition. We demonstrate that the sparse temporal sampling strategy used in VLMs is inherently misaligned with the fleeting nature of micro-expressions (0.25-0.5 seconds), which are often the most critical affective signal. As a diagnostic probe, we propose a multi-stage context enrichment strategy that utilizes the information from “in-between” frames by first converting them into natural language summaries. This enriched textual context is provided as input to the VLM alongside sparse keyframes, preventing attentional dilution from excessive visual data while preserving the emotional trajectory.

49. Prism: Symbolic Superoptimization of Tensor Programs

Authors: Mengdi Wu , Xiaoyu Jiang , Oded Padon , Zhihao Jia
URL: https://arxiv.org/abs/2604.15272
Abstract:

This paper presents Prism, the first symbolic superoptimizer for tensor programs. The key idea is sGraph, a symbolic, hierarchical representation that compactly encodes large classes of tensor programs by symbolically representing some execution parameters. Prism organizes optimization as a two-level search: it constructs symbolic graphs that represent families of programs, and then instantiates them into concrete implementations. This formulation enables structured pruning of provably suboptimal regions of the search space using symbolic reasoning over operator semantics, algebraic identities, and hardware constraints. We develop techniques for efficient symbolic graph generation, equivalence verification via e-graph rewriting, and parameter instantiation through auto-tuning. Together, these components allow Prism to bridge the rigor of exhaustive search with the scalability required for modern ML workloads. Evaluation on five commonly used LLM workloads shows that Prism achieves up to $2.2\times$ speedup over best superoptimizers and $4.9\times$ over best compiler-based approaches, while reducing end-to-end optimization time by up to $3.4\times$.

Authors: Emanuel Tewolde , Xiao Zhang , David Guzman Piedrahita , Vincent Conitzer , Zhijing Jin
URL: https://arxiv.org/abs/2604.15267
Abstract:

It is increasingly important that LLM agents interact effectively and safely with other goal-pursuing agents, yet, recent works report the opposite trend: LLMs with stronger reasoning capabilities behave less cooperatively in mixed-motive games such as the prisoner’s dilemma and public goods settings. Indeed, our experiments show that recent models – with or without reasoning enabled – consistently defect in single-shot social dilemmas. To tackle this safety concern, we present the first comparative study of game-theoretic mechanisms that are designed to enable cooperative outcomes between rational agents in equilibrium. Across four social dilemmas testing distinct components of robust cooperation, we evaluate the following mechanisms: (1) repeating the game for many rounds, (2) reputation systems, (3) third-party mediators to delegate decision making to, and (4) contract agreements for outcome-conditional payments between players. Among our findings, we establish that contracting and mediation are most effective in achieving cooperative outcomes between capable LLM models, and that repetition-induced cooperation deteriorates drastically when co-players vary. Moreover, we demonstrate that these cooperation mechanisms become more effective under evolutionary pressures to maximize individual payoffs.

51. VisPCO: Visual Token Pruning Configuration Optimization via Budget-Aware Pareto-Frontier Learning for Vision-Language Models

Authors: Huawei Ji , Yuanhao Sun , Yuan Jin , Cheng Deng , Jiaxin Ding , Luoyi Fu , Xinbing Wang
URL: https://arxiv.org/abs/2604.15188
Abstract:

Visual token pruning methods effectively mitigate the quadratic computational growth caused by processing high-resolution images and video frames in vision-language models (VLMs). However, existing approaches rely on predefined pruning configurations without determining whether they achieve computation-performance optimality. In this work, we introduce , a novel framework that formulates visual token pruning as a Pareto configuration optimization problem to automatically identify optimal configurations. Our approach employs continuous relaxation and straight-through estimators to enable gradient-based search, solved via the Augmented Lagrangian method. Extensive experiments across 8 visual benchmarks demonstrate that effectively approximates the empirical Pareto frontier obtained through grid search and generalizes well across various pruning methods and VLM architectures. Furthermore, through learnable kernel functions, we investigate layer-wise pruning patterns and reveal that multi-step progressive pruning captures VLMs’ hierarchical compression structure, achieving superior accuracy-efficiency trade-offs compared to single-layer approaches.

52. Scepsy: Serving Agentic Workflows Using Aggregate LLM Pipelines

Authors: Marcel Wagenländer , Otto White , Britannio Jarrett , Pedro Silvestre , Yanda Tao , Guo Li , Huanzhou Zhu , Llúis Vilanova , Peter Pietzuch
URL: https://arxiv.org/abs/2604.15186
Abstract:

Agentic workflows carry out complex tasks by orchestrating multiple large language models (LLMs) and tools. Serving such workflows at a target throughput with low latency is challenging because they can be defined using arbitrary agentic frameworks and exhibit unpredictable execution times: execution may branch, fan-out, or recur in data-dependent ways. Since LLMs in workflows often outnumber available GPUs, their execution also leads to GPU oversubscription. We describe Scepsy, a new agentic serving system that efficiently schedules arbitrary multi-LLM agentic workflows onto a GPU cluster. Scepsy exploits the insight that, while agentic workflows have unpredictable end-to-end latencies, the shares of each LLM’s total execution times are comparatively stable across executions. Scepsy decides on GPU allocations based on these aggregate shares: first, it profiles the LLMs under different parallelism degrees. It then uses these statistics to construct an Aggregate LLM Pipeline, which is a lightweight latency/throughput predictor for allocations. To find a GPU allocation that minimizes latency while achieving a target throughput, Scepsy uses the Aggregate LLM Pipeline to explore a search space over fractional GPU shares, tensor parallelism degrees, and replica counts. It uses a hierarchical heuristic to place the best allocation onto the GPU cluster, minimizing fragmentation, while respecting network topology constraints. Our evaluation on realistic agentic workflows shows that Scepsy achieves up to 2.4x higher throughput and 27x lower latency compared to systems that optimize LLMs independently or rely on user-specified allocations.

53. Compressing Sequences in the Latent Embedding Space: $K$-Token Merging for Large Language Models

Authors: Zihao Xu , John Harvill , Ziwei Fan , Yizhou Sun , Hao Ding , Hao Wang
URL: https://arxiv.org/abs/2604.15153
Abstract:

Large Language Models (LLMs) incur significant computational and memory costs when processing long prompts, as full self-attention scales quadratically with input length. Token compression aims to address this challenge by reducing the number of tokens representing inputs. However, existing prompt-compression approaches primarily operate in token space and overlook inefficiencies in the latent embedding space. In this paper, we propose K-Token Merging, a latent-space compression framework that merges each contiguous block of K token embeddings into a single embedding via a lightweight encoder. The compressed sequence is processed by a LoRA-adapted LLM, while generation remains in the original vocabulary. Experiments on structural reasoning (Textualized Tree), sentiment classification (Amazon Reviews), and code editing (CommitPackFT) show that K-Token Merging lies on the Pareto frontier of performance vs. compression, achieving up to 75% input length reduction with minimal performance degradation.

54. IUQ: Interrogative Uncertainty Quantification for Long-Form Large Language Model Generation

Authors: Haozhi Fan , Jinhao Duan , Kaidi Xu
URL: https://arxiv.org/abs/2604.15109
Abstract:

Despite the rapid advancement of Large Language Models (LLMs), uncertainty quantification in LLM generation is a persistent challenge. Although recent approaches have achieved strong performance by restricting LLMs to produce short or constrained answer sets, many real-world applications require long-form and free-form text generation. A key difficulty in this setting is that LLMs often produce responses that are semantically coherent yet factually inaccurate, while the underlying semantics are multifaceted and the linguistic structure is complex. To tackle this challenge, this paper introduces Interrogative Uncertainty Quantification (IUQ), a novel framework that leverages inter-sample consistency and intra-sample faithfulness to quantify the uncertainty in long-form LLM outputs. By utilizing an interrogate-then-respond paradigm, our method provides reliable measures of claim-level uncertainty and the model’s faithfulness. Experimental results across diverse model families and model sizes demonstrate the superior performance of IUQ over two widely used long-form generation datasets. The code is available at this https URL .

55. Autonomous Evolution of EDA Tools: Multi-Agent Self-Evolved ABC

Authors: Cunxi Yu , Haoxing Ren
URL: https://arxiv.org/abs/2604.15082
Abstract:

This paper introduces the first \emph{self-evolving} logic synthesis framework, which leverages Large Language Model (LLM) agents to autonomously improve the source code of \textsc{ABC}, the widely adopted logic synthesis system. Our framework operates on the \emph{entire integrated ABC codebase}, and the output repository preserves its single-binary execution model and command interface. In the initial evolution cycle, we bootstrap the system using existing prior open-source synthesis components, covering flow tuning, logic minimization, and technology mapping, but without manually injecting new heuristics. On top of this foundation, a team of LLM-based agents iteratively rewrites and evolves specific sub-components of ABC following our programming guidance prompts under a unified correctness and QoR-driven evaluation loop. Each evolution cycle proposes code modifications, compiles the integrated binary, validates correctness, and evaluates quality-of-results (QoR) on \emph{multi-suite benchmarks including ISCAS~85/89/99, VTR, EPFL, and IWLS~2005}. Through continuous feedback, the system discovers optimizations beyond human-designed heuristics, effectively \emph{learning new synthesis strategies} that enhance QoR. We detail the architecture of this self-improving system, its integration with \textsc{ABC}, and results demonstrating that the framework can autonomously and progressively improve EDA tool at full million-line scale.

56. Route to Rome Attack: Directing LLM Routers to Expensive Models via Adversarial Suffix Optimization

Authors: Haochun Tang , Yuliang Yan , Jiahua Lu , Huaxiao Liu , Enyan Dai
URL: https://arxiv.org/abs/2604.15022
Abstract:

Cost-aware routing dynamically dispatches user queries to models of varying capability to balance performance and inference cost. However, the routing strategy introduces a new security concern that adversaries may manipulate the router to consistently select expensive high-capability models. Existing routing attacks depend on either white-box access or heuristic prompts, rendering them ineffective in real-world black-box scenarios. In this work, we propose R$^2$A, which aims to mislead black-box LLM routers to expensive models via adversarial suffix optimization. Specifically, R$^2$A deploys a hybrid ensemble surrogate router to mimic the black-box router. A suffix optimization algorithm is further adapted for the ensemble-based surrogate. Extensive experiments on multiple open-source and commercial routing systems demonstrate that {R$^2$A} significantly increases the routing rate to expensive models on queries of different distributions. Code and examples: this https URL .

57. UniDoc-RL: Coarse-to-Fine Visual RAG with Hierarchical Actions and Dense Rewards

Authors: Jun Wang , Shuo Tan , Zelong Sun , Tiancheng Gu , Yongle Zhao , Ziyong Feng , Kaicheng Yang , Cewu Lu
URL: https://arxiv.org/abs/2604.14967
Abstract:

Retrieval-Augmented Generation (RAG) extends Large Vision-Language Models (LVLMs) with external visual knowledge. However, existing visual RAG systems typically rely on generic retrieval signals that overlook the fine-grained visual semantics essential for complex reasoning. To address this limitation, we propose UniDoc-RL, a unified reinforcement learning framework in which an LVLM agent jointly performs retrieval, reranking, active visual perception, and reasoning. UniDoc-RL formulates visual information acquisition as a sequential decision-making problem with a hierarchical action space. Specifically, it progressively refines visual evidence from coarse-grained document retrieval to fine-grained image selection and active region cropping, allowing the model to suppress irrelevant content and attend to information-dense regions. For effective end-to-end training, we introduce a dense multi-reward scheme that provides task-aware supervision for each action. Based on Group Relative Policy Optimization (GRPO), UniDoc-RL aligns agent behavior with multiple objectives without relying on a separate value network. To support this training paradigm, we curate a comprehensive dataset of high-quality reasoning trajectories with fine-grained action annotations. Experiments on three benchmarks demonstrate that UniDoc-RL consistently surpasses state-of-the-art baselines, yielding up to 17.7% gains over prior RL-based methods.

58. Calibration-Gated LLM Pseudo-Observations for Online Contextual Bandits

Authors: Maksim Pershin , Ivan Golovanov , Pavel Baltabaev , Natalia Trankova
URL: https://arxiv.org/abs/2604.14961
Abstract:

Contextual bandit algorithms suffer from high regret during cold-start, when the learner has insufficient data to distinguish good arms from bad. We propose augmenting Disjoint LinUCB with LLM pseudo-observations: after each round, a large language model predicts counterfactual rewards for the unplayed arms, and these predictions are injected into the learner as weighted pseudo-observations. The injection weight is controlled by a calibration-gated decay schedule that tracks the LLM’s prediction accuracy on played arms via an exponential moving average; high calibration error suppresses the LLM’s influence, while accurate predictions receive higher weight during the critical early rounds. We evaluate on two contextual bandit environments - UCI Mushroom (2-arm, asymmetric rewards) and MIND-small (5-arm news recommendation) - and find that when equipped with a task-specific prompt, LLM pseudo-observations reduce cumulative regret by 19% on MIND relative to pure LinUCB. However, generic counterfactual prompt framing increases regret on both environments, demonstrating that prompt design is the dominant factor, more important than the choice of decay schedule or calibration gating parameters. We analyze the failure modes of calibration gating on domains with small prediction errors and provide a theoretical motivation for the bias-variance trade-off governing pseudo-observation weight.

59. RaTA-Tool: Retrieval-based Tool Selection with Multimodal Large Language Models

Authors: Gabriele Mattioli , Evelyn Turri , Sara Sarto , Lorenzo Baraldi , Marcella Cornia , Lorenzo Baraldi , Rita Cucchiara
URL: https://arxiv.org/abs/2604.14951
Abstract:

Tool learning with foundation models aims to endow AI systems with the ability to invoke external resources – such as APIs, computational utilities, and specialized models – to solve complex tasks beyond the reach of standalone language generation. While recent advances in Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) have expanded their reasoning and perception capabilities, existing tool-use methods are predominantly limited to text-only inputs and closed-world settings. Consequently, they struggle to interpret multimodal user instructions and cannot generalize to tools unseen during training. In this work, we introduce RaTA-Tool, a novel framework for open-world multimodal tool selection. Rather than learning direct mappings from user queries to fixed tool identifiers, our approach enables an MLLM to convert a multimodal query into a structured task description and subsequently retrieve the most appropriate tool by matching this representation against semantically rich, machine-readable tool descriptions. This retrieval-based formulation naturally supports extensibility to new tools without retraining. To further improve alignment between task descriptions and tool selection, we incorporate a preference-based optimization stage using Direct Preference Optimization (DPO). To support research in this setting, we also introduce the first dataset for open-world multimodal tool use, featuring standardized tool descriptions derived from Hugging Face model cards. Extensive experiments demonstrate that our approach significantly improves tool-selection performance, particularly in open-world, multimodal scenarios.

60. Can LLMs Score Medical Diagnoses and Clinical Reasoning as well as Expert Panels?

Authors: Amy Rouillard , Sitwala Mundiab , Linda Camarab , Michael Cameron Gramaniec , Ziyaad Dangorc , Ismail Kallad , Shabir A. Madhic , Kajal Morarc , Marlvin T. Ncubec , Haroon Saloojeee , Bruce A. Bassett
URL: https://arxiv.org/abs/2604.14892
Abstract:

Evaluating medical AI systems using expert clinician panels is costly and slow, motivating the use of large language models (LLMs) as alternative adjudicators. Here, we evaluate an LLM jury composed of three frontier AI models scoring 3333 diagnoses on 300 real-world middle-income country (MIC) hospital cases. Model performance was benchmarked against expert clinician panel and independent human re-scoring panel evaluations. Both LLM and clinician-generated diagnoses are scored across four dimensions: diagnosis, differential diagnosis, clinical reasoning and negative treatment risk. For each of these, we assess scoring difference, inter-rater agreement, scoring stability, severe safety errors and the effect of post-hoc calibration. We find that: (i) the uncalibrated LLM jury scores are systematically lower than clinician panels scores; (ii) the LLM Jury preserves ordinal agreement and exhibits better concordance with the primary expert panels than the human expert re-score panels do; (iii) the probability of severe errors is lower in \lj models compared to the human expert re-score panels; (iv) the LLM Jury shows excellent agreement with primary expert panels’ rankings. We find that the LLM jury combined with AI model diagnoses can be used to identify ward diagnoses at high risk of error, enabling targeted expert review and improved panel efficiency; (v) LLM jury models show no self-preference bias. They did not score diagnoses generated by their own underlying model or models from the same vendor more (or less) favourably than those generated by other models. Finally, we demonstrate that LLM jury calibration using isotonic regression improves alignment with human expert panel evaluations. Together, these results provide compelling evidence that a calibrated, multi-model LLM jury can serve as a trustworthy and reliable proxy for expert clinician evaluation in medical AI benchmarking.

61. Reasoning Dynamics and the Limits of Monitoring Modality Reliance in Vision-Language Models

Authors: Danae Sánchez Villegas , Samuel Lewis-Lim , Nikolaos Aletras , Desmond Elliott
URL: https://arxiv.org/abs/2604.14888
Abstract:

Recent advances in vision language models (VLMs) offer reasoning capabilities, yet how these unfold and integrate visual and textual information remains unclear. We analyze reasoning dynamics in 18 VLMs covering instruction-tuned and reasoning-trained models from two different model families. We track confidence over Chain-of-Thought (CoT), measure the corrective effect of reasoning, and evaluate the contribution of intermediate reasoning steps. We find that models are prone to answer inertia, in which early commitments to a prediction are reinforced, rather than revised during reasoning steps. While reasoning-trained models show stronger corrective behavior, their gains depend on modality conditions, from text-dominant to vision-only settings. Using controlled interventions with misleading textual cues, we show that models are consistently influenced by these cues even when visual evidence is sufficient, and assess whether this influence is recoverable from CoT. Although this influence can appear in the CoT, its detectability varies across models and depends on what is being monitored. Reasoning-trained models are more likely to explicitly refer to the cues, but their longer and fluent CoTs can still appear visually grounded while actually following textual cues, obscuring modality reliance. In contrast, instruction-tuned models refer to the cues less explicitly, but their shorter traces reveal inconsistencies with the visual input. Taken together, these findings indicate that CoT provides only a partial view of how different modalities drive VLM decisions, with important implications for the transparency and safety of multimodal systems.

62. RACER: Retrieval-Augmented Contextual Rapid Speculative Decoding

Authors: Zihong Zhang , Zuchao Li , Lefei Zhang , Ping Wang , Hai Zhao
URL: https://arxiv.org/abs/2604.14885
Abstract:

Autoregressive decoding in Large Language Models (LLMs) generates one token per step, causing high inference latency. Speculative decoding (SD) mitigates this through a guess-and-verify strategy, but existing training-free variants face trade-offs: retrieval-based drafts break when no exact match exists, while logits-based drafts lack structural guidance. We propose $\textbf{RACER}$ ($\textbf{R}$etrieval-$\textbf{A}$ugmented $\textbf{C}$ont$\textbf{e}$xtual $\textbf{R}$apid Speculative Decoding), a lightweight and training-free method that integrates retrieved exact patterns with logit-driven future cues. This unification supplies both reliable anchors and flexible extrapolation, yielding richer speculative drafts. Experiments on Spec-Bench, HumanEval, and MGSM-ZH demonstrate that RACER consistently accelerates inference, achieving more than $2\times$ speedup over autoregressive decoding, and outperforms prior training-free methods, offering a scalable, plug-and-play solution for efficient LLM decoding. Our source code is available at $\href{ this https URL }{ this https URL }$.

63. Vibe-Coding: Feedback-Based Automated Verification with no Human Code Inspection, a Feasibility Study

Authors: Michal Töpfer , František Plášil , Tomáš Bureš , Petr Hnětynka
URL: https://arxiv.org/abs/2604.14867
Abstract:

Vibe coding inherently assumes iterative refinement of LLM-generated code through feedback loops. While effective for conventional software tasks, its reliability in runtime-adaptive systems is unclear – especially when generated code is not manually inspected. This paper studies feedback-based automated verification of LLM-generated adaptation managers in Collective Adaptive Systems (CAS). We focus on the key challenges of verification in the loop: how to detect failures of generated code at runtime and how to report them precisely enough for an LLM to fix them. We combine the adaptation loop with a vibe-coding feedback loop where correctness is checked against (i) generic architectural constraints and (ii) functional constraints formalized in Functional Constraints Logic (FCL), a novel first-order temporal logic over potentially finite traces. Conducting the Dragon Hunt CAS case study, we show that fine-grained constraint violations provide actionable feedback that typically yields a valid adaptation manager within a few iterations, while simple coarse metric-based feedback often stalls. Our findings suggest that feedback precision is the dominant factor for reliable vibe coding in systems designed by domain experts with no programming skills, thereby obviating the need for human code inspection.

64. MetaDent: Labeling Clinical Images for Vision-Language Models in Dentistry

Authors: Meng-Xun Li , Wen-Hui Deng , Zhi-Xing Wu , Chun-Xiao Jin , Jia-Min Wu , Yue Han , James Kit Hon Tsoi , Gui-Song Xia , Cui Huang
URL: https://arxiv.org/abs/2604.14866
Abstract:

Vision-Language Models (VLMs) have demonstrated significant potential in medical image analysis, yet their application in intraoral photography remains largely underexplored due to the lack of fine-grained, annotated datasets and comprehensive benchmarks. To address this, we present MetaDent, a comprehensive resource that includes (1) a novel and large-scale dentistry image dataset collected from clinical, public, and web sources; (2) a semi-structured annotation framework designed to capture the hierarchical and clinically nuanced nature of dental photography; and (3) comprehensive benchmark suites for evaluating state-of-the-art VLMs on clinical image understanding. Our labeling approach combines a high-level image summary with point-by-point, free-text descriptions of abnormalities. This method enables rich, scalable, and task-agnostic representations. We curated 60,669 dental images from diverse sources and annotated a representative subset of 2,588 images using this meta-labeling scheme. Leveraging Large Language Models (LLMs), we derive standardized benchmarks: approximately 15K Visual Question Answering (VQA) pairs and an 18-class multi-label classification dataset, which we validated with human review and error analysis to justify that the LLM-driven transition reliably preserves fidelity and semantic accuracy. We then evaluate state-of-the-art VLMs across VQA, classification, and image captioning tasks. Quantitative results reveal that even the most advanced models struggle with a fine-grained understanding of intraoral scenes, achieving moderate accuracy and producing inconsistent or incomplete descriptions in image captioning. We publicly release our dataset, annotations, and tools to foster reproducible research and accelerate the development of vision-language systems for dental applications.

65. Schema Key Wording as an Instruction Channel in Structured Generation under Constrained Decoding

Authors: Yifan Le
URL: https://arxiv.org/abs/2604.14862
Abstract:

Constrained decoding has been widely adopted for structured generation with large language models (LLMs), ensuring that outputs satisfy predefined formats such as JSON and XML. However, existing approaches largely treat schemas as purely structural constraints and overlook the possibility that their linguistic formulation may affect model behavior. In this work, we study how instruction placement influences model performance in structured generation and show that merely changing the wording of schema keys, without modifying the prompt or model parameters, can significantly alter model performance under constrained decoding. Based on this observation, we propose to reinterpret structured generation as a multi-channel instruction problem, where instructions can be conveyed explicitly through prompts and implicitly through schema keys during decoding. To the best of our knowledge, this is the first work to systematically study how schema key formulation acts as an implicit instruction channel and affects model performance under constrained decoding. Experiments on multiple mathematical reasoning benchmarks show that different model families exhibit distinct sensitivities to these instruction channels: Qwen models consistently benefit from schema-level instructions, while LLaMA models rely more heavily on prompt-level guidance. We further observe non-additive interaction effects between instruction channels, showing that combining multiple channels does not always lead to further improvement. These findings suggest that schema design not only determines output structure, but also carries instruction signals, offering a new perspective on structured generation in LLMs.

66. ClimateCause: Complex and Implicit Causal Structures in Climate Reports

Authors: Liesbeth Allein , Nataly Pineda-Castañeda , Andrea Rocci , Marie-Francine Moens
URL: https://arxiv.org/abs/2604.14856
Abstract:

Understanding climate change requires reasoning over complex causal networks. Yet, existing causal discovery datasets predominantly capture explicit, direct causal relations. We introduce ClimateCause, a manually expert-annotated dataset of higher-order causal structures from science-for-policy climate reports, including implicit and nested causality. Cause-effect expressions are normalized and disentangled into individual causal relations to facilitate graph construction, with unique annotations for cause-effect correlation, relation type, and spatiotemporal context. We further demonstrate ClimateCause’s value for quantifying readability based on the semantic complexity of causal graphs underlying a statement. Finally, large language model benchmarking on correlation inference and causal chain reasoning highlights the latter as a key challenge.

67. Zero-Shot Retail Theft Detection via Orchestrated Vision Models: A Model-Agnostic, Cost-Effective Alternative to Trained Single-Model Systems

Authors: Haileab Yagersew
URL: https://arxiv.org/abs/2604.14846
Abstract:

Retail theft costs the global economy over $100 billion annually, yet existing AI-based detection systems require expensive custom model training on proprietary datasets and charge $200-500/month per store. We present Paza, a zero-shot retail theft detection framework that achieves practical concealment detection without training any model. Our approach orchestrates multiple existing models in a layered pipeline - cheap object detection and pose estimation running continuously, with an expensive vision-language model (VLM) invoked only when behavioral pre-filters trigger. A multi-signal suspicion pre-filter (requiring dwell time plus at least one behavioral signal) reduces VLM invocations by 240x compared to per-frame analysis, bounding calls to <=10/minute and enabling a single GPU to serve 10-20 stores. The architecture is model-agnostic: the VLM component accepts any OpenAI-compatible endpoint, enabling operators to swap between models such as Gemma 4, Qwen3.5-Omni, GPT-4o, or future releases without code changes - ensuring the system improves as the VLM landscape evolves. We evaluate the VLM component on the DCSASS synthesized shoplifting dataset (169 clips, controlled environment), achieving 89.5% precision and 92.8% specificity at 59.3% recall zero-shot - where the recall gap is attributable to sparse frame sampling in offline evaluation rather than VLM reasoning failures, as precision and specificity are the operationally critical metrics determining false alarm rates. We present a detailed cost model showing viability at $50-100/month per store (3-10x cheaper than commercial alternatives), and introduce a privacy-preserving design that obfuscates faces in the detection pipeline. The source code is available at this https URL .

Authors: Midan Shim , Seokju Hwang , Kaehyun Um , Kyong-Ho Lee
URL: https://arxiv.org/abs/2604.14749
Abstract:

Large language models still struggle with faithfulness and hallucinations despite their remarkable reasoning abilities. In Knowledge Graph Question Answering (KGQA), semantic parsing-based approaches address the limitations by understanding constraints in a user’s question and converting them into a logical form to execute on a knowledge graph. However, existing KGQA benchmarks and methods are biased toward positive and calculation constraints. Negative constraints are neglected, although they frequently appear in real-world questions. In this paper, we introduce a new task, NEgative-conSTrained (NEST) KGQA, where each question contains at least one negative constraint, and a corresponding dataset, NestKGQA. We also design PyLF, a Python-formatted logical form, since existing logical forms are hardly suitable to express negation clearly while maintaining readability. Furthermore, NEST questions naturally contain multiple constraints. To mitigate their semantic complexity, we present a novel framework named CUCKOO, specialized to multiple-constrained questions and ensuring semantic executability. CUCKOO first generates a constraint-aware logical form draft and performs schema-guided semantic matching. It then selectively applies self-directed refinement only when executing improper logical forms yields an empty result, reducing cost while improving robustness. Experimental results demonstrate that CUCKOO consistently outperforms baselines on both conventional KGQA and NEST-KGQA benchmarks under few-shot settings.

69. Bounded Autonomy for Enterprise AI: Typed Action Contracts and Consumer-Side Execution

Authors: Sarmad Sohail , Ghufran Haider
URL: https://arxiv.org/abs/2604.14723
Abstract:

Large language models are increasingly used as natural-language interfaces to enterprise software, but their direct use as system operators remains unsafe. Model errors can propagate into unauthorized actions, malformed requests, cross-workspace execution, and other costly failures. We argue this is primarily an execution architecture problem. We present a bounded-autonomy architecture in which language models may interpret intent and propose actions, but all executable behavior is constrained by typed action contracts, permission-aware capability exposure, scoped context, validation before side effects, consumer-side execution boundaries, and optional human approval. The enterprise application remains the source of truth for business logic and authorization, while the orchestration engine operates over an explicit published actions manifest. We evaluate the architecture in a deployed multi-tenant enterprise application across three conditions: manual operation, unconstrained AI with safety layers disabled, and full bounded autonomy. Across 25 scenario trials spanning seven failure families, the bounded-autonomy system completed 23 of 25 tasks with zero unsafe executions, while the unconstrained configuration completed only 17 of 25. Two wrong-entity mutations escaped all consumer-contributed layers; only disambiguation and confirmation mechanisms intercept this class. Both AI conditions delivered 13-18x speedup over manual operation. Critically, removing safety layers made the system less useful: structured validation feedback guided the model to correct outcomes in fewer turns, while the unconstrained system hallucinated success. Several safety properties are structurally enforced by code and intercepted all targeted violations regardless of model output. The result is a practical, deployed architecture for making imperfect language models operationally useful in enterprise systems.

70. Fact4ac at the Financial Misinformation Detection Challenge Task: Reference-Free Financial Misinformation Detection via Fine-Tuning and Few-Shot Prompting of Large Language Models

Authors: Cuong Hoang , Le-Minh Nguyen
URL: https://arxiv.org/abs/2604.14640
Abstract:

The proliferation of financial misinformation poses a severe threat to market stability and investor trust, misleading market behavior and creating critical information asymmetry. Detecting such misleading narratives is inherently challenging, particularly in real-world scenarios where external evidence or supplementary references for cross-verification are strictly unavailable. This paper presents our winning methodology for the “Reference-Free Financial Misinformation Detection” shared task. Built upon the recently proposed RFC-BENCH framework (Jiang et al. 2026), this task challenges models to determine the veracity of financial claims by relying solely on internal semantic understanding and contextual consistency, rather than external fact-checking. To address this formidable evaluation setup, we propose a comprehensive framework that capitalizes on the reasoning capabilities of state-of-the-art Large Language Models (LLMs). Our approach systematically integrates in-context learning, specifically zero-shot and few-shot prompting strategies, with Parameter-Efficient Fine-Tuning (PEFT) via Low-Rank Adaptation (LoRA) to optimally align the models with the subtle linguistic cues of financial manipulation. Our proposed system demonstrated superior efficacy, successfully securing the first-place ranking on both official leaderboards. Specifically, we achieved an accuracy of 95.4% on the public test set and 96.3% on the private test set, highlighting the robustness of our method and contributing to the acceleration of context-aware misinformation detection in financial Natural Language Processing. Our models (14B and 32B) are available at this https URL .

71. StoryCoder: Narrative Reformulation for Structured Reasoning in LLM Code Generation

Authors: Geonhui Jang , Dongyoon Han , YoungJoon Yoo
URL: https://arxiv.org/abs/2604.14631
Abstract:

Effective code generation requires both model capability and a problem representation that carefully structures how models reason and plan. Existing approaches augment reasoning steps or inject specific structure into how models think, but leave scattered problem conditions unchanged. Inspired by the way humans organize fragmented information into coherent explanations, we propose StoryCoder, a narrative reformulation framework that transforms code generation questions into coherent natural language narratives, providing richer contextual structure than simple rephrasings. Each narrative consists of three components: a task overview, constraints, and example test cases, guided by the selected algorithm and genre. Experiments across 11 models on HumanEval, LiveCodeBench, and CodeForces demonstrate consistent improvements, with an average gain of 18.7% in zero-shot pass@10. Beyond accuracy, our analyses reveal that narrative reformulation guides models toward correct algorithmic strategies, reduces implementation errors, and induces a more modular code structure. The analyses further show that these benefits depend on narrative coherence and genre alignment, suggesting that structured problem representation is important for code generation regardless of model scale or architecture. Our code is available at this https URL .

72. ELMoE-3D: Leveraging Intrinsic Elasticity of MoE for Hybrid-Bonding-Enabled Self-Speculative Decoding in On-Premises Serving

Authors: Yuseon Choi , Jingu Lee , Jungjun Oh , Sunjoo Whang , Byeongcheol Kim , Minsung Kim , Hoi-Jun Yoo , Sangjin Kim
URL: https://arxiv.org/abs/2604.14626
Abstract:

Mixture-of-Experts (MoE) models have become the dominant architecture for large-scale language models, yet on-premises serving remains fundamentally memory-bound as batching turns sparse per-token compute into dense memory activation. Memory-centric architectures (PIM, NMP) improve bandwidth but leave compute underutilized under MoE’s low arithmetic intensity at high batch sizes. Speculative decoding (SD) trades idle compute for fewer target invocations, yet verification must load experts even for rejected tokens, severely limiting its benefit in MoE especially at low batch sizes. We propose ELMoE-3D, a hybrid-bonding (HB)-based HW-SW co-designed framework that unifies cache-based acceleration and speculative decoding to offer overall speedup across batch sizes. We identify two intrinsic elasticity axes of MoE-expert and bit-and jointly scale them to construct Elastic Self-Speculative Decoding (Elastic-SD), which serves as both an expert cache and a strongly aligned self-draft model accelerated by high HB bandwidth. Our LSB-augmented bit-sliced architecture exploits inherent redundancy in bit-slice representations to natively support bit-nested execution. On our 3D-stacked hardware, ELMoE-3D achieves an average $6.6\times$ speedup and $4.4\times$ energy efficiency gain over naive MoE serving on xPU across batch sizes 1-16, and delivers $2.2\times$ speedup and $1.4\times$ energy efficiency gain over the best-performing prior accelerator baseline.

73. Retrieve, Then Classify: Corpus-Grounded Automation of Clinical Value Set Authoring

Authors: Sumit Mukherjee , Juan Shu , Nairwita Mazumder , Tate Kernell , Celena Wheeler , Shannon Hastings , Chris Sidey-Gibbons
URL: https://arxiv.org/abs/2604.14616
Abstract:

Clinical value set authoring – the task of identifying all codes in a standardized vocabulary that define a clinical concept – is a recurring bottleneck in clinical quality measurement and phenotyping. A natural approach is to prompt a large language model (LLM) to generate the required codes directly, but structured clinical vocabularies are large, version-controlled, and not reliably memorized during pretraining. We propose Retrieval-Augmented Set Completion (RASC): retrieve the $K$ most similar existing value sets from a curated corpus to form a candidate pool, then apply a classifier to each candidate code. Theoretically, retrieve-and-select can reduce statistical complexity by shrinking the effective output space from the full vocabulary to a much smaller retrieved candidate pool. We demonstrate the utility of RASC on 11,803 publicly available VSAC value sets, constructing the first large-scale benchmark for this task. A cross-encoder fine-tuned on SAPBert achieves AUROC~0.852 and value-set-level F1~0.298, outperforming a simpler three-layer Multilayer Perceptron (AUROC~0.799, F1~0.250) and both reduce the number of irrelevant candidates per true positive from 12.3 (retrieval-only) to approximately 3.2 and 4.4 respectively. Zero-shot GPT-4o achieves value-set-level F1~0.105, with 48.6\% of returned codes absent from VSAC entirely. This performance gap widens with increasing value set size, consistent with RASC’s theoretical advantage. We observe similar performance gains across two other classifier model types, namely a cross-encoder initialized from pre-trained SAPBert and a LightGBM model, demonstrating that RASC’s benefits extend beyond a single model class. The code to download and create the benchmark dataset, as well as the model training code is available at: \href{ this https URL }{ this https URL }.

74. Hijacking Large Audio-Language Models via Context-Agnostic and Imperceptible Auditory Prompt Injection

Authors: Meng Chen , Kun Wang , Li Lu , Jiaheng Zhang , Tianwei Zhang
URL: https://arxiv.org/abs/2604.14604
Abstract:

Modern Large audio-language models (LALMs) power intelligent voice interactions by tightly integrating audio and text. This integration, however, expands the attack surface beyond text and introduces vulnerabilities in the continuous, high-dimensional audio channel. While prior work studied audio jailbreaks, the security risks of malicious audio injection and downstream behavior manipulation remain underexamined. In this work, we reveal a previously overlooked threat, auditory prompt injection, under realistic constraints of audio data-only access and strong perceptual stealth. To systematically analyze this threat, we propose \textit{AudioHijack}, a general framework that generates context-agnostic and imperceptible adversarial audio to hijack LALMs. \textit{AudioHijack} employs sampling-based gradient estimation for end-to-end optimization across diverse models, bypassing non-differentiable audio tokenization. Through attention supervision and multi-context training, it steers model attention toward adversarial audio and generalizes to unseen user contexts. We also design a convolutional blending method that modulates perturbations into natural reverberation, making them highly imperceptible to users. Extensive experiments on 13 state-of-the-art LALMs show consistent hijacking across 6 misbehavior categories, achieving average success rates of 79\%-96\% on unseen user contexts with high acoustic fidelity. Real-world studies demonstrate that commercial voice agents from Mistral AI and Microsoft Azure can be induced to execute unauthorized actions on behalf of users. These findings expose critical vulnerabilities in LALMs and highlight the urgent need for dedicated defense.

75. CausalDetox: Causal Head Selection and Intervention for Language Model Detoxification

Authors: Yian Wang , Yuen Chen , Agam Goyal , Hari Sundaram
URL: https://arxiv.org/abs/2604.14602
Abstract:

Large language models (LLMs) frequently generate toxic content, posing significant risks for safe deployment. Current mitigation strategies often degrade generation quality or require costly human annotation. We propose CAUSALDETOX, a framework that identifies and intervenes on the specific attention heads causally responsible for toxic generation. Using the Probability of Necessity and Sufficiency (PNS), we isolate a minimal set of heads that are necessary and sufficient for toxicity. We utilize these components via two complementary strategies: (1) Local Inference-Time Intervention, which constructs dynamic, input-specific steering vectors for context-aware detoxification, and (2) PNS-Guided Fine-Tuning, which permanently unlearns toxic representations. We also introduce PARATOX, a novel benchmark of aligned toxic/non-toxic sentence pairs enabling controlled counterfactual evaluation. Experiments on ToxiGen, ImplicitHate, and ParaDetox show that CAUSALDETOX achieves up to 5.34% greater toxicity reduction compared to baselines while preserving linguistic fluency, and offers a 7x speedup in head selection.

76. Mechanistic Decoding of Cognitive Constructs in LLMs

Authors: Yitong Shou , Manhao Guan
URL: https://arxiv.org/abs/2604.14593
Abstract:

While Large Language Models (LLMs) demonstrate increasingly sophisticated affective capabilities, the internal mechanisms by which they process complex emotions remain unclear. Existing interpretability approaches often treat models as black boxes or focus on coarse-grained basic emotions, leaving the cognitive structure of more complex affective states underexplored. To bridge this gap, we propose a Cognitive Reverse-Engineering framework based on Representation Engineering (RepE) to analyze social-comparison jealousy. By combining appraisal theory with subspace orthogonalization, regression-based weighting, and bidirectional causal steering, we isolate and quantify two psychological antecedents of jealousy, Superiority of Comparison Person and Domain Self-Definitional Relevance, and examine their causal effects on model judgments. Experiments on eight LLMs from the Llama, Qwen, and Gemma families suggest that models natively encode jealousy as a structured linear combination of these constituent factors. Their internal representations are broadly consistent with the human psychological construct, treating Superiority as the foundational trigger and Relevance as the ultimate intensity multiplier. Our framework also demonstrates that toxic emotional states can be mechanically detected and surgically suppressed, suggesting a possible route toward representational monitoring and intervention for AI safety in multi-agent environments.

77. AgileLog: A Forkable Shared Log for Agents on Data Streams

Authors: Shreesha G. Bhat , Tony Hong , Michael Noguera , Ramnatthan Alagappan , Aishwarya Ganesan
URL: https://arxiv.org/abs/2604.14590
Abstract:

In modern data-streaming systems, alongside traditional programs, a new type of entity has emerged that can interact with streaming data: AI agents. Unlike traditional programs, AI agents use LLM reasoning to accomplish high-level tasks specified in natural language over streaming data. Unfortunately, current streaming systems cannot fully support agents: they lack the fundamental mechanisms to avoid the performance interference caused by agentic tasks and to safely handle agentic writes. We argue that the shared log, the core abstraction underlying streaming data, must support creating forks of itself, and that such a forkable shared log serves as a great substrate for agents acting on streaming data. We propose AgileLog, a new shared log abstraction that provides novel forking primitives for agentic use cases. We design Bolt, an implementation of the AgileLog abstraction, that uses novel techniques to make forks cheap, and provide logical and performance isolation.

78. CPGRec+: A Balance-oriented Framework for Personalized Video Game Recommendations

Authors: Xiping Li , Aier Yang , Jianghong Ma , Kangzhe Liu , Shanshan Feng , Haijun Zhang , Yi Zhao
URL: https://arxiv.org/abs/2604.14586
Abstract:

The rapid expansion of gaming industry requires advanced recommender systems tailored to its dynamic landscape. Existing Graph Neural Network (GNN)-based methods primarily prioritize accuracy over diversity, overlooking their inherent trade-off. To address this, we previously proposed CPGRec, a balance-oriented gaming recommender system. However, CPGRec fails to account for critical disparities in player-game interactions, which carry varying significance in reflecting players’ personal preferences and may exacerbate over-smoothness issues inherent in GNN-based models. Moreover, existing approaches underutilize the reasoning capabilities and extensive knowledge of large language models (LLMs) in addressing these limitations. To bridge this gap, we propose two new modules. First, Preference-informed Edge Reweighting (PER) module assigns signed edge weights to qualitatively distinguish significant player interests and disinterests while then quantitatively measuring preference strength to mitigate over-smoothing in graph convolutions. Second, Preference-informed Representation Generation (PRG) module leverages LLMs to generate contextualized descriptions of games and players by reasoning personal preferences from comparing global and personal interests, thereby refining representations of players and games. Experiments on \textcolor{black}{two Steam datasets} demonstrate CPGRec+’s superior accuracy and diversity over state-of-the-art models. The code is accessible at this https URL .

79. Generative Augmented Inference

Authors: Cheng Lu , Mengxin Wang , Dennis J. Zhang , Heng Zhang
URL: https://arxiv.org/abs/2604.14575
Abstract:

Data-driven operations management often relies on parameters estimated from costly human-generated labels. Recent advances in large language models (LLMs) and other AI systems offer inexpensive auxiliary data, but introduce a new challenge: AI outputs are not direct observations of the target outcomes, but could involve high-dimensional representations with complex and unknown relationships to human labels. Conventional methods leverage AI predictions as direct proxies for true labels, which can be inefficient or unreliable when this relationship is weak or misspecified. We propose Generative Augmented Inference (GAI), a general framework that incorporates AI-generated outputs as informative features for estimating models of human-labeled outcomes. GAI uses an orthogonal moment construction that enables consistent estimation and valid inference with flexible, nonparametric relationship between LLM-generated outputs and human labels. We establish asymptotic normality and show a “safe default” property: relative to human-data-only estimators, GAI weakly improves estimation efficiency under arbitrary auxiliary signals and yields strict gains whenever the auxiliary information is predictive. Empirically, GAI outperforms benchmarks across diverse settings. In conjoint analysis with weak auxiliary signals, GAI reduces estimation error by about 50% and lowers human labeling requirements by over 75%. In retail pricing, where all methods access the same auxiliary inputs, GAI consistently outperforms alternative estimators, highlighting the value of its construction rather than differences in information. In health insurance choice, it cuts labeling requirements by over 90% while maintaining decision accuracy. Across applications, GAI improves confidence interval coverage without inflating width. Overall, GAI provides a principled and scalable approach to integrating AI-generated information.

80. Don’t Retrieve, Navigate: Distilling Enterprise Knowledge into Navigable Agent Skills for QA and RAG

Authors: Yiqun Sun , Pengfei Wei , Lawrence B. Hsieh
URL: https://arxiv.org/abs/2604.14572
Abstract:

Retrieval-Augmented Generation (RAG) grounds LLM responses in external evidence but treats the model as a passive consumer of search results: it never sees how the corpus is organized or what it has not yet retrieved, limiting its ability to backtrack or combine scattered evidence. We present Corpus2Skill, which distills a document corpus into a hierarchical skill directory offline and lets an LLM agent navigate it at serve time. The compilation pipeline iteratively clusters documents, generates LLM-written summaries at each level, and materializes the result as a tree of navigable skill files. At serve time, the agent receives a bird’s-eye view of the corpus, drills into topic branches via progressively finer summaries, and retrieves full documents by ID. Because the hierarchy is explicitly visible, the agent can reason about where to look, backtrack from unproductive paths, and combine evidence across branches. On WixQA, an enterprise customer-support benchmark for RAG, Corpus2Skill outperforms dense retrieval, RAPTOR, and agentic RAG baselines across all quality metrics.

81. VeriGraphi: A Multi-Agent Framework of Hierarchical RTL Generation for Large Hardware Designs

Authors: Sazzadul Islam , Tasnim Tabassum , Hao Zheng
URL: https://arxiv.org/abs/2604.14550
Abstract:

Generating synthesizable Verilog for large, hierarchical hardware designs remains a significant challenge for large language models (LLMs), which struggle to replicate the structured reasoning that human experts employ when translating complex specifications into RTL. When tasked with producing hierarchical Verilog, LLMs frequently lose context across modules, hallucinate interfaces, fabricate inter-module wiring, and fail to maintain structural coherence - failures that intensify as design complexity grows and specifications involve informal prose, figures, and tables that resist direct operationalization. To address these challenges, we present VeriGraphi, a framework that introduces a spec-anchored Knowledge Graph as the architectural substrate driving the RTL generation pipeline. VeriGraphi constructs a HDA, a structured knowledge graph that explicitly encodes module hierarchy, port-level interfaces, wiring semantics, and inter-module dependencies as first-class graph entities and relations. Built through iterative multi-agent analysis of the specification, this Knowledge Graph provides a deterministic, machine-checkable structural scaffold before code generation. Guided by the KG, a progressive coding module incrementally generates pseudo-code and synthesizable RTL while enforcing interface consistency and dependency correctness at each submodule stage. We evaluate VeriGraphi on a benchmark of three representative specification documents from the National Institute of Standards and Technology and their corresponding implementations, and we present a RV32I processor as a detailed case study to illustrate the full pipeline. The results demonstrate that VeriGraphi enables reliable hierarchical RTL generation with minimal human intervention for RISC-V, marking a significant milestone for LLM-generated hardware design while maintaining strong functional correctness.

82. Hierarchical vs. Flat Iteration in Shared-Weight Transformers

Authors: Sang-Il Han
URL: https://arxiv.org/abs/2604.14442
Abstract:

We present an empirical study of whether hierarchically structured, shared-weight recurrence can match the representational quality of independent-layer stacking in a Transformer-based language model. HRM-LM replaces L independent Transformer layers with a two-speed recurrent pair: a Fast module operating at every step for local refinement, and a Slow module operating every T steps for global compression. This recurrent hierarchy is unrolled for M = N x T steps with shared parameters. The central and most robust finding, supported by a parameter-matched Universal Transformer ablation (UniTF, 1.2B) across five independent runs, is a sharp empirical gap between the two approaches.

83. LLMs taking shortcuts in test generation: A study with SAP HANA and LevelDB

Authors: Vekil Bekmyradov , Noah C. Pütz , Thomas Bartz-Beielstein
URL: https://arxiv.org/abs/2604.14437
Abstract:

Large Language Models (LLMs) have achieved impressive results on public benchmarks, often leading to claims of advanced reasoning and understanding. However, recent research in cognitive science reveals that these models sometimes rely on shallow heuristics and memorization, taking shortcuts rather than demonstrating genuine cognitive abilities. This paper investigates LLM behavior in automated test generation for software, contrasting performance on an open-source system (LevelDB) with SAP HANA, one of the most widely deployed commercial database systems worldwide, whose proprietary codebase is guaranteed to be absent from training data. We combine cognitive evaluation principles, drawing on Mitchell’s mechanism-focused assessment methodology, with empirical software testing, employing mutation score and iterative compiler-feedback repair loops to assess both accuracy and underlying reasoning strategies. Results show that LLMs excel on familiar, open-source benchmarks but struggle with unseen, complex domains, often prioritizing compilability over semantic effectiveness. These findings provide independent software engineering evidence for the broader claim that current LLMs lack robust reasoning, and highlight the need for evaluation frameworks that penalize trivial shortcuts and reward true generalization.

84. SpaceMind: A Modular and Self-Evolving Embodied Vision-Language Agent Framework for Autonomous On-orbit Servicing

Authors: Aodi Wu , Haodong Han , Xubo Luo , Ruisuo Wang , Shan He , Xue Wan
URL: https://arxiv.org/abs/2604.14399
Abstract:

Autonomous on-orbit servicing demands embodied agents that perceive through visual sensors, reason about 3D spatial situations, and execute multi-phase tasks over extended horizons. We present SpaceMind, a modular and self-evolving vision-language model (VLM) agent framework that decomposes knowledge, tools, and reasoning into three independently extensible dimensions: skill modules with dynamic routing, Model Context Protocol (MCP) tools with configurable profiles, and injectable reasoning-mode skills. An MCP-Redis interface layer enables the same codebase to operate across simulation and physical hardware without modification, and a Skill Self-Evolution mechanism distills operational experience into persistent skill files without model fine-tuning. We validate SpaceMind through 192 closed-loop runs across five satellites, three task types, and two environments, a UE5 simulation and a physical laboratory, deliberately including degraded conditions to stress-test robustness. Under nominal conditions all modes achieve 90–100% navigation success; under degradation, the Prospective mode uniquely succeeds in search-and-approach tasks where other modes fail. A self-evolution study shows that the agent recovers from failure in four of six groups from a single failed episode, including complete failure to 100% success and inspection scores improving from 12 to 59 out of 100. Real-world validation confirms zero-code-modification transfer to a physical robot with 100% rendezvous success. Code: this https URL

85. Generating Concept Lexicalizations via Dictionary-Based Cross-Lingual Sense Projection

Authors: David Basil , Chirooth Girigowda , Bradley Hauer , Sahir Momin , Ning Shi , Grzegorz Kondrak
URL: https://arxiv.org/abs/2604.14397
Abstract:

We study the task of automatically expanding WordNet-style lexical resources to new languages through sense generation. We generate senses by associating target-language lemmas with existing lexical concepts via semantic projection. Given a sense-tagged English corpus and its translation, our method projects English synsets onto aligned target-language tokens and assigns the corresponding lemmas to those synsets. To generate these alignments and ensure their quality, we augment a pre-trained base aligner with a bilingual dictionary, which is also used to filter out incorrect sense projections. We evaluate the method on multiple languages, comparing it to prior methods, as well as dictionary-based and large language model baselines. Results show that the proposed project-and-filter strategy improves precision while remaining interpretable and requiring few external resources. We plan to make our code, documentation, and generated sense inventories accessible.

86. BiCon-Gate: Consistency-Gated De-colloquialisation for Dialogue Fact-Checking

Authors: Hyunkyung Park , Arkaitz Zubiaga
URL: https://arxiv.org/abs/2604.14389
Abstract:

Automated fact-checking in dialogue involves multi-turn conversations where colloquial language is frequent yet understudied. To address this gap, we propose a conservative rewrite candidate for each response claim via staged de-colloquialisation, combining lightweight surface normalisation with scoped in-claim coreference resolution. We then introduce BiCon-Gate, a semantics-aware consistency gate that selects the rewrite candidate only when it is semantically supported by the dialogue context, otherwise falling back to the original claim. This gated selection stabilises downstream fact-checking and yields gains in both evidence retrieval and fact verification. On the DialFact benchmark, our approach improves retrieval and verification, with particularly strong gains on SUPPORTS, and outperforms competitive baselines, including a decoder-based one-shot LLM rewrite that attempts to perform all de-colloquialisation steps in a single pass.

87. Coalition Formation in LLM Agent Networks: Stability Analysis and Convergence Guarantees

Authors: Dongxin Guo , Jikun Wu , Siu-Ming Yiu
URL: https://arxiv.org/abs/2604.14386
Abstract:

Large Language Model (LLM) agents are increasingly deployed in multi-agent systems requiring strategic coordination. While recent work has analyzed LLM behavior in two-player games, coalition formation, where $n$ agents dynamically form cooperative groups, remains theoretically uncharacterized. We present the first framework grounding coalition formation in LLM agent networks in hedonic game theory with formal stability guarantees. We introduce the LLM Coalition Formation Game (LCFG), establish sufficient conditions for Nash-stable partitions, and prove complexity results. Our analysis reveals that LLM agents exhibit bounded rationality characterized by $\epsilon$-rational preferences; we provide both deterministic existence guarantees and consistency-driven stability bounds whose predictions are consistent with empirical outcomes. Experiments with GPT-4, Claude-3, and Llama-3 across 2,400 episodes validate our framework: LLM coalitions achieve Nash stability in 73.2% of cases under our Coalition-of-Thought (CoalT) protocol, compared to 58.4% under chain-of-thought and 41.8% under standard prompting ($p < 0.001$). Our framework provides theoretical foundations for designing stable multi-agent LLM systems.

88. Modular Continual Learning via Zero-Leakage Reconstruction Routing and Autonomous Task Discovery

Authors: Noureddine Kermiche
URL: https://arxiv.org/abs/2604.14375
Abstract:

Catastrophic forgetting remains a primary hurdle in sequential task learning for artificial neural networks. We propose a silicon-native modular architecture that achieves structural parameter isolation using Task-Specific Experts and a distributed, outlier-based Gatekeeper. Moving beyond traditional sequential consolidation, our framework utilizes a Simultaneous Pipeline where Teacher learning, Student distillation, and Router manifold acquisition occur in parallel while raw data is present in a localized training session. This approach ensures computational efficiency and complies with privacy mandates like GDPR by deleting raw data as soon as a task is learned. We demonstrate that a Tight-Bottleneck Autoencoder (TB-AE) can effectively distinguish semantically crowded manifolds in high-dimensional latent spaces, overcoming the posterior collapse inherent to standard variational methods. By establishing strict topological boundaries, our TB-AE resolves latent space crowding in 4096-D LLM embeddings to provide a robust, unsupervised novelty signal. Furthermore, we validate an Autonomous Retrieval mechanism that confidently identifies returning manifolds, enabling stable lifelong learning without redundant module instantiation. Empirical results demonstrate that our ``Live Distillation’’ approach acts as a natural regularizer, achieving strong retention across computer vision and natural language processing domains without suffering a student fidelity gap.

89. SatBLIP: Context Understanding and Feature Identification from Satellite Imagery with Vision-Language Learning

Authors: Xue Wu , Shengting Cao , Jiaqi Gong
URL: https://arxiv.org/abs/2604.14373
Abstract:

Rural environmental risks are shaped by place-based conditions (e.g., housing quality, road access, land-surface patterns), yet standard vulnerability indices are coarse and provide limited insight into risk contexts. We propose SatBLIP, a satellite-specific vision-language framework for rural context understanding and feature identification that predicts county-level Social Vulnerability Index (SVI). SatBLIP addresses limitations of prior remote sensing pipelines-handcrafted features, manual virtual audits, and natural-image-trained VLMs-by coupling contrastive image-text alignment with bootstrapped captioning tailored to satellite semantics. We use GPT-4o to generate structured descriptions of satellite tiles (roof type/condition, house size, yard attributes, greenery, and road context), then fine-tune a satellite-adapted BLIP model to generate captions for unseen images. Captions are encoded with CLIP and fused with LLM-derived embeddings via attention for SVI estimation under spatial aggregation. Using SHAP, we identify salient attributes (e.g., roof form/condition, street width, vegetation, cars/open space) that consistently drive robust predictions, enabling interpretable mapping of rural risk environments.

Authors: Akshay Paruchuri , Ishan Chatterjee , Henry Fuchs , Ehsan Adeli , Piotr Didyk
URL: https://arxiv.org/abs/2604.14363
Abstract:

Multimodal language models systematically underperform on visual perception tasks, yet the structure underlying this failure remains poorly understood. We propose centroid replacement, collapsing each token to its nearest K-means centroid, as a controlled probe for modal dependence. Across seven models spanning three architecture families, erasing text centroid structure costs 4$\times$ more accuracy than erasing visual centroid structure, exposing a universal imbalance where language representations overshadow vision even on tasks that demand visual reasoning. We exploit this asymmetry through text centroid contrastive decoding, recovering up to +16.9% accuracy on individual tasks by contrastively decoding against a text-centroid-erased reference. This intervention varies meaningfully with training approaches: standard fine-tuned models show larger gains (+5.6% on average) than preference-optimized models (+1.5% on average). Our findings suggest that modal competition is structurally localized, correctable at inference time without retraining, and quantifiable as a diagnostic signal to guide future multimodal training.

91. APEX-MEM: Agentic Semi-Structured Memory with Temporal Reasoning for Long-Term Conversational AI

Authors: Pratyay Banerjee , Masud Moshtaghi , Shivashankar Subramanian , Amita Misra , Ankit Chadha
URL: https://arxiv.org/abs/2604.14362
Abstract:

Large language models still struggle with reliable long-term conversational memory: simply enlarging context windows or applying naive retrieval often introduces noise and destabilizes responses. We present APEX-MEM, a conversational memory system that combines three key innovations: (1) a property graph which uses domain-agnostic ontology to structure conversations as temporally grounded events in an entity-centric framework, (2) append-only storage that preserves the full temporal evolution of information, and (3) a multi-tool retrieval agent that understands and resolves conflicting or evolving information at query time, producing a compact and contextually relevant memory summary. This retrieval-time resolution preserves the full interaction history while suppressing irrelevant details. APEX-MEM achieves 88.88% accuracy on LOCOMO’s Question Answering task and 86.2% on LongMemEval, outperforming state-of-the-art session-aware approaches and demonstrating that structured property graphs enable more temporally coherent long-term conversational reasoning.

92. When PCOS Meets Eating Disorders: An Explainable AI Approach to Detecting the Hidden Triple Burden

Authors: Apoorv Prasad , Susan McRoy
URL: https://arxiv.org/abs/2604.14356
Abstract:

Women with polycystic ovary syndrome (PCOS) face substantially elevated risks of body image distress, disordered eating, and metabolic challenges, yet existing natural language processing approaches for detecting these conditions lack transparency and cannot identify co-occurring presentations. We developed small, open-source language models to automatically detect this triple burden in social media posts with grounded explainability. We collected 1,000 PCOS-related posts from six subreddits, with two trained annotators labeling posts using guidelines operationalizing Lee et al. (2017) clinical framework. Three models (Gemma-2-2B, Qwen3-1.7B, DeepSeek-R1-Distill-Qwen-1.5B) were fine-tuned using Low-Rank Adaptation to generate structured explanations with textual evidence. The best model achieved 75.3 percent exact match accuracy on 150 held-out posts, with robust comorbidity detection and strong explainability. Performance declined with diagnostic complexity, indicating their best use is for screening rather than autonomous diagnosis.

Authors: Pushpa Kumar Balan , Aijing Feng
URL: https://arxiv.org/abs/2604.14334
Abstract:

Gradient saliency from deep sequence models surfaces candidate biomarkers efficiently, but the resulting gene lists are contaminated by tissue-composition confounders that degrade downstream classifiers. We study whether LLM chain-of-thought (CoT) reasoning can faithfully filter these confounders, and whether reasoning quality drives downstream performance. We train a Mamba SSM on TCGA-BRCA RNA-seq and extract the top-50 genes by gradient saliency; DeepSeek-R1 evaluates every candidate with structured CoT to produce a final 17-gene set. The raw 50-gene saliency set (no LLM) performs worse than a 5,000-gene variance baseline (AUC 0.832 vs. 0.903), while the LLM-filtered set surpasses it (AUC 0.927), using 294x fewer features. A faithfulness audit (COSMIC CGC, OncoKB, PAM50) reveals only 6 of 17 selected genes (35.3%) are validated BRCA biomarkers, yet 10 of 16 known BRCA genes in the input were missed - including FOXA1. This gap between downstream performance and reasoning faithfulness suggests selective faithfulness: targeted confounder removal is sufficient for performance gains even without comprehensive recall.

94. Faithfulness Serum: Mitigating the Faithfulness Gap in Textual Explanations of LLM Decisions via Attribution Guidance

Authors: Bar Alon , Itamar Zimerman , Lior Wolf
URL: https://arxiv.org/abs/2604.14325
Abstract:

Large language models (LLMs) achieve strong performance and have revolutionized NLP, but their lack of explainability keeps them treated as black boxes, limiting their use in domains that demand transparency and trust. A promising direction to address this issue is post-hoc text-based explanations, which aim to explain model decisions in natural language. Prior work has focused on generating convincing rationales that appear to be subjectively faithful, but it remains unclear whether these explanations are epistemically faithful, whether they reflect the internal evidence the model actually relied on for its decision. In this paper, we first assess the epistemic faithfulness of LLM-generated explanations via counterfactuals and show that they are often unfaithful. We then introduce a training-free method that enhances faithfulness by guiding explanation generation through attention-level interventions, informed by token-level heatmaps extracted via a faithful attribution method. This method significantly improves epistemic faithfulness across multiple models, benchmarks, and prompts.

95. Challenges and Future Directions in Agentic Reverse Engineering Systems

Authors: Salem Radey , Jack West , Kassem Fawaz
URL: https://arxiv.org/abs/2604.14317
Abstract:

Agentic systems built on large language models (LLMs) are increasingly being used for complex security tasks, including binary reverse engineering (RE). Despite recent growth in popularity and capability, these systems continue to face limitations in realistic settings. Cutting-edge systems still fail in complex RE scenarios that involve obfuscation, timing, and unique architecture. In this work, we examine how agentic systems perform reverse engineering tasks with static, dynamic, and hybrid agents. Through an analysis of existing agentic tool usage, we identify several limitations, including token constraints, struggles with obfuscation, and a lack of program guardrails. From these findings, we outline current challenges and position future directions for system designers to overcome from a security perspective.

96. DharmaOCR: Specialized Small Language Models for Structured OCR that outperform Open-Source and Commercial Baselines

Authors: Gabriel Pimenta de Freitas Cardoso , Caio Lucas da Silva Chacon , Jonas Felipe da Fonseca Oliveira , Paulo Henrique de Medeiros Araujo
URL: https://arxiv.org/abs/2604.14314
Abstract:

This manuscript introduces DharmaOCR Full and Lite, a pair of specialized small language models (SSLMs) for structured OCR that jointly optimize transcription quality, generation stability, and inference cost. It also presents DharmaOCR-Benchmark, a benchmark that covers printed, handwritten, and legal/administrative documents, and proposes a unified evaluation protocol that measures fidelity and structure while explicitly tracking text degeneration as a first-class benchmark metric (alongside unit cost). Beyond reporting degeneration rates, the manuscript empirically shows degeneration is not merely a quality failure, since it materially worsens production performance by increasing response time, reducing throughput, and inflating computational cost due to abnormally long generations. To the best of the author’s knowledge, as a methodological contribution, this is the first application of Direct Preference Optimization (DPO) for OCR, explicitly using degenerate generations as rejected examples to penalize looping behavior. Combined with Supervised Fine-Tuning (SFT) for enforcing a strict JSON schema (header, margin, footer, and text), DPO consistently reduces degeneration rate across model families (up to 87.6% relative) while preserving or improving extraction quality. The resulting models, namely, DharmaOCR Full (7B) and DharmaOCR Lite (3B), set a new state-of-the-art on DharmaOCR-Benchmark, outperforming each open-source and commercial baseline model evaluated regarding extraction quality, reaching 0.925 and 0.911 scores with 0.40% and 0.20% degeneration rates. AWQ quantization reduced up to 22% per-page cost with negligible quality loss, enabling a strong quality-cost trade-off in comparison to proprietary OCR APIs and open-source alternatives.

97. EuropeMedQA Study Protocol: A Multilingual, Multimodal Medical Examination Dataset for Language Model Evaluation

Authors: Francesco Andrea Causio , Vittorio De Vita , Olivia Riccomi , Michele Ferramola , Federico Felizzi , Antonio Cristiano , Lorenzo De Mori , Chiara Battipaglia , Melissa Sawaya , Luigi De Angelis , Marcello Di Pumpo , Alessandra Piscitelli , Pietro Eric Risuleo , Alessia Longo , Giulia Vojvodic , Mariapia Vassalli , Bianca Destro Castaniti , Nicolò Scarsi , Manuel Del Medico
URL: https://arxiv.org/abs/2604.14306
Abstract:

While Large Language Models (LLMs) have demonstrated high proficiency on English-centric medical examinations, their performance often declines when faced with non-English languages and multimodal diagnostic tasks. This study protocol describes the development of EuropeMedQA, the first comprehensive, multilingual, and multimodal medical examination dataset sourced from official regulatory exams in Italy, France, Spain, and Portugal. Following FAIR data principles and SPIRIT-AI guidelines, we describe a rigorous curation process and an automated translation pipeline for comparative analysis. We evaluate contemporary multimodal LLMs using a zero-shot, strictly constrained prompting strategy to assess cross-lingual transfer and visual reasoning. EuropeMedQA aims to provide a contamination-resistant benchmark that reflects the complexity of European clinical practices and fosters the development of more generalizable medical AI.

98. Enhancing LLM-based Search Agents via Contribution Weighted Group Relative Policy Optimization

Authors: Junzhe Wang , Zhiheng Xi , yajie yang , Hao Luo , Shihan Dou , Tao Gui , Qi Zhang
URL: https://arxiv.org/abs/2604.14267
Abstract:

Search agents extend Large Language Models (LLMs) beyond static parametric knowledge by enabling access to up-to-date and long-tail information unavailable during pretraining. While reinforcement learning has been widely adopted for training such agents, existing approaches face key limitations: process supervision often suffers from unstable value estimation, whereas outcome supervision struggles with credit assignment due to sparse, trajectory-level rewards. To bridge this gap, we propose Contribution-Weighted GRPO (CW-GRPO), a framework that integrates process supervision into group relative policy optimization. Instead of directly optimizing process rewards, CW-GRPO employs an LLM judge to assess the retrieval utility and reasoning correctness at each search round, producing per-round contribution scores. These scores are used to rescale outcome-based advantages along the trajectory, enabling fine-grained credit assignment without sacrificing optimization stability. Experiments on multiple knowledge-intensive benchmarks show that CW-GRPO outperforms standard GRPO by 5.0\% on Qwen3-8B and 6.3\% on Qwen3-1.7B, leading to more effective search behaviors. Additional analysis reveals that successful trajectories exhibit concentrated contributions across rounds, providing empirical insight into search agent tasks.

99. Reinforcement Learning via Value Gradient Flow

Authors: Haoran Xu , Kaiwen Hu , Somayeh Sojoudi , Amy Zhang
URL: https://arxiv.org/abs/2604.14265
Abstract:

We study behavior-regularized reinforcement learning (RL), where regularization toward a reference distribution (the dataset in offline RL or the base model in LLM RL finetuning) is essential to prevent value over-optimization caused by erroneous out-of-distribution extrapolation. Existing methods either rely on reparameterized policy gradient, which are difficult to scale to large generative models, or on reject sampling, which can be overly conservative when attempting to move beyond the behavior support. In this paper, we propose Value Gradient Flow (VGF), a scalable new paradigm for behavior-regularized RL. VGF casts behavior-regularized RL as an optimal transport problem that maps the reference distribution to the value-induced optimal policy distribution. We solve this transport problem via discrete gradient flow, where value gradients guide particles initialized from the reference distribution. Our analysis shows that VGF imposes regularization implicitly by controlling the transport budget. VGF eliminates explicit policy parameterization while remaining expressive and flexible, this enables adaptive test-time scaling by adjusting the transport budget. Extensive experiments demonstrate that VGF significantly outperforms prior methods, achieving state-of-the-art results on offline RL benchmarks (D4RL, OGBench) and LLM RL tasks. Code and runs can be found at this https URL .

100. ReviewGrounder: Improving Review Substantiveness with Rubric-Guided, Tool-Integrated Agents

Authors: Zhuofeng Li , Yi Lu , Dongfu Jiang , Haoxiang Zhang , Yuyang Bai , Chuan Li , Yu Wang , Shuiwang Ji , Jianwen Xie , Yu Zhang
URL: https://arxiv.org/abs/2604.14261
Abstract:

The rapid rise in AI conference submissions has driven increasing exploration of large language models (LLMs) for peer review support. However, LLM-based reviewers often generate superficial, formulaic comments lacking substantive, evidence-grounded feedback. We attribute this to the underutilization of two key components of human reviewing: explicit rubrics and contextual grounding in existing work. To address this, we introduce REVIEWBENCH, a benchmark evaluating review text according to paper-specific rubrics derived from official guidelines, the paper’s content, and human-written reviews. We further propose REVIEWGROUNDER, a rubric-guided, tool-integrated multi-agent framework that decomposes reviewing into drafting and grounding stages, enriching shallow drafts via targeted evidence consolidation. Experiments on REVIEWBENCH show that REVIEWGROUNDER, using a Phi-4-14B-based drafter and a GPT-OSS-120B-based grounding stage, consistently outperforms baselines with substantially stronger/larger backbones (e.g., GPT-4.1 and DeepSeek-R1-670B) in both alignment with human judgments and rubric-based review quality across 8 dimensions. The code is available \href{ this https URL }{here}.

101. Evaluation of Agents under Simulated AI Marketplace Dynamics

Authors: To Eun Kim , Alireza Salemi , Hamed Zamani , Fernando Diaz
URL: https://arxiv.org/abs/2604.14256
Abstract:

Modern information access ecosystems consist of mixtures of systems, such as retrieval systems and large language models, and increasingly rely on marketplaces to mediate access to models, tools, and data, making competition between systems inherent to deployment. In such settings, outcomes are shaped not only by benchmark quality but also by competitive pressure, including user switching, routing decisions, and operational constraints. Yet evaluation is still largely conducted on static benchmarks with accuracy-focused measures that assume systems operate in isolation. This mismatch makes it difficult to predict post-deployment success and obscures competitive effects such as early-adoption advantages and market dominance. We introduce Marketplace Evaluation, a simulation-based paradigm that evaluates information access systems as participants in a competitive marketplace. By simulating repeated interactions and evolving user and agent preferences, the framework enables longitudinal evaluation and marketplace-level metrics, such as retention and market share, that complement and can extend beyond traditional accuracy-based metrics. We formalize the framework and outline a research agenda, motivated by business and economics, around marketplace simulation, metrics, optimization, and adoption in evaluation campaigns like TREC.

102. FRESCO: Benchmarking and Optimizing Re-rankers for Evolving Semantic Conflict in Retrieval-Augmented Generation

Authors: Sohyun An (1 and 2), Hayeon Lee (1), Shuibenyang Yuan (1), Chun-cheng Jason Chen (1), Cho-Jui Hsieh (2), Vijai Mohan (1), Alexander Min (1) ((1) Meta Superintelligence Labs, (2) UCLA)
URL: https://arxiv.org/abs/2604.14227
Abstract:

Retrieval-Augmented Generation (RAG) is a key approach to mitigating the temporal staleness of large language models (LLMs) by grounding responses in up-to-date evidence. Within the RAG pipeline, re-rankers play a pivotal role in selecting the most useful documents from retrieved candidates. However, existing benchmarks predominantly evaluate re-rankers in static settings and do not adequately assess performance under evolving information – a critical gap, as real-world systems often must choose among temporally different pieces of evidence. To address this limitation, we introduce FRESCO (Factual Recency and Evolving Semantic COnflict), a benchmark for evaluating re-rankers in temporally dynamic contexts. By pairing recency-seeking queries with historical Wikipedia revisions, FRESCO tests whether re-rankers can prioritize factually recent evidence while maintaining semantic relevance. Our evaluation reveals a consistent failure mode across existing re-rankers: a strong bias toward older, semantically rich documents, even when they are factually obsolete. We further investigate an instruction optimization framework to mitigate this issue. By identifying Pareto-optimal instructions that balance Evolving and Non-Evolving Knowledge tasks, we obtain gains of up to 27% on Evolving Knowledge tasks while maintaining competitive performance on Non-Evolving Knowledge tasks.

103. TRACE: A Conversational Framework for Sustainable Tourism Recommendation with Agentic Counterfactual Explanations

Authors: Ashmi Banerjee , Adithi Satish , Wolfgang Wörndl , Yashar Deldjoo
URL: https://arxiv.org/abs/2604.14223
Abstract:

Traditional conversational travel recommender systems primarily optimize for user relevance and convenience, often reinforcing popular, overcrowded destinations and carbon-intensive travel choices. To address this, we present TRACE (Tourism Recommendation with Agentic Counterfactual Explanations), a multi-agent, LLM-based framework that promotes sustainable tourism through interactive nudging. TRACE uses a modular orchestrator-worker architecture where specialized agents elicit latent sustainability preferences, construct structured user personas, and generate recommendations that balance relevance with environmental impact. A key innovation lies in its use of agentic counterfactual explanations and LLM-driven clarifying questions, which together surface greener alternatives and refine understanding of intent, fostering user reflection without coercion. User studies and semantic alignment analyses demonstrate that TRACE effectively supports sustainable decision-making while preserving recommendation quality and interactive responsiveness. TRACE is implemented on Google’s Agent Development Kit, with full code, Docker setup, prompts, and a publicly available demo video to ensure reproducibility. A project summary, including all resources, prompts, and demo access, is available at this https URL .

104. Adaptive Query Routing: A Tier-Based Framework for Hybrid Retrieval Across Financial, Legal, and Medical Documents

Authors: Afshan Hashmi
URL: https://arxiv.org/abs/2604.14222
Abstract:

Retrieval-Augmented Generation (RAG) has become the standard paradigm for grounding Large Language Model outputs in external knowledge. Lumer et al. [1] presented the first systematic evaluation comparing vector-based agentic RAG against hierarchical node-based reasoning systems for financial document QA across 1,200 SEC filings, finding vector-based systems achieved a 68% win rate. Concurrently, the PageIndex framework [2] demonstrated 98.7% accuracy on FinanceBench through purely reasoning-based retrieval. This paper extends their work by: (i) implementing and evaluating three retrieval architectures: Vector RAG, Tree Reasoning, and the proposed Adaptive Hybrid Retrieval (AHR) across financial, legal, and medical domains; (ii) introducing a four-tier query complexity benchmark; and (iii) employing GPT-4-powered LLM-as-judge evaluation. Experiments reveal that Tree Reasoning achieves the highest overall score (0.900), but no single paradigm dominates across all tiers: Vector RAG wins on multi-document synthesis (Tier 4, score 0.900), while the Hybrid AHR achieves the best performance on cross-reference (0.850) and multi-section queries (0.929). Cross-reference recall reaches 100% for tree-based and hybrid approaches versus 91.7% for vector search, quantifying a critical capability gap. Validation on FinanceBench (150 expert-annotated questions on real SEC 10-K and 10-Q filings) confirms and strengthens these findings: Tree Reasoning scores 0.938, Hybrid AHR 0.901, and Vector RAG 0.821, with the Tree–Vector quality gap widening to 11.7 percentage points on real-world documents. These findings support the development of adaptive retrieval systems that dynamically select strategies based on query complexity and document structure. All code and data are publicly available.

105. PriHA: A RAG-Enhanced LLM Framework for Primary Healthcare Assistant in Hong Kong

Authors: Richard Wai Cheung Chan , Shanru Lin , Ya-nan Ma , Hao Chen , Liangjun Jiang , Wenqi Fan
URL: https://arxiv.org/abs/2604.14215
Abstract:

To address the unsustainable rise in public health expenditures, the Hong Kong SAR Government is shifting its strategic focus to primary healthcare and encouraging citizens to use community resources to self-manage their health. However, official clinical guidelines are fragmented across disparate departments and formats, creating significant access barriers. While general-purpose Large Language Models (LLMs) such as ChatGPT and DeepSeek offer potential solutions for information accessibility, they are prone to generating factually inaccurate content due to a lack of localized and domain-specific knowledge. To this end, we propose a Retrieval-Augmented Generation-Enhanced LLM system as Primary Healthcare Assistant (PriHA) in Hong Kong. Specifically, a tri-stage pipeline is proposed that leverages a query optimizer to generalize user intent-oriented sub-queries, followed by a novel Dual Retrieval Augmented Generation (DRAG) architecture for mixed-source retrieval and context-reorganized generation. Comprehensive experiments and a detailed case study demonstrate that our proposed method can outperform both ablations and baseline in terms of accuracy and clarity. Our research provides a reliable and traceable dialogue retrieval framework for exploring other high-risk, localized application scenarios.

106. CROP: Token-Efficient Reasoning in Large Language Models via Regularized Prompt Optimization

Authors: Deep Shah , Sanket Badhe , Nehal Kathrotia , Priyanka Tiwari
URL: https://arxiv.org/abs/2604.14214
Abstract:

Large Language Models utilizing reasoning techniques improve task performance but incur significant latency and token costs due to verbose generation. Existing automatic prompt optimization(APO) frameworks target task accuracy exclusively at the expense of generating long reasoning traces. We propose Cost-Regularized Optimization of Prompts (CROP), an APO method that introduces regularization on response length by generating textual feedback in addition to standard accuracy feedback. This forces the optimization process to produce prompts that elicit concise responses containing only critical information and reasoning. We evaluate our approach on complex reasoning datasets, specifically GSM8K, LogiQA and BIG-Bench Hard. We achieved an 80.6\% reduction in token consumption while maintaining competitive accuracy, seeing only a nominal decline in performance. This presents a pragmatic solution for deploying token-efficient and cost-effective agentic AI systems in production pipelines.

107. PolyBench: Benchmarking LLM Forecasting and Trading Capabilities on Live Prediction Market Data

Authors: Pu Cheng , Juncheng Liu , Yunshen Long
URL: https://arxiv.org/abs/2604.14199
Abstract:

Predicting real-world events from live market signals demands systems that fuse qualitative news with quantitative order-book dynamics under strict temporal discipline – a challenge existing benchmarks fail to capture. We present \textbf{PolyBench}, a multimodal benchmark derived from Polymarket that records point-in-time cross-sections of 38,666 binary prediction markets spanning 4,997 events, synchronously coupling each snapshot with a Central Limit Order Book (CLOB) state and a real-time news stream. Using PolyBench, we evaluate seven state-of-the-art Large Language Models – spanning open- and closed-source families – generating 36,165 predictions under identical, timestamp-locked market states collected between February 6 and 12, 2026. Our multidimensional framework assesses directional accuracy, our proposed Confidence-Weighted Return (CWR), Annualized Percentage Yield (APY), and Sharpe ratio via realistic order-book execution simulation. The results reveal a pronounced performance divergence: only two of seven models achieve positive financial returns – MiMo-V2-Flash at \textbf{17.6% } CWR and Gemini-3-Flash at 6.2% CWR – while the remaining five incur losses despite uniformly high stated confidence. These findings highlight the gap between surface-level language fluency and genuine probabilistic reasoning under live market uncertainty, and establish PolyBench as a contamination-proof, financially-grounded evaluation standard for future LLM research. Our dataset and code available at \underline{\href{ this https URL }{ this https URL } }.

108. MixAtlas: Uncertainty-aware Data Mixture Optimization for Multimodal LLM Midtraining

Authors: Bingbing Wen , Sirajul Salekin , Feiyang Kang , Bill Howe , Lucy Lu Wang , Javier Movellan , Manjot Bilkhu
URL: https://arxiv.org/abs/2604.14198
Abstract:

Domain reweighting can improve sample efficiency and downstream generalization, but data-mixture optimization for multimodal midtraining remains largely unexplored. Current multimodal training recipes tune mixtures along a single dimension, typically data format or task type. We introduce MixAtlas, a method that produces benchmark-targeted data recipes that can be inspected, adapted, and transferred to new corpora. MixAtlas decomposes the training corpus along two axes: image concepts (10 visual-domain clusters discovered via CLIP embeddings) and task supervision (5 objective types including captioning, OCR, grounding, detection, and VQA). Using small proxy models (Qwen2-0.5B) paired with a Gaussian-process surrogate and GP-UCB acquisition, MixAtlas searches the resulting mixture space with the same proxy budget as regression-based baselines but finds better-performing mixtures. We evaluate on 10 benchmarks spanning visual understanding, document reasoning, and multimodal reasoning. On Qwen2-7B, optimized mixtures improve average performance by 8.5%-17.6% over the strongest baseline; on Qwen2.5-7B, gains are 1.0%-3.3%. Both settings reach baseline-equivalent training loss in up to 2 times fewer steps. Recipes discovered on 0.5B proxies transfer to 7B-scale training across Qwen model families.

109. The PICCO Framework for Large Language Model Prompting: A Taxonomy and Reference Architecture for Prompt Structure

Authors: David A. Cook
URL: https://arxiv.org/abs/2604.14197
Abstract:

Large language model (LLM) performance depends heavily on prompt design, yet prompt construction is often described and applied inconsistently. Our purpose was to derive a reference framework for structuring LLM prompts. This paper presents PICCO, a framework derived through a rigorous synthesis of 11 previously published prompting frameworks identified through a multi-database search. The analysis yields two main contributions. First, it proposes a taxonomy that distinguishes prompt frameworks, prompt elements, prompt generation, prompting techniques, and prompt engineering as related but non-equivalent concepts. Second, it derives a five-element reference architecture for prompt generation: Persona, Instructions, Context, Constraints, and Output (PICCO). For each element, we define its function, scope, and relationship to other elements, with the goal of improving conceptual clarity and supporting more systematic prompt design. Finally, to support application of the framework, we outline key concepts relevant to implementation, including prompting techniques (e.g., zero-shot, few-shot, chain-of-thought, ensembling, decomposition, and self-critique, with selected variants), human and automated approaches to iterative prompt engineering, responsible prompting considerations such as security, privacy, bias, and trust, and priorities for future research. This work is a conceptual and methodological contribution: it formalizes a common structure for prompt specification and comparison, but does not claim empirical validation of PICCO as an optimization method.

110. Grading the Unspoken: Evaluating Tacit Reasoning in Quantum Field Theory and String Theory with LLMs

Authors: Xingyang Yu , Yinghuan Zhang , Yufei Zhang , Zijun Cui
URL: https://arxiv.org/abs/2604.14188
Abstract:

Large language models have demonstrated impressive performance across many domains of mathematics and physics. One natural question is whether such models can support research in highly abstract theoretical fields such as quantum field theory and string theory. Evaluating this possibility faces an immediate challenge: correctness in these domains is layered, tacit, and fundamentally non-binary. Standard answer-matching metrics fail to capture whether intermediate conceptual steps are properly reconstructed or whether implicit structural constraints are respected. We construct a compact expert-curated dataset of twelve questions spanning core areas of quantum field theory and string theory, and introduce a five-level grading rubric separating statement correctness, key concept awareness, reasoning chain presence, tacit step reconstruction, and enrichment. Evaluating multiple contemporary LLMs, we observe near-ceiling performance on explicit derivations within stable conceptual frames, but systematic degradation when tasks require reconstruction of omitted reasoning steps or reorganization of representations under global consistency constraints. These failures are driven not only by missing intermediate steps, but by an instability in representation selection: models often fail to identify the correct conceptual framing required to resolve implicit tensions. We argue that highly abstract theoretical physics provides a uniquely sensitive lens on the epistemic limits of current evaluation paradigms.

111. Internal Knowledge Without External Expression: Probing the Generalization Boundary of a Classical Chinese Language Model

Authors: Jiuting Chen , Yuan Lian , Hao Wu , Tianqi Huang , Hiroshi Sasaki , Makoto Kouno , Jongil Choi
URL: https://arxiv.org/abs/2604.14180
Abstract:

We train a 318M-parameter Transformer language model from scratch on a curated corpus of 1.56 billion tokens of pure Classical Chinese, with zero English characters or Arabic numerals. Through systematic out-of-distribution (OOD) testing, we investigate whether the model can distinguish known from unknown inputs, and crucially, whether it can express this distinction in its generated text. We find a clear dissociation between internal and external uncertainty. Internally, the model exhibits a perplexity jump ratio of 2.39x between real and fabricated historical events (p = 8.9e-11, n = 92 per group), with semi-fabricated events (real figures + fictional events) showing the highest perplexity (4.24x, p = 1.1e-16), demonstrating genuine factual encoding beyond syntactic pattern matching. Externally, however, the model never learns to express uncertainty: classical Chinese epistemic markers appear at lower rates for OOD questions (3.5%) than for in-distribution questions (8.3%, p = 0.023), reflecting rhetorical conventions rather than genuine metacognition. We replicate both findings across three languages (Classical Chinese, English, Japanese), three writing systems, and eight models from 110M to 1.56B parameters. We further show that uncertainty expression frequency is determined entirely by training data conventions, with Classical Chinese models showing a “humility paradox” (more hedging for known topics), while Japanese models almost never hedge. We argue that metacognitive expression – the ability to say “I don’t know” – does not emerge from language modeling alone and requires explicit training signals such as RLHF.

112. An Underexplored Frontier: Large Language Models for Rare Disease Patient Education and Communication – A scoping review

Authors: Zaifu Zhan , Yu Hou , Kai Yu , Min Zeng , Anita Burgun , Xiaoyi Chen , Rui Zhang
URL: https://arxiv.org/abs/2604.14179
Abstract:

Rare diseases affect over 300 million people worldwide and are characterized by complex care pathways, limited clinical expertise, and substantial unmet communication needs throughout the long patient journey. Recent advances in large language models (LLMs) offer new opportunities to support patient education and communication, yet their application in rare diseases remains unclear. We conducted a scoping review of studies published between January 2022 and March 2026 across major databases, identifying 12 studies on LLM-based rare disease patient education and communication. Data were extracted on study characteristics, application scenarios, model usage, and evaluation methods, and synthesized using descriptive and qualitative analyses. The literature is highly recent and dominated by general-purpose models, particularly ChatGPT. Most studies focus on patient question answering using curated question sets, with limited use of real-world data or longitudinal communication scenarios. Evaluations are primarily centered on accuracy, with limited attention to patient-centered dimensions such as readability, empathy, and communication quality. Multilingual communication is rarely addressed. Overall, the field remains at an early stage. Future research should prioritize patient-centered design, domain-adapted methods, and real-world deployment to support safe, adaptive, and effective communication in rare diseases.

113. Tug-of-War within A Decade: Conflict Resolution in Vulnerability Analysis via Teacher-Guided Retrieval-Augmented Generations

Authors: Ziyin Zhou , Jianyi Zhang , Xu ji , Yilong Li , Jiameng Han , Zhangchi Zhao
URL: https://arxiv.org/abs/2604.14172
Abstract:

Large Language Models (LLMs) are essential for analyzing and addressing vulnerabilities in cybersecurity. However, among over 200,000 vulnerabilities were discovered in the past decade, more than 30,000 have been changed or updated. This necessitates frequent updates to the training datasets and internal knowledge bases of LLMs to maintain knowledge consistency. In this paper, we focus on the problem of knowledge discrepancy and conflict within CVE (Common Vulnerabilities and Exposures) detection and analysis. This problem hinders LLMs’ ability to retrieve the latest knowledge from original training datasets, leading to knowledge conflicts, fabrications of factually incorrect results, and generation hallucinations. To address this problem, we propose an innovative two-stage framework called CRVA-TGRAG (Conflict Resolution in Vulnerability Analysis via Teacher-Guided Retrieval-Augmented Generation). First, to improve document retrieval accuracy during the retrieval stage, we utilize Parent Document Segmentation and an ensemble retrieval scheme based on semantic similarity and inverted indexing. Second, to enhance LLMs’ capabilities based on the retrieval of CVE dataset in generation stage, we employ a teacher-guided preference optimization technique to fine-tune LLMs. Our framework not only enhances the quality of content retrieval through RAG but also leverages the advantages of preference fine-tuning in LLMs to answer questions more effectively and precisely. Experiments demonstrate our method achieves higher accuracy in retrieving the latest CVEs compared to external knowledge bases. In conclusion, our framework significantly mitigates potential knowledge conflicts and inconsistencies that may arise from relying solely on LLMs for knowledge retrieval.

114. Benchmarking Linguistic Adaptation in Comparable-Sized LLMs: A Study of Llama-3.1-8B, Mistral-7B-v0.1, and Qwen3-8B on Romanized Nepali

Authors: Ananda Rimal (Nepal Engineering College), Adarsha Rimal (Tribhuvan University)
URL: https://arxiv.org/abs/2604.14171
Abstract:

Romanized Nepali, the Nepali language written in the Latin alphabet, is the dominant medium for informal digital communication in Nepal, yet it remains critically underresourced in the landscape of Large Language Models (LLMs). This study presents a systematic benchmarking of linguistic adaptation across three comparable-sized open-weight models: Llama-3.1-8B, Mistral-7B-v0.1, and Qwen3-8B. We evaluate these architectures under zero-shot and fine-tuned settings using a curated bilingual dataset of 10,000 transliterated instruction-following samples. Performance is quantified across five metrics spanning seven measurement dimensions: Perplexity (PPL), BERTScore, chrF++, ROUGE-1, ROUGE-2, ROUGE-L, and BLEU, capturing fluency, phonetic consistency, and semantic integrity. Models were fine-tuned using Quantized Low-Rank Adaptation (QLoRA) with Rank-Stabilized LoRA (rsLoRA) at rank r=32 on dual NVIDIA Tesla T4 GPUs, training only approximately 1% of each model’s parameters in under 27 total GPU-hours. At zero-shot, all three models fail to generate Romanized Nepali, each exhibiting a distinct architecture-specific failure mode. Following fine-tuning, all three resolve these failures and converge to BERTScore approximately 0.75 and chrF++ greater than 23. Overall dimension-wise assessment across ten criteria identifies Qwen3-8B as the overall recommended architecture, being the only model to produce semantically relevant zero-shot output and leading all structural alignment metrics post-SFT. The adaptation headroom hypothesis is confirmed: Llama-3.1-8B, despite its weakest zero-shot baseline, achieves the largest absolute fine-tuning gains in PPL (Delta = -49.77) and BERTScore (Delta = +0.3287), making it the preferred choice for iterative low-resource development pipelines. This work establishes the first rigorous baseline for Romanized Nepali adaptation in comparable-sized open-weight LLMs.

115. Stateful Evidence-Driven Retrieval-Augmented Generation with Iterative Reasoning

Authors: Qi Dong , Ziheng Lin , Ning Ding
URL: https://arxiv.org/abs/2604.14170
Abstract:

Retrieval-Augmented Generation (RAG) grounds Large Language Models (LLMs) in external knowledge but often suffers from flat context representations and stateless retrieval, leading to unstable performance. We propose Stateful Evidence-Driven RAG with Iterative Reasoning, a framework that models question answering as a progressive evidence accumulation process. Retrieved documents are converted into structured reasoning units with explicit relevance and confidence signals and maintained in a persistent evidence pool capturing both supportive and non-supportive information. The framework performs evidence-driven deficiency analysis to identify gaps and conflicts and iteratively refines queries to guide subsequent retrieval. This iterative reasoning process enables stable evidence aggregation and improves robustness to noisy retrieval. Experiments on multiple question answering benchmarks demonstrate consistent improvements over standard RAG and multi-step baselines, while effectively accumulating high-quality evidence and maintaining stable performance under substantial retrieval noise.

116. Chinese Essay Rhetoric Recognition Using LoRA, In-context Learning and Model Ensemble

Authors: Yuxuan Lai , Xiajing Wang , Chen Zheng
URL: https://arxiv.org/abs/2604.14167
Abstract:

Rhetoric recognition is a critical component in automated essay scoring. By identifying rhetorical elements in student writing, AI systems can better assess linguistic and higher-order thinking skills, making it an essential task in the area of AI for education. In this paper, we leverage Large Language Models (LLMs) for the Chinese rhetoric recognition task. Specifically, we explore Low-Rank Adaptation (LoRA) based fine-tuning and in-context learning to integrate rhetoric knowledge into LLMs. We formulate the outputs as JSON to obtain structural outputs and translate keys to Chinese. To further enhance the performance, we also investigate several model ensemble methods. Our method achieves the best performance on all three tracks of CCL 2025 Chinese essay rhetoric recognition evaluation task, winning the first prize.

117. SeaAlert: Critical Information Extraction From Maritime Distress Communications with Large Language Models

Authors: Tomer Atia , Yehudit Aperstein , Alexander Apartsin
URL: https://arxiv.org/abs/2604.14163
Abstract:

Maritime distress communications transmitted over very high frequency (VHF) radio are safety-critical voice messages used to report emergencies at sea. Under the Global Maritime Distress and Safety System (GMDSS), such messages follow standardized procedures and are expected to convey essential details, including vessel identity, position, nature of the distress, and required assistance. In practice, however, automatic analysis remains difficult because distress messages are often brief, noisy, and produced under stress, may deviate from the prescribed format, and are further degraded by automatic speech recognition (ASR) errors caused by channel noise and speaker stress. This paper presents SeaAlert, an LLM-based framework for robust analysis of maritime distress communications. To address the scarcity of labeled real-world data, we develop a synthetic data generation pipeline in which an LLM produces realistic and diverse maritime messages, including challenging variants in which standard distress codewords are omitted or replaced with less explicit expressions. The generated utterances are synthesized into speech, degraded with simulated VHF noise, and transcribed by an ASR system to obtain realistic noisy transcripts.

118. Can Large Language Models Detect Methodological Flaws? Evidence from Gesture Recognition for UAV-Based Rescue Operation Based on Deep Learning

Authors: Domonkos Varga
URL: https://arxiv.org/abs/2604.14161
Abstract:

Reliable evaluation is essential in machine learning research, yet methodological flaws-particularly data leakage-continue to undermine the validity of reported results. In this work, we investigate whether large language models (LLMs) can act as independent analytical agents capable of identifying such issues in published studies. As a case study, we analyze a gesture-recognition paper reporting near-perfect accuracy on a small, human-centered dataset. We first show that the evaluation protocol is consistent with subject-level data leakage due to non-independent training and test splits. We then assess whether this flaw can be detected independently by six state-of-the-art LLMs, each analyzing the original paper without prior context using an identical prompt. All models consistently identify the evaluation as flawed and attribute the reported performance to non-independent data partitioning, supported by indicators such as overlapping learning curves, minimal generalization gap, and near-perfect classification results. These findings suggest that LLMs can detect common methodological issues based solely on published artifacts. While not definitive, their consistent agreement highlights their potential as complementary tools for improving reproducibility and supporting scientific auditing.

119. HUOZIIME: An On-Device LLM-enhanced Input Method for Deep Personalization

Authors: Baocai Shan , Yuzhuang Xu , Wanxiang Che
URL: https://arxiv.org/abs/2604.14159
Abstract:

Mobile input method editors (IMEs) are the primary interface for text input, yet they remain constrained to manual typing and struggle to produce personalized text. While lightweight large language models (LLMs) make on-device auxiliary generation feasible, enabling deeply personalized, privacy-preserving, and real-time generative IMEs poses fundamental this http URL this end, we present HUOZIIME, a personalized on-device IME powered by LLM. We endow HUOZIIME with initial human-like prediction ability by post-training a base LLM on synthesized personalization data. Notably, a hierarchical memory mechanism is designed to continually capture and leverage user-specific input history. Furthermore, we perform systemic optimizations tailored to on-device LLMbased IME deployment, ensuring efficient and responsive operation under mobile this http URL demonstrate efficient on-device execution and high-fidelity memory-driven personalization. Code and package are available at this https URL .

120. MemGround: Long-Term Memory Evaluation Kit for Large Language Models in Gamified Scenarios

Authors: Yihang Ding , Wanke Xia , Yiting Zhao , Jinbo Su , Jialiang Yang , Zhengbo Zhang , Ke Wang , Wenming Yang
URL: https://arxiv.org/abs/2604.14158
Abstract:

Current evaluations of long-term memory in LLMs are fundamentally static. By fixating on simple retrieval and short-context inference, they neglect the multifaceted nature of complex memory systems, such as dynamic state tracking and hierarchical reasoning in continuous interactions. To overcome these limitations, we propose MemGround, a rigorous long-term memory benchmark natively grounded in rich, gamified interactive scenarios. To systematically assess these capabilities, MemGround introduces a three-tier hierarchical framework that evaluates Surface State Memory, Temporal Associative Memory, and Reasoning-Based Memory through specialized interactive tasks. Furthermore, to comprehensively quantify both memory utilization and behavioral trajectories, we propose a multi-dimensional metric suite comprising Question-Answer Score (QA Overall), Memory Fragments Unlocked (MFU), Memory Fragments with Correct Order (MFCO), and Exploration Trajectory Diagrams (ETD). Extensive experiments reveal that state-of-the-art LLMs and memory agents still struggle with sustained dynamic tracking, temporal event association, and complex reasoning derived from long-term accumulated evidence in interactive environments.

LLM 관련 주요 논문 - 2026-04-17

1. Generalization in LLM Problem Solving: The Case of the Shortest Path

2. Diagnosing LLM Judge Reliability: Conformal Prediction Sets and Transitivity Violations

3. RadAgent: A tool-using AI agent for stepwise interpretation of chest computed tomography

4. Context Over Content: Exposing Evaluation Faking in Automated Judges

5. Meituan Merchant Business Diagnosis via Policy-Guided Dual-Process User Simulation

6. IG-Search: Step-Level Information Gain Rewards for Search-Augmented Reasoning

7. OpenMobile: Building Open Mobile Agents with Task and Trajectory Synthesis

8. From Reactive to Proactive: Assessing the Proactivity of Voice Agents via ProVoice-Bench

9. Autogenesis: A Self-Evolving Agent Protocol

10. Towards Faster Language Model Inference Using Mixture-of-Experts Flow Matching

11. COEVO: Co-Evolutionary Framework for Joint Functional Correctness and PPA Optimization in LLM-Based RTL Generation

12. Dr.~RTL: Autonomous Agentic RTL Optimization through Tool-Grounded Self-Improvement

13. Discovering Novel LLM Experts via Task-Capability Coevolution

14. Dual-Axis Generative Reward Model Toward Semantic and Turn-taking Robustness in Interactive Spoken Dialogue Models

15. ADAPT: Benchmarking Commonsense Planning under Unspecified Affordance Constraints

16. Governing Reflective Human-AI Collaboration: A Framework for Epistemic Scaffolding and Traceable Reasoning

17. The Missing Knowledge Layer in AI: A Framework for Stable Human-AI Reasoning

18. Beyond Literal Summarization: Redefining Hallucination for Medical SOAP Note Evaluation

19. The LLM Fallacy: Misattribution in AI-Assisted Cognitive Workflows

20. MirrorBench: Evaluating Self-centric Intelligence in MLLMs by Introducing a Mirror

21. CoTEvol: Self-Evolving Chain-of-Thoughts for Data Synthesis in Mathematical Reasoning

22. Disentangle-then-Refine: LLM-Guided Decoupling and Structure-Aware Refinement for Graph Contrastive Learning

23. The Agentification of Scientific Research: A Physicist’s Perspective

24. SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval

25. HWE-Bench: Benchmarking LLM Agents on Real-World Hardware Bug Repair Tasks

26. CAMO: An Agentic Framework for Automated Causal Discovery from Micro Behaviors to Macro Emergence in LLM Agent Simulations

27. DR$^{3}$-Eval: Towards Realistic and Reproducible Deep Research Evaluation

28. Acceptance Dynamics Across Cognitive Domains in Speculative Decoding

29. Rethinking Patient Education as Multi-turn Multi-modal Interaction

30. Targeted Exploration via Unified Entropy Control for Reinforcement Learning

31. Learning to Draw ASCII Improves Spatial Reasoning in Language Models

32. El Agente Forjador: Task-Driven Agent Generation for Quantum Simulation

33. GDPR Auto-Formalization with AI Agents and Human Verification

34. Enhancing Mental Health Counseling Support in Bangladesh using Culturally-Grounded Knowledge

35. TRACER: Trace-Based Adaptive Cost-Efficient Routing for LLM Classification

36. Dissecting Failure Dynamics in Large Language Model Reasoning

37. Quantifying Cross-Query Contradictions in Multi-Query LLM Reasoning

38. Pushing the Limits of On-Device Streaming ASR: A Compact, High-Accuracy English Model for Low-Latency Inference

39. Seeing Through Circuits: Faithful Mechanistic Interpretability for Vision Transformers

40. Evo-MedAgent: Beyond One-Shot Diagnosis with Agents That Remember, Reflect, and Improve

41. Response-Aware User Memory Selection for LLM Personalization

42. AIBuildAI: An AI Agent for Automatically Building AI Models

43. Credo: Declarative Control of LLM Pipelines via Beliefs and Policies

44. Seeing Through Experts Eyes A Foundational Vision Language Model Trained on Radiologists Gaze and Reasoning

45. GFT: From Imitation to Reward Fine-Tuning with Unbiased Group Advantages and Dynamic Coefficient Rectification

46. Simulating Human Cognition: Heartbeat-Driven Autonomous Thinking Activity Scheduling for LLM-based AI systems

47. NuHF Claw: A Risk Constrained Cognitive Agent Framework for Human Centered Procedure Support in Digital Nuclear Control Rooms

48. Why Do Vision Language Models Struggle To Recognize Human Emotions?

49. Prism: Symbolic Superoptimization of Tensor Programs

50. CoopEval: Benchmarking Cooperation-Sustaining Mechanisms and LLM Agents in Social Dilemmas

51. VisPCO: Visual Token Pruning Configuration Optimization via Budget-Aware Pareto-Frontier Learning for Vision-Language Models

52. Scepsy: Serving Agentic Workflows Using Aggregate LLM Pipelines

53. Compressing Sequences in the Latent Embedding Space: $K$-Token Merging for Large Language Models

54. IUQ: Interrogative Uncertainty Quantification for Long-Form Large Language Model Generation

55. Autonomous Evolution of EDA Tools: Multi-Agent Self-Evolved ABC

56. Route to Rome Attack: Directing LLM Routers to Expensive Models via Adversarial Suffix Optimization

57. UniDoc-RL: Coarse-to-Fine Visual RAG with Hierarchical Actions and Dense Rewards

58. Calibration-Gated LLM Pseudo-Observations for Online Contextual Bandits

59. RaTA-Tool: Retrieval-based Tool Selection with Multimodal Large Language Models

60. Can LLMs Score Medical Diagnoses and Clinical Reasoning as well as Expert Panels?

61. Reasoning Dynamics and the Limits of Monitoring Modality Reliance in Vision-Language Models

62. RACER: Retrieval-Augmented Contextual Rapid Speculative Decoding

63. Vibe-Coding: Feedback-Based Automated Verification with no Human Code Inspection, a Feasibility Study

64. MetaDent: Labeling Clinical Images for Vision-Language Models in Dentistry

65. Schema Key Wording as an Instruction Channel in Structured Generation under Constrained Decoding

66. ClimateCause: Complex and Implicit Causal Structures in Climate Reports

67. Zero-Shot Retail Theft Detection via Orchestrated Vision Models: A Model-Agnostic, Cost-Effective Alternative to Trained Single-Model Systems

68. Which bird does not have wings: Negative-constrained KGQA with Schema-guided Semantic Matching and Self-directed Refinement

69. Bounded Autonomy for Enterprise AI: Typed Action Contracts and Consumer-Side Execution

70. Fact4ac at the Financial Misinformation Detection Challenge Task: Reference-Free Financial Misinformation Detection via Fine-Tuning and Few-Shot Prompting of Large Language Models

71. StoryCoder: Narrative Reformulation for Structured Reasoning in LLM Code Generation

72. ELMoE-3D: Leveraging Intrinsic Elasticity of MoE for Hybrid-Bonding-Enabled Self-Speculative Decoding in On-Premises Serving

73. Retrieve, Then Classify: Corpus-Grounded Automation of Clinical Value Set Authoring

74. Hijacking Large Audio-Language Models via Context-Agnostic and Imperceptible Auditory Prompt Injection

75. CausalDetox: Causal Head Selection and Intervention for Language Model Detoxification

76. Mechanistic Decoding of Cognitive Constructs in LLMs

77. AgileLog: A Forkable Shared Log for Agents on Data Streams

78. CPGRec+: A Balance-oriented Framework for Personalized Video Game Recommendations

79. Generative Augmented Inference