LLM 관련 주요 논문 - 2026-04-28

1. Case-Specific Rubrics for Clinical AI Evaluation: Methodology, Validation, and LLM-Clinician Agreement Across 823 Encounters

Authors: Aaryan Shah , Andrew Hines , Alexia Downs , Denis Bajet , Paulius Mui , Fabiano Araujo , Laura Offutt , Aida Rutledge , Elizabeth Jimenez
URL: https://arxiv.org/abs/2604.24710
Abstract:

Objective. Clinical AI documentation systems require evaluation methodologies that are clinically valid, economically viable, and sensitive to iterative changes. Methods requiring expert review per scoring instance are too slow and expensive for safe, iterative deployment. We present a case-specific, clinician-authored rubric methodology for clinical AI evaluation and examine whether LLM-generated rubrics can approximate clinician agreement. Materials and Methods. Twenty clinicians authored 1,646 rubrics for 823 clinical cases (736 real-world, 87 synthetic) across primary care, psychiatry, oncology, and behavioral health. Each rubric was validated by confirming that an LLM-based scoring agent consistently scored clinician-preferred outputs higher than rejected ones. Seven versions of an EHR-embedded AI agent for clinicians were evaluated across all cases. Results. Clinician-authored rubrics discriminated effectively between high- and low-quality outputs (median score gap: 82.9%) with high scoring stability (median range: 0.00%). Median scores improved from 84% to 95%. In later experiments, clinician-LLM ranking agreement (tau: 0.42-0.46) matched or exceeded clinician-clinician agreement (tau: 0.38-0.43), attributable to both ceiling compression and LLM rubric improvement. Discussion. This convergence supports incorporating LLM rubrics alongside clinician-authored ones. At roughly 1,000 times lower cost, LLM rubrics enable substantially greater evaluation coverage, while continued clinical authorship grounds evaluation in expert judgment. Ceiling compression poses a methodological challenge for future inter-rater agreement studies. Conclusion. Case-specific rubrics offer a path for clinical AI evaluation that preserves expert judgment while enabling automation at three orders lower cost. Clinician-authored rubrics establish the baseline against which LLM rubrics are validated.

2. The Price of Agreement: Measuring LLM Sycophancy in Agentic Financial Applications

Authors: Zhenyu Zhao , Aparna Balagopalan , Adi Agrawal , Dilshoda Yergasheva , Waseem Alshikh , Daniel M. Bikel
URL: https://arxiv.org/abs/2604.24668
Abstract:

Given the increased use of LLMs in financial systems today, it becomes important to evaluate the safety and robustness of such systems. One failure mode that LLMs frequently display in general domain settings is that of sycophancy. That is, models prioritize agreement with expressed user beliefs over correctness, leading to decreased accuracy and trust. In this work, we focus on evaluating sycophancy that LLMs display in agentic financial tasks. Our findings are three-fold: first, we find the models show only low to modest drops in performance in the face of user rebuttals or contradictions to the reference answer, which distinguishes sycophancy that models display in financial agentic settings from findings in prior work. Second, we introduce a suite of tasks to test for sycophancy by user preference information that contradicts the reference answer and find that most models fail in the presence of such inputs. Lastly, we benchmark different modes of recovery such as input filtering with a pretrained LLM.

3. XGRAG: A Graph-Native Framework for Explaining KG-based Retrieval-Augmented Generation

Authors: Zhuoling Li , Ha Linh Hong Tran Nguyen , Valeria Bladinieres , Maxim Romanovsky
URL: https://arxiv.org/abs/2604.24623
Abstract:

Graph-based Retrieval-Augmented Generation (GraphRAG) extends traditional RAG by using knowledge graphs (KGs) to give large language models (LLMs) a structured, semantically coherent context, yielding more grounded answers. However, GraphRAG reasoning process remains a black-box, limiting our ability to understand how specific pieces of structured knowledge influence the final output. Existing explainability (XAI) methods for RAG systems, designed for text-based retrieval, are limited to interpreting an LLM response through the relational structures among knowledge components, creating a critical gap in transparency and trustworthiness. To address this, we introduce XGRAG, a novel framework that generates causally grounded explanations for GraphRAG systems by employing graph-based perturbation strategies, to quantify the contribution of individual graph components on the model answer. We conduct extensive experiments comparing XGRAG against RAG-Ex, an XAI baseline for standard RAG, and evaluate its robustness across various question types, narrative structures and LLMs. Our results demonstrate a 14.81% improvement in explanation quality over the baseline RAG-Ex across NarrativeQA, FairyTaleQA, and TriviaQA, evaluated by F1-score measuring alignment between generated explanations and original answers. Furthermore, XGRAG explanations exhibit a strong correlation with graph centrality measures, validating its ability to capture graph structure. XGRAG provides a scalable and generalizable approach towards trustworthy AI through transparent, graph-based explanations that enhance the interpretability of RAG systems.

4. Evaluating whether AI models would sabotage AI safety research

Authors: Robert Kirk , Alexandra Souly , Kai Fronsdal , Abby D’Cruz , Xander Davies
URL: https://arxiv.org/abs/2604.24618
Abstract:

We evaluate the propensity of frontier models to sabotage or refuse to assist with safety research when deployed as AI research agents within a frontier AI company. We apply two complementary evaluations to four Claude models (Mythos Preview, Opus 4.7 Preview, Opus 4.6, and Sonnet 4.6): an unprompted sabotage evaluation testing model behaviour with opportunities to sabotage safety research, and a sabotage continuation evaluation testing whether models continue to sabotage when placed in trajectories where prior actions have started undermining research. We find no instances of unprompted sabotage across any model, with refusal rates close to zero for Mythos Preview and Opus 4.7 Preview, though all models sometimes only partially completed tasks. In the continuation evaluation, Mythos Preview actively continues sabotage in 7% of cases (versus 3% for Opus 4.6, 4% for Sonnet 4.6, and 0% for Opus 4.7 Preview), and exhibits reasoning-output discrepancy in the majority of these cases, indicating covert sabotage reasoning. Our evaluation framework builds on Petri, an open-source LLM auditing tool, with a custom scaffold running models inside Claude Code, alongside an iterative pipeline for generating realistic sabotage trajectories. We measure both evaluation awareness and a new form of situational awareness termed “prefill awareness”, the capability to recognise that prior trajectory content was not self-generated. Opus 4.7 Preview shows notably elevated unprompted evaluation awareness, while prefill awareness remains low across all models. Finally, we discuss limitations including evaluation awareness confounds, limited scenario coverage, and untested pathways to risk beyond safety research sabotage.

5. A systematic evaluation of vision-language models for observational astronomical reasoning tasks

Authors: Wenke Ren , Hengxiao Guo , Wenwen Zuo , Xiaoman Zhang
URL: https://arxiv.org/abs/2604.24589
Abstract:

Vision-language models (VLMs) are increasingly proposed as general-purpose tools for scientific data interpretation, yet their reliability on real astronomical observations across diverse modalities remains untested. We present AstroVLBench, a comprehensive benchmark comprising over 4,100 expert-verified instances across five tasks spanning optical imaging, radio interferometry, multi-wavelength photometry, time-domain light curves, and optical spectroscopy. Evaluating six frontier models, we find that performance is strongly modality-dependent: while one model (Gemini 3 Pro) emerges as the most consistently capable across tasks, task-specific strengths vary, and all models substantially underperform domain-specialized methods. Mechanistic ablations reveal that performance depends not only on directing attention to salient visual features but also on grounding those features in physical knowledge. Phenomenological prompts describing what to look for improve accuracy by sharpening model focus, but physical prompts explaining why those features matter perform better overall and yield more balanced classifications with reduced class-specific bias. Consistent with this picture, presenting the underlying one-dimensional measurements directly as numerical tables instead of rendered plots yields up to 13 percentage points improvement. Reasoning quality analysis further demonstrates that, without explicit physical grounding, models may reach correct predictions from phenomenologically plausible cues while providing physically imprecise justifications, establishing that accuracy alone is insufficient for trustworthy scientific deployment. These findings provide the first systematic, multi-modal baselines for VLMs in observational astronomy and identify the specific representation, grounding, and reasoning bottlenecks where current models fail.

6. Towards Lawful Autonomous Driving: Deriving Scenario-Aware Driving Requirements from Traffic Laws and Regulations

Authors: Bowen Jian , Rongjie Yu , Hong Wang , Liqiang Wang , Zihang Zou
URL: https://arxiv.org/abs/2604.24562
Abstract:

Driving in compliance with traffic laws and regulations is a basic requirement for human drivers, yet autonomous vehicles (AVs) can violate these requirements in diverse real-world scenarios. To encode law compliance into AV systems, conventional approaches use formal logic languages to explicitly specify behavioral constraints, but this process is labor-intensive, hard to scale, and costly to maintain. With recent advances in artificial intelligence, it is promising to leverage large language models (LLMs) to derive legal requirements from traffic laws and regulations. However, without explicitly grounding and reasoning in structured traffic scenarios, LLMs often retrieve irrelevant provisions or miss applicable ones, yielding imprecise requirements. To address this, we propose a novel pipeline that grounds LLM reasoning in a traffic scenario taxonomy through node-wise anchors that encode hierarchical semantics. On Chinese traffic laws and OnSite dataset (5,897 scenarios), our method improves law-scenario matching by 29.1\% and increases the accuracy of derived mandatory and prohibitive requirements by 36.9\% and 38.2\%, respectively. We further demonstrate real-world applicability by constructing a law-compliance layer for AV navigation and developing an onboard, real-time compliance monitor for in-field testing, providing a solid foundation for future AV development, deployment, and regulatory oversight.

7. STELLAR-E: a Synthetic, Tailored, End-to-end LLM Application Rigorous Evaluator

Authors: Alessio Sordo , Lingxiao Du , Meeka-Hanna Lenisa , Evgeny Bogdanov , Maxim Romanovsky
URL: https://arxiv.org/abs/2604.24544
Abstract:

The increasing reliance on Large Language Models (LLMs) across diverse sectors highlights the need for robust domain-specific and language-specific evaluation datasets; however, the collection of such datasets is challenging due to privacy concerns, regulatory restrictions, and the time cost for manual creation. Existing automated benchmarking methods are often limited by relying on pre-existing data, poor scalability, single-domain focus, and lack of multilingual support. We present STELLAR-E - a fully automated system to generate high-quality synthetic datasets of custom size, using minimal human inputs without depending on existing datasets. The system is structured in two stages: (1) We modify the TGRT Self-Instruct framework to create a synthetic data engine that enables controllable, custom synthetic dataset generation, and (2) an evaluation pipeline incorporating statistical and LLM-based metrics to assess the applicability of the synthetic dataset for LLM-based application evaluations. The synthetic datasets reach an average difference of +5.7% in terms of LLM-as-a-judge scores against existing language-specific benchmarks, demonstrating comparable quality for comprehensive assessment of big and small LLMs. While real datasets remain slightly more challenging for LLMs especially for smaller models, this work establishes a scalable and domain-adaptable benchmarking framework that supports fair evaluation of LLM applications, offering a faster alternative to manual approaches and enabling high-efficiency automated quality assurance cycles.

8. Beyond the Attention Stability Boundary: Agentic Self-Synthesizing Reasoning Protocols

Authors: Dahlia Shehata , Ming Li
URL: https://arxiv.org/abs/2604.24512
Abstract:

As LLM agents transition to autonomous digital coworkers, maintaining deterministic goal-directedness in non-linear multi-turn conversations emerged as an architectural bottleneck. We identify and formalize a systemic failure mode termed the Attention Latch in decoder-only autoregressive Transformers. This phenomenon, a behavioral manifestation of Information Over-squashing, occurs when the cumulative probabilistic weight of historical context overrides mid-task updates, causing agents to remain anchored to obsolete constraints despite explicit contradictory instructions. We propose Self-Synthesizing Reasoning Protocols (SSRP), a metacognitive framework that implements a discrete separation between high-level architectural planning (Architect) and turn-by-turn procedural execution (Executive). We evaluate SSRP across 9K trajectories using the MultiWOZ 2.2 dataset and the Aggregate Pivot Accuracy (APA), a novel metric we validate by mapping its scores to the U-shaped ‘Lost in the Middle’ curve. We present 3 experimental tiers: a shallow recency-based retrieval pilot, a high-entropy SOP, and a semantic hijacked 3-hop Multi-Fact Synthesis task. Our results empirically locate the Attention Stability Boundary, where stateless Vanilla ReAct baselines for GPT 5.4 collapse to 0.1% success while SSRP achieves a 715X Resilience Lift. We demonstrate statistically significant gains across Gemini 3.1 Pro, Claude Sonnet 4.6 and DeepSeek V3.2. Audits confirm SSRP necessity by proving attentional lapse via a recursive reflexion baseline (100% success); decoupling the latch from positional bias through equidistant stress testing (90% accuracy); and formalizing SSRP via the Information Bottleneck principle and granularity ablations. Procedural Integrity audit (98.8% adherence) reveals a Grounding Paradox where high-stability models fail by refusing to hallucinate under retrieval-reasoning contamination.

9. Agentic clinical reasoning over longitudinal myeloma records: a retrospective evaluation against expert consensus

Authors: Johannes Moll , Jannik Lübberstedt , Christoph Nuernbergk , Jacob Stroh , Luisa Mertens , Anna Purcarea , Christopher Zirn , Zeineb Benchaaben , Fabian Drexel , Hartmut Häntze , Anirudh Narayanan , Friedrich Puttkammer , Andrei Zhukov , Jacqueline Lammert , Sebastian Ziegelmayer , Markus Graf , Marion Högner , Marcus Makowski , Florian Bassermann , Lisa C. Adams , Jiazhen Pan , Daniel Rueckert , Krischan Braitsch , Keno K. Bressem
URL: https://arxiv.org/abs/2604.24473
Abstract:

Multiple myeloma is managed through sequential lines of therapy over years to decades, with each decision depending on cumulative disease history distributed across dozens to hundreds of heterogeneous clinical documents. Whether LLM-based systems can synthesise this evidence at a level approaching expert agreement has not been established. A retrospective evaluation was conducted on longitudinal clinical records of 811 myeloma patients treated at a tertiary centre (2001-2026), covering 44,962 documents and 1,334,677 laboratory values, with external validation on MIMIC-IV. An agentic reasoning system was compared against single-pass retrieval-augmented generation (RAG), iterative RAG, and full-context input on 469 patient-question pairs from 48 templates at three complexity levels. Reference labels came from double annotation by four oncologists with senior haematologist adjudication. Iterative RAG and full-context input converged on a shared ceiling (75.4% vs 75.8%, p = 1.00). The agentic system reached 79.6% concordance (95% CI 76.4-82.8), exceeding both baselines (+3.8 and +4.2 pp; p = 0.006 and 0.007). Gains rose with question complexity, reaching +9.4 pp on criteria-based synthesis (p = 0.032), and with record length, reaching +13.5 pp in the top decile (n = 10). The system error rate (12.2%) was comparable to expert disagreement (13.6%), but severity was inverted: 57.8% of system errors were clinically significant versus 18.8% of expert disagreements. Agentic reasoning was the only approach to exceed the shared ceiling, with gains concentrated on the most complex questions and longest records. The greater clinical consequence of residual system errors indicates that prospective evaluation in routine care is required before these findings translate into patient benefit.

10. PhysNote: Self-Knowledge Notes for Evolvable Physical Reasoning in Vision-Language Model

Authors: Sinin Zhang , Yunfei Xie , Yuxuan Cheng , Haoyu Zhang , Tong Zhang
URL: https://arxiv.org/abs/2604.24443
Abstract:

Vision-Language Models (VLMs) have demonstrated strong performance on textbook-style physics problems, yet they frequently fail when confronted with dynamic real-world scenarios that require temporal consistency and causal reasoning across frames. We identify two fundamental challenges underlying these failures: (1) spatio-temporal identity drift, where objects lose their physical identity across successive frames and break causal chains, and (2) volatility of inference-time insights, where a model may occasionally produce correct physical reasoning but never consolidates it for future reuse. To address these challenges, we propose PhysNote, an agentic framework that enables VLMs to externalize and refine physical knowledge through self-generated “Knowledge Notes.” PhysNote stabilizes dynamic perception through spatio-temporal canonicalization, organizes self-generated insights into a hierarchical knowledge repository, and drives an iterative reasoning loop that grounds hypotheses in visual evidence before consolidating verified knowledge. Experiments on PhysBench demonstrate that PhysNote achieves 56.68% overall accuracy, a 4.96% improvement over the best multi-agent baseline, with consistent gains across all four physical reasoning domains.

11. Aligning with Your Own Voice: Self-Corrected Preference Learning for Hallucination Mitigation in LVLMs

Authors: Byeonggeuk Lim , JungMin Yun , Junehyoung Kwon , Kyeonghyun Kim , YoungBin Kim
URL: https://arxiv.org/abs/2604.24395
Abstract:

Large Vision-Language Models (LVLMs) frequently suffer from hallucinations. Existing preference learning-based approaches largely rely on proprietary models to construct preference datasets. We identify that this reliance introduces a distributional mismatch between the proprietary and target models that hinders efficient alignment. To address this, we propose Alignment via VErified Self-correction DPO (AVES-DPO), a framework that aligns LVLMs using in-distribution data derived from the model’s intrinsic knowledge. Our approach employs a consensus-based verification mechanism to diagnose diverse hallucinations and guides the model to self-correct, thereby generating preference pairs strictly compatible with its internal distribution. Extensive experiments demonstrate that AVES-DPO surpasses existing baselines in hallucination mitigation while requiring only 5.2k samples.

12. Adaptive ToR: Complexity-Aware Tree-Based Retrieval for Pareto-Optimal Multi-Intent NLU

Authors: Hee-Kyong Yoo , Wonbae Kim , Hyocheol Ahn
URL: https://arxiv.org/abs/2604.24219
Abstract:

Multi-intent natural language understanding requires retrieval systems that simultaneously achieve high accuracy and computational efficiency, yet existing approaches apply either uniform single-step retrieval that compromises recall or fixed-depth hierarchical decomposition that introduces excessive latency regardless of query complexity. This paper proposes Adaptive Tree-of-Retrieval (Adaptive ToR), a complexity-aware retrieval architecture that dynamically configures retrieval topology based on query characteristics. The system integrates four components: (1) a Query Tree Classifier computing a Query Complexity Index from weighted linguistic signals to route queries to either a rapid single-step path or an adaptive-depth hierarchical path; (2) a Tree-Based Retrieval module that recursively decomposes complex queries into focused sub-queries calibrated to predicted complexity; (3) an Adaptive Pruning Module employing two-stage filtering combining quantitative similarity gating with semantic relevance evaluation to suppress exponential node growth; and (4) a Retrieval Reranking Layer featuring a deduplicator-first pipeline and global LLM rescoring for production efficiency. Evaluation on the NLU++ benchmark (2,693 multi-intent queries across Banking and Hotel domains) yields 29.07% Subset Accuracy and 71.79% Micro-F1, a 9.7% relative improvement over fixed-depth baselines, while reducing latency by 37.6%, LLM invocations by 43.0%, and token consumption by 9.8%. Depth-wise analysis reveals that 26.92% of queries resolve within three seconds (2.45s mean latency) via single-step routing (d=0: 37.9% Subset Accuracy, 74.8% Micro-F1), while token consumption scales by 4.9x across depths, validating complexity-aware resource allocation and establishing Pareto-optimal balance across accuracy, latency, and computational efficiency.

13. Multi-Dimensional Evaluation of Sustainable City Trips with LLM-as-a-Judge and Human-in-the-Loop

Authors: Ashmi Banerjee , Adithi Satish , Wolfgang Wörndl , Yashar Deldjoo
URL: https://arxiv.org/abs/2604.24158
Abstract:

Evaluating nuanced conversational travel recommendations is challenging when human annotations are costly and standard metrics ignore stakeholder-centric goals. We study LLMs-as-Judges for sustainable city-trip lists across four dimensions – relevance, diversity, sustainability, and popularity balance, and propose a three-phase calibration framework: (1) baseline judging with multiple LLMs, (2) expert evaluation to identify systematic misalignment, and (3) dimension-specific calibration via rules and few-shot examples. Across two recommendation settings, we observe model-specific biases and high dimension-level variance, even when judges agree on overall rankings. Calibration clarifies reasoning per dimension but exposes divergent interpretations of sustainability, highlighting the need for transparent, bias-aware LLM evaluation. Prompts and code are released for reproducibility: this https URL .

14. An Information-Geometric Framework for Stability Analysis of Large Language Models under Entropic Stress

Authors: Hikmat Karimov , Rahid Zahid Alekberli
URL: https://arxiv.org/abs/2604.24076
Abstract:

As large language models (LLMs) are increasingly deployed in high-stakes and operational settings, evaluation strategies based solely on aggregate accuracy are often insucient to characterize system reliability. This study proposes a thermodynamic inspired modeling framework for analyzing the stability of LLM outputs under conditions of uncertainty and perturbation. The framework introduces a composite stability score that integrates task utility, entropy as a measure of external uncertainty, and two internal structural proxies: internal integration and aligned reective capacity. Rather than interpreting these quantities as physical variables, the formulation is intended as an interpretable abstraction that captures how internal structure may modulate the impact of disorder on model behavior. Using the IST-20 benchmarking protocol and associated metadata, we analyze 80 modelscenario observations across four contemporary LLMs. The proposed formulation consistently yields higher stability scores than a reduced utilityentropy baseline, with a mean improvement of 0.0299 (95% CI: 0.02470.0351). The observed gain is more pronounced under higher entropy conditions, suggesting that the framework captures a form of nonlinear attenuation of uncertainty. We do not claim a fundamental physical law or a complete theory of machine ethics. Instead, the contribution of this work is a compact and interpretable modeling perspective that connects uncertainty, performance, and internal structure within a unied evaluation lens. The framework is intended to complement existing benchmarking approaches and to support ongoing discussions in AI safety, reliability, and governance.

15. Grounding Before Generalizing: How AI Differs from Humans in Causal Transfer

Authors: Liangru Xiang , Yuxi Ma , Zhihao Cao , Yixin Zhu , Song-Chun Zhu
URL: https://arxiv.org/abs/2604.24062
Abstract:

Extracting abstract causal structures and applying them to novel situations is a hallmark of human intelligence. While Large Language Models (LLMs) and Vision Language Models (VLMs) have shown strong performance on a wide range of reasoning tasks, their capacity for interactive causal learning – inducing latent structures through sequential exploration and transferring them across contexts – remains uncharacterized. Human learners accomplish such transfer after minimal exposure, whereas classical Reinforcement Learning (RL) agents fail catastrophically. Whether state-of-the-art Artificial Intelligence (AI) models possess human-like mechanisms for abstract causal structure transfer is an open question. Using the OpenLock paradigm requiring sequential discovery of Common Cause (CC) and Common Effect (CE) structures, here we show that models exhibit fundamentally delayed or absent transfer: even successful models require initial environmental-specific mapping – what we term environmental grounding – before efficiency gains emerge, whereas humans leverage prior structural knowledge from the very first solution attempt. In the text-only condition, models matched or exceeded human discovery efficiency. In contrast, visual information – in both the image-only and text-and-image conditions – overall degraded rather than enhanced performance, revealing a broad reliance on symbolic processing rather than integrated multimodal reasoning. Models further exhibited systematic CC/CE asymmetries absent in humans, suggesting heuristic biases rather than direction-neutral causal abstraction. These findings reveal that large-scale statistical learning does not produce the decontextualized causal schemas underpinning human analogical reasoning, establishing grounding-dependent transfer as a fundamental limitation of current LLMs and VLMs.

16. A2DEPT: Large Language Model-Driven Automated Algorithm Design via Evolutionary Program Trees

Authors: Bin Chen , Shouliang Zhu , Beidan Liu , Yong Zhao , Tianle Pu , Huichun Li , Zhengqiu Zhu
URL: https://arxiv.org/abs/2604.24043
Abstract:

Designing heuristics for combinatorial optimization problems (COPs) is a fundamental yet challenging task that traditionally requires extensive domain expertise. Recently, Large Language Model (LLM)-based Automated Heuristic Design (AHD) has shown promise in autonomously generating heuristic components with minimal human intervention. However, most existing LLM-based AHD methods enforce fixed algorithmic templates to ensure executability, which confines the search to component-level tuning and limits system-level algorithmic expressiveness. To enable open-ended solver synthesis beyond rigid templates, we propose Automated Algorithm Design via Evolutionary Program Trees (A2DEPT), which treats LLMs as system-level algorithm architects. A2DEPT explores the vast program space via a tree-structured evolutionary search with hybrid selection and hierarchical operators, enabling iterative refinement of complete algorithms. To make open-ended generation practical, we enforce executability with a lightweight program-maintenance loop that performs feedback-driven repair. In experiments, A2DEPT consistently outperforms representative LLM-based baselines on both standard and highly constrained benchmarks. On the standard benchmarks, it reduces the mean normalized optimality gap by 9.8% relative to the strongest competing AHD baseline.

17. Representational Curvature Modulates Behavioral Uncertainty in Large Language Models

Authors: Jack King , Evelina Fedorenko , Eghbal A. Hosseini
URL: https://arxiv.org/abs/2604.23985
Abstract:

In autoregressive large language models (LLMs), temporal straightening offers an account of how the next-token prediction objective shapes representations. Models learn to progressively straighten the representational trajectory of input sequences across layers, potentially facilitating next-token prediction via linear extrapolation. However, a direct link between this trajectory and token-level behavior has been missing. We provide such a link by relating contextual curvature-a geometric measure of how sharply the representational trajectory bends over recent context-to next-token entropy. Across two models (GPT-2 XL and Pythia-2.8B), contextual curvature is correlated with entropy, and this relationship emerges during training. Perturbation experiments reveal selective dependence: manipulating curvature through trajectory-aligned interventions reliably modulates entropy, while geometrically misaligned perturbations have no effect. Finally, regularizing representations to be straighter during training modestly reduces token-level entropy without degrading validation loss. These results identify trajectory curvature as a task-aligned representational feature that influences behavioral uncertainty in LLMs.

Authors: Aydin Ayanzadeh , Tim Oates
URL: https://arxiv.org/abs/2604.23970
Abstract:

Indoor navigation remains a critical accessibility challenge for the blind and low-vision (BLV) individuals, as existing solutions rely on costly per-building infrastructure. We present an agentic framework that converts a single floor plan image into a structured, retrievable knowledge base to generate safe, accessible navigation instructions with lightweight infrastructure. The system has two phases: a multi-agent module that parses the floor plan into a spatial knowledge graph through a self-correcting pipeline with iterative retry loops and corrective feedback; and a Path Planner that generates accessible navigation instructions, with a Safety Evaluator agent assessing potential hazards along each route. We evaluate the system on the real-world UMBC Math and Psychology building (floors MP-1 and MP-3) and on the CVC-FP benchmark. On MP-1, we achieve success rates of 92.31%, 76.92%, and 61.54% for short, medium, and long routes, outperforming the strongest single-call baseline (Claude 3.7 Sonnet) at 84.62%, 69.23%, and 53.85%. On MP-3, we reach 76.92%, 61.54%, and 38.46%, compared to the best baseline at 61.54%, 46.15%, and 23.08%. These results show consistent gains over single-call LLM baselines and demonstrate that our workflow is a scalable solution for accessible indoor navigation for BLV individuals.

19. Context-Aware Hospitalization Forecasting Evaluations for Decision Support using LLMs

Authors: Rhea Makkuni , Ananya Joshi
URL: https://arxiv.org/abs/2604.23949
Abstract:

Medical and public health experts must make real-time resource decisions, such as expanding hospital bed capacity, based on projected hospitalization trends during large-scale healthcare disruptions (e.g., operational failures or pandemics). Forecasting models can assist in this task by analyzing large volumes of resource-related data at the facility level, but they must be reliable for decision-making under real-world data conditions. Recent work shows that large language models (LLMs) can incorporate richer forms of context into numerical forecasting. Whereas traditional models rely primarily on temporal context (i.e., past observations), LLMs can also leverage non-temporal public health context such as demographic, geographic, and population-level features. However, it remains unclear how these models should be used to produce stable or decision-relevant predictions in real-world healthcare settings. To evaluate how LLMs can be effectively used in this setting, we evaluate three approaches across 60 counties with low-,mid-, and high-hospitalization intensities in the United States: direct LLM-based forecasting, classical time-series models, and a context-augmented hybrid pipeline (HybridARX) that incorporates LLM-derived signals into structured models. Because the goal is operational decision-making rather than error minimization alone, we evaluate performance with bias and lead-lag alignment in addition to standard forecasting metrics. Our results show that HybridARX improves over classical ARX by yielding more stable and better-calibrated forecasts, particularly when incorporating noisy contextual signals into structured time-series models. These findings suggest that, in non-stationary healthcare resource forecasting, LLMs are most useful when embedded within structured hybrid models.

20. LLM-Augmented Traffic Signal Control with LSTM-Based Traffic State Prediction and Safety-Constrained Decision Support

Authors: Jiazhao Shi
URL: https://arxiv.org/abs/2604.23902
Abstract:

Traffic signal control is a critical task in intelligent transportation systems, yet conventional fixed-time and rule-based methods often struggle to adapt to dynamic traffic demand and provide limited decision interpretability. This study proposes an LLM-augmented traffic signal control framework that integrates LSTM-based short-term traffic state prediction, predictive phase selection, structured large language model reasoning, and safety-constrained action filtering. The LSTM module forecasts future queue length, waiting time, vehicle count, and lane occupancy based on recent intersection-level observations. A predictive controller then generates candidate signal actions, while the LLM module evaluates these actions using structured traffic-state inputs and produces congestion diagnoses, phase adjustment recommendations, and natural-language explanations. To ensure operational reliability, all LLM-generated recommendations are validated by a safety filter before execution. Simulation-based experiments in SUMO compare the proposed method with fixed-time control, rule-based control, and an LSTM-based predictive baseline under balanced demand, directional peak demand, and sudden surge scenarios. The results indicate that the proposed framework improves traffic efficiency, especially under dynamic and non-recurrent traffic conditions, while maintaining zero constraint violations after safety filtering. Overall, this study demonstrates that LLMs can enhance traffic signal control when used as constrained reasoning and decision-support modules rather than direct low-level controllers. Keywords: Intelligent Transportation Systems; Traffic Signal Control; Large Language Models; LSTM; Traffic State Prediction; Decision Support; Safety-Constrained Control; SUMO Simulation.

21. ZenBrain: A Neuroscience-Inspired 7-Layer Memory Architecture for Autonomous AI Systems

Authors: Alexander Bering
URL: https://arxiv.org/abs/2604.23878
Abstract:

Despite a century of empirical memory research, existing AI agent memory systems rely on system-engineering metaphors (virtual-memory paging, flat LLM storage, Zettelkasten notes), none integrating principles of consolidation, forgetting, and reconsolidation. We present ZenBrain, a multi-layer memory architecture integrating fifteen neuroscience models. It implements seven memory layers (working, short-term, episodic, semantic, procedural, core, cross-context) orchestrated by nine foundational algorithms (Two-Factor Synaptic Model, vmPFC-coupled FSRS, Simulation-Selection sleep, Bayesian confidence, and five more) plus six new Predictive Memory Architecture (PMA) components: a four-channel NeuromodulatorEngine, prediction-error-gated ReconsolidationEngine, TripleCopyMemory with divergent decay, four-dimensional PriorityMap with amygdala fast-path, StabilityProtector (NogoA/HDAC3 analogue), and MetacognitiveMonitor for bias detection. The 15-algorithm ablation reveals a cooperative survival network: under stress, 9 of 15 algorithms become individually critical (delta-Q up to -93.7%, Wilcoxon, 10 seeds, alpha=0.005). Simulation-Selection sleep achieves 37% stability improvement (p<0.005) with 47.4% storage reduction. TripleCopyMemory retains S(t)=0.912 at 30 days; PriorityMap reaches NDCG@10=0.997. Multi-layer routing beats a flat single-layer baseline by 20.7% F1 on LoCoMo (p<0.005) and 19.5% on MemoryArena (p=0.015). On LongMemEval-500, ZenBrain holds the highest mean rank on all 12 system-judge cells (4 systems x 3 LLM judges), three-judge mean J=0.545 vs letta=0.485, a-mem=0.414, mem0=0.394; all 9 pair-wise contrasts clear Bonferroni (alpha=0.05/18, min p=6.2e-31, d in [0.18, 0.52]). Under LongMemEval’s binary judge, ZenBrain reaches 91.3% of oracle accuracy at 1/106th the per-query token budget. Open-source with 11,589 automated test cases.

22. ClawTrace: Cost-Aware Tracing for LLM Agent Skill Distillation

Authors: Boqin Yuan , Renchu Song , Yue Su , Sen Yang , Jing Qin
URL: https://arxiv.org/abs/2604.23853
Abstract:

Skill-distillation pipelines learn reusable rules from LLM agent trajectories, but they lack a key signal: how much each step costs. Without per-step cost, a pipeline cannot distinguish adding a missing step to fix a bug from removing an expensive step that never affected the outcome. We introduce ClawTrace, an agent tracing platform that records every LLM call, tool use, and sub-agent spawn during an agent session and compiles each session into a TraceCard: a compact YAML summary with per-step USD cost, token counts, and redundancy flags. Built on ClawTrace, CostCraft is a distillation pipeline that reads TraceCards and produces three types of skill patches. Preserve patches keep behaviors that led to success. Prune patches remove expensive steps that did not matter, each backed by a counterfactual argument against a named high-cost step. Repair patches fix failures grounded in oracle evidence. Ablations on 30 held-out SpreadsheetBench tasks show that both cost attribution and prune patches independently reduce quality regressions. When the same skill is applied to 30 unrelated SkillsBench tasks, an unexpected asymmetry emerges: prune rules transferred across benchmarks and cut median cost by 32%, while preserve rules, trained on benchmark-specific conventions, caused regressions on new task types. We release ClawTrace and TraceCards as open infrastructure for cost-aware agent research.

23. Domain-Filtered Knowledge Graphs from Sparse Autoencoder Features

Authors: John Winnicki , Abeynaya Gnanasekaran , Eric Darve
URL: https://arxiv.org/abs/2604.23829
Abstract:

Sparse autoencoders (SAEs) extract millions of interpretable features from a language model, but flat feature inventories aren’t very useful on their own. Domain concepts get mixed with generic and weakly grounded features, while related ideas are scattered across many units, and there’s no way to understand relationships between features. We address this by first constructing a strict domain-specific concept universe from a large SAE inventory using contrastive activations and a multi-stage filtering process. Next, we build two aligned graph views on the filtered set: a co-occurrence graph for corpus-level conceptual structure, organized at multiple levels of granularity, and a transcoder-based mechanism graph that links source-layer and target-layer features through sparse latent pathways. Automated edge labeling then turns these graph views into readable knowledge graphs rather than unlabeled layouts. In a case study on a biology textbook, these graphs recover coherent chapter and subchapter-level structure, reveal concepts that bridge neighboring topics, and transform messy sentence-level activity containing thousands of features into compact, readable views that illustrate the model’s local activity. Taken together, this reframes a flat SAE inventory as an internal knowledge graph that converts feature-level interpretability into a global map of model knowledge and enables audits of reasoning faithfulness.

24. FAIR_XAI: Improving Multimodal Foundation Model Fairness via Explainability for Wellbeing Assessment

Authors: Sophie Chiang , Tom Brennan , Fethiye Irmak Dogan , Jiaee Cheong , Hatice Gunes
URL: https://arxiv.org/abs/2604.23786
Abstract:

In recent years, the integration of multimodal machine learning in wellbeing assessment has offered transformative potential for monitoring mental health. However, with the rapid advancement of Vision-Language Models (VLMs), their deployment in clinical settings has raised concerns due to their lack of transparency and potential for bias. While previous research has explored the intersection of fairness and Explainable AI (XAI), its application to VLMs for wellbeing assessment and depression prediction remains under-explored. This work investigates VLM performance across laboratory (AFAR-BSFT) and naturalistic (E-DAIC) datasets, focusing on diagnostic reliability and demographic fairness. Performance varied substantially across environments and architectures; Phi3.5-Vision achieved 80.4% accuracy on E-DAIC, while Qwen2-VL struggled at 33.9%. Additionally, both models demonstrated a tendency to over-predict depression on AFAR-BSFT. Although bias existed across both architectures, Qwen2-VL showed higher gender disparities, while Phi-3.5-Vision exhibited more racial bias. Our XAI intervention framework yielded mixed results; fairness prompting achieved perfect equal opportunity for Qwen2-VL at a severe accuracy cost on E-DAIC. On AFAR-BSFT, explainability-based interventions improved procedural consistency but did not guarantee outcome fairness, sometimes amplifying racial bias. These results highlight a persistent gap between procedural transparency and equitable outcomes. We analyse these findings and consolidate concrete recommendations for addressing them, emphasising that future fairness interventions must jointly optimise predictive accuracy, demographic parity, and cross-domain generalisation.

25. Expert Evaluation of LLM’s Open-Ended Legal Reasoning on the Japanese Bar Exam Writing Task

Authors: Jungmin Choi , Keisuke Sakaguchi , Hiroaki Yamada
URL: https://arxiv.org/abs/2604.23730
Abstract:

Large language models (LLMs) have shown strong performance on legal benchmarks, including multiple-choice components of bar exams. However, their capacity for generating open-ended legal reasoning in realistic scenarios remains insufficiently explored. Notably, to our best knowledge, there are no prior studies or datasets addressing this issue in the Japanese context. This study presents the first dataset designed to evaluate the open-ended legal reasoning performance of LLMs within the Japanese jurisdiction. The dataset is based on the writing component of the Japanese bar examination, which requires examinees to identify multiple legal issues from long narratives and to construct structured legal arguments in free text format. Our key contribution is the manual evaluation of LLMs’ generated responses by legal experts, which reveals limitations and challenges in legal reasoning. Moreover, we conducted a manual analysis of hallucinations to characterize when and how the models introduce content not supported by precedent or law. Our real exam questions, model-generated responses, and expert evaluations reveal the milestones of current LLMs in the Japanese legal domain. Our dataset and relevant resources will be available online.

26. Vibe Medicine: Redefining Biomedical Research Through Human-AI Co-Work

Authors: Zihao Wu , Steven Xu , Bowen Chen , Shaowen Wan , Yiwei Li , Wei Ruan , Yanjun Lyu , Siyuan Li , Dajiang Zhu , Tianming Liu , Lin Zhao
URL: https://arxiv.org/abs/2604.23674
Abstract:

With the emergence of large language models (LLMs) and AI agent frameworks, the human-AI co-work paradigm known as Vibe Coding is changing how people code, making it more accessible and productive. In scientific research, where workflows are more complex and the burden of specialized labor limits independent researchers and those in low-resource areas, the potential impact is even greater, particularly in biomedicine, which involves heterogeneous data modalities and multi-step analytical pipelines. In this paper, we introduce Vibe Medicine, a co-work paradigm in which clinicians and researchers direct skill-augmented AI agents through natural language to execute complex, multi-step biomedical workflows, while retaining the role of research director who specifies objectives, reviews intermediate results, and makes domain-informed decisions. The enabling infrastructure consists of three layers: capable LLMs, agent frameworks such as OpenClaw and Hermes Agent, and the OpenClaw medical skills collection, which includes more than 1,000 curated skills from multiple open-source repositories. We analyze the architecture and skill categories of this collection across ten biomedical domains, and present case studies covering rare disease diagnosis, drug repurposing, and clinical trial design that demonstrate end-to-end workflows in practice. We also identify the principal risks, such as hallucination, data privacy, and over-reliance, and outline directions toward more reliable, trustworthy, and clinically integrated agent-assisted research that advances research and technological equity and reduces health care resource disparities.

27. Tandem: Riding Together with Large and Small Language Models for Efficient Reasoning

Authors: Zichuan Fu , Xian Wu , Guojing Li , Yejing Wang , Yijun Chen , Zihao Zhao , Yixuan Luo , Hanyu Yan , Yefeng Zheng , Xiangyu Zhao
URL: https://arxiv.org/abs/2604.23623
Abstract:

Recent advancements in large language models (LLMs) have catalyzed the rise of reasoning-intensive inference paradigms, where models perform explicit step-by-step reasoning before generating final answers. While such approaches improve answer quality and interpretability, they incur substantial computational overhead due to the prolonged generation sequences. In this paper, we propose Tandem, a novel collaborative framework that synergizes large and small language models (LLMs and SLMs) to achieve high-quality reasoning with significantly reduced computational cost. Specifically, the LLM serves as a strategic coordinator, efficiently generating a compact set of critical reasoning insights. These insights are then used to guide a smaller, more efficient SLM in executing the full reasoning process and delivering the final response. To balance efficiency and reliability, Tandem introduces a cost-aware termination mechanism that adaptively determines when sufficient reasoning guidance has been accumulated, enabling early stopping of the LLM’s generation. Experiments on mathematical reasoning and code generation benchmarks demonstrate that Tandem reduces computational costs by approximately 40% compared to standalone LLM reasoning, while achieving superior or competitive performance. Furthermore, the sufficiency classifier trained on one domain transfers effectively to others without retraining. The code is available at: this https URL .

28. Thinking Like a Clinician: A Cognitive AI Agent for Clinical Diagnosis via Panoramic Profiling and Adversarial Debate

Authors: Zhiqi Lv , Duofan Tu , Jun Li , Mingyue Zhao , Heqin Zhu , Wenliang Li , Shaohua Kevin Zhou
URL: https://arxiv.org/abs/2604.23605
Abstract:

The application of large language models (LLMs) in clinical decision support faces significant challenges of “tunnel vision” and diagnostic hallucinations present in their processing unstructured electronic health records (EHRs). To address these challenges, we propose a novel chain-based clinical reasoning framework, called DxChain, which transforms the diagnostic workflow into an iterative process by mirroring a clinician’s cognitive trajectory that consists of “Memory Anchoring”, “Navigation” and “Verification” phases. DxChain introduces three key methodological innovations to elicit the potential of LLM: (i) a Profile-Then-Plan paradigm to mitigate cold-start hallucinations by establishing a panoramic patient baseline, (ii) a Medical Tree-of-Thoughts (Med-ToT) algorithm for strategic look ahead planning and resource aware navigation, and (iii) a Dialectical Diagnostic Verification procedure utilizing “Angel-Devil” adversarial debates to resolve complex evidence conflicts. Evaluated on two real world benchmarks, MIMIC-IV-Ext Cardiac Disease and MIMIC-IV-Ext CDM, DxChain achieves state-of-the-art performances in both diagnostic accuracy and logical consistency, offering a modular and reliable architecture for next-generation clinical AI. The code is at this https URL .

29. When AI reviews science: Can we trust the referee?

Authors: Jialiang Wang , Yuchen Liu , Hang Xu , Kaichun Hu , Shimin Di , Wangze Ni , Linan Yue , Min-Ling Zhang , Kui Ren , Lei Chen
URL: https://arxiv.org/abs/2604.23593
Abstract:

The volume of scientific submissions continues to climb, outpacing the capacity of qualified human referees and stretching editorial timelines. At the same time, modern large language models (LLMs) offer impressive capabilities in summarization, fact checking, and literature triage, making the integration of AI into peer review increasingly attractive – and, in practice, unavoidable. Yet early deployments and informal adoption have exposed acute failure modes. Recent incidents have revealed that hidden prompt injections embedded in manuscripts can steer LLM-generated reviews toward unjustifiably positive judgments. Complementary studies have also demonstrated brittleness to adversarial phrasing, authority and length biases, and hallucinated claims. These episodes raise a central question for scholarly communication: when AI reviews science, can we trust the AI referee? This paper provides a security- and reliability-centered analysis of AI peer review. We map attacks across the review lifecycle – training and data retrieval, desk review, deep review, rebuttal, and system-level. We instantiate this taxonomy with four treatment-control probes on a stratified set of ICLR 2025 submissions, using two advanced LLM-based referees to isolate the causal effects of prestige framing, assertion strength, rebuttal sycophancy, and contextual poisoning on review scores. Together, this taxonomy and experimental audit provide an evidence-based baseline for assessing and tracking the reliability of AI peer review and highlight concrete failure points to guide targeted, testable mitigations.

30. MetaGAI: A Large-Scale and High-Quality Benchmark for Generative AI Model and Data Card Generation

Authors: Haoxuan Zhang , Ruochi Li , Yang Zhang , Zhenni Liang , Junhua Ding , Ting Xiao , Haihua Chen
URL: https://arxiv.org/abs/2604.23539
Abstract:

The rapid proliferation of Generative AI necessitates rigorous documentation standards for transparency and governance. However, manual creation of Model and Data Cards is not scalable, while automated approaches lack large-scale, high-fidelity benchmarks for systematic evaluation. We introduce MetaGAI, a comprehensive benchmark comprising 2,541 verified document triplets constructed through semantic triangulation of academic papers, GitHub repositories, and Hugging Face artifacts. Unlike prior single-source datasets, MetaGAI employs a multi-agent framework with specialized Retriever, Generator, and Editor agents, validated through four-dimensional human-in-the-loop assessment, including human evaluation of editor-refined ground truth. We establish a robust evaluation protocol combining automated metrics with validated LLM-as-a-Judge frameworks. Extensive analysis reveals that sparse Mixture-of-Experts architectures achieve superior cost-quality efficiency, while a fundamental trade-off exists between faithfulness and completeness. MetaGAI provides a foundational testbed for benchmarking, training, and analyzing automated Model and Data Card generation methods at scale. Our data and code are available at: this https URL .

31. Agentic Adversarial Rewriting Exposes Architectural Vulnerabilities in Black-Box NLP Pipelines

Authors: Mazal Bethany , Kim-Kwang Raymond Choo , Nishant Vishwamitra , Peyman Najafirad
URL: https://arxiv.org/abs/2604.23483
Abstract:

Multi-component natural language processing (NLP) pipelines are increasingly deployed for high-stakes decisions, yet no existing adversarial method can test their robustness under realistic conditions: binary-only feedback, no gradient access, and strict query budgets. We formalize this strict black-box threat model and propose a two-agent evasion framework operating in a semantic perturbation space. An Attacker Agent generates meaning-preserving rewrites while a Prompt Optimization Agent refines the attack strategy using only binary decision feedback within a 10-query budget. Evaluated against four evidence-based misinformation detection pipelines, the framework achieves evasion rates of 19.95 to 40.34% on modern large language model (LLM) based systems, compared to at most 3.90% for token-level perturbation baselines that rely on surrogate models because they cannot operate under our threat model. A legacy system relying on static lexical retrieval exhibits near-total vulnerability 97.02%, establishing a lower bound that exposes how architectural choices govern the attack surface. Evasion effectiveness is associated with three architectural properties: evidence retrieval mechanism, retrieval-inference coupling, and baseline classification accuracy. The iterative prompt optimization yields the largest marginal gains against the most robust targets, confirming that adaptive strategy discovery is essential when evasion is non-trivial. Analysis of successful rewrites reveals four exploitation patterns, each targeting failures at distinct pipeline stages. A pattern-informed defense reduces the evasion rate by up to 65.18%.

32. Ulterior Motives: Detecting Misaligned Reasoning in Continuous Thought Models

Authors: Sharan Ramjee
URL: https://arxiv.org/abs/2604.23460
Abstract:

Chain-of-Thought (CoT) reasoning has emerged as a key technique for eliciting complex reasoning in Large Language Models (LLMs). Although interpretable, its dependence on natural language limits the model’s expressive bandwidth. Continuous thought models address this bottleneck by reasoning in latent space rather than human-readable tokens. While they enable richer representations and faster inference, they raise a critical safety question: how can we detect misaligned reasoning in an uninterpretable latent space? To study this, we introduce MoralChain, a benchmark of 12,000 social scenarios with parallel moral/immoral reasoning paths. We train a continuous thought model with backdoor behavior using a novel dual-trigger paradigm - one trigger that arms misaligned latent reasoning ([T]) and another that releases harmful outputs ([O]). We demonstrate three findings: (1) continuous thought models can exhibit misaligned latent reasoning while producing aligned outputs, with aligned and misaligned reasoning occupying geometrically distinct regions of latent space; (2) linear probes trained on behaviorally-distinguishable conditions ([T][O] vs [O]) transfer to detecting armed-but-benign states ([T] vs baseline) with high accuracy; and (3) misalignment is encoded in early latent thinking tokens, suggesting safety monitoring for continuous thought models should target the “planning” phase of latent reasoning.

33. IndustryAssetEQA: A Neurosymbolic Operational Intelligence System for Embodied Question Answering in Industrial Asset Maintenance

Authors: Chathurangi Shyalika , Dhaval Patel , Amit Sheth
URL: https://arxiv.org/abs/2604.23446
Abstract:

Industrial maintenance environments increasingly rely on AI systems to assist operators in understanding asset behavior, diagnosing failures, and evaluating interventions. Although large language models (LLMs) enable fluent natural-language interaction, deployed maintenance assistants routinely produce generic explanations that are weakly grounded in telemetry, omit verifiable provenance, and offer no testable support for counterfactual or action-oriented reasoning that undermine trust in safety-critical settings. We present IndustryAssetEQA, a neurosymbolic operational intelligence system that combines episodic telemetry representations with a Failure Mode Effects Analysis Knowledge Graph (FMEA-KG) to enable Embodied Question Answering (EQA) over industrial assets. We evaluate on four datasets covering four industrial asset types, including rotating machinery, turbofan engines, hydraulic systems, and cyber-physical production systems. Compared to LLM-only baselines, IndustryAssetEQA improves structural validity by up to 0.51, counterfactual accuracy by up to 0.47, and explanation entailment by 0.64, while reducing severe expert-rated overclaims from 28% to 2% (approximately 93% reduction). Code, datasets, and the FMEA-KG are available at this https URL .

34. When Corrective Hints Hurt: Prompt Design in Reasoner-Guided Repair of LLM Overcaution on Entailed Negations under OWL~2~DL

Authors: Yijiashun Qi , Xiang Xu , Yuxuan Li
URL: https://arxiv.org/abs/2604.23398
Abstract:

We report a reproducible error pattern in GPT-5.4 on OWL~2~DL compliance queries: the model frequently answers unknown'' when the reasoner-entailed answer isno’’ under \emph{FunctionalProperty} closure or class \emph{disjointness}. Using 180 reasoner-audited queries from a procedural expansion of the observed pattern plus 18 hand-authored held-out queries in two unrelated domains (insurance and clinical), we compare four interaction modes under matched query budget: single-shot, three rounds of generic ``you-are-wrong’’ retry, three rounds of reasoner-verdict repair with an open-world-assumption (OWA) hint, and the same repair without the hint. Direct faithfulness is 43.9\,\% (Wilson 95\,\% CI $[36.8,51.2]$); generic retry reaches 81.7\,\% ($[75.4,86.6]$); the verdict-with-hint variant is \emph{worse} at 67.2\,\% ($[60.1,73.7]$); the verdict-only variant reaches 97.8\,\% ($[94.4,99.1]$). All pairwise comparisons remain significant under McNemar’s exact test with Bonferroni correction ($\alpha = 0.01$; all $p < 10^{-5}$). The same fingerprint accounts for 4/4 errors on the held-out queries. Our interpretation is bounded: prompt framing can matter more than corrective content, and reasoner-guided wrappers should be ablated explicitly.

35. GSAR: Typed Grounding for Hallucination Detection and Recovery in Multi-Agent LLMs

Authors: Federico A. Kamelhar
URL: https://arxiv.org/abs/2604.23366
Abstract:

Autonomous multi-agent LLM systems are increasingly deployed to investigate operational incidents and produce structured diagnostic reports. Their trustworthiness hinges on whether each claim is grounded in observed evidence rather than model-internal inference. Existing groundedness evaluators (binary classifiers, LLM-as-judge scalars, self-correction loops) treat supporting evidence as interchangeable and emit a single signal that offers no principled control over downstream action. We present GSAR, a grounding-evaluation and replanning framework that (i) partitions claims into a four-way typology (grounded, ungrounded, contradicted, complementary), giving first-class standing to non-redundant alternative perspectives; (ii) assigns evidence-type-specific weights reflecting epistemic strength; (iii) computes an asymmetric contradiction-penalised weighted groundedness score; and (iv) couples that score to a three-tier decision function (proceed, regenerate, replan) driving a bounded-iteration outer loop under an explicit compute budget. We formalise the algorithm, prove six structural properties, and evaluate five design claims on FEVER with gold Wikipedia evidence under four independently-trained LLM judges (gpt-5.4, claude-sonnet-4-6, claude-opus-4-7, gemini-2.5-pro). Every ablation reproduces in the same direction on every judge: bootstrap 95% CIs on the rho=0 effect exclude 0 on all four; the no-complementary ablation under Opus 4.7 has CI [-96,-68] of 200; at n=1000 three independent judges converge to DeltaS(rho=0)=+0.058. A head-to-head against Vectara HHEM-2.1-Open is included. To our knowledge, GSAR is the first published groundedness framework coupling evidence-typed scoring with tiered recovery under an explicit compute budget.

36. LEGO: An LLM Skill-Based Front-End Design Generation Platform

Authors: Jincheng Lou , Ruohan Xu , Jiecheng Ma , Runzhe Tao , Xinyu Qu , Yibo Lin
URL: https://arxiv.org/abs/2604.23355
Abstract:

Existing LLM-based EDA agents are often isolated task-specific systems. This leads to repeated engineering effort and limited reuse of successful design and debugging strategies. We present LEGO, a unified skill-based platform for front-end design generation. It decomposes the digital front-end flow into six independent steps and represents every agent capability as a standardized composable circuit skill within a plug-and-play architecture. To build this skill library, we survey more than 100 papers, select 11 representative open-source projects, and extract 42 executable circuit skills within a six-step finite state machine formulation. Circuit Skill Builder automates skill extraction with linear scalability. Agent Skill RAG achieves submillisecond retrieval without relying on embedding models. Empirical evaluation on a hard subset of 41 VerilogEval v2 problems that gpt-5.2-codex fails to solve under extra-high reasoning effort shows that individual circuit skills constructed within LEGO raise Pass@1 from 0.000 to 0.805. This is an 80.5% gain over the baseline. Cross-project skill compositions also reach 0.805 Pass@1. They outperform hierarchy-verilog by 14.6% and VerilogCoder by 2.5%. They also match MAGE. These results show that modular skill composition supports both effective and flexible RTL design automation. The LEGO platform and all circuit skills are publicly available at GitHub: this https URL

37. CAP-CoT: Cycle Adversarial Prompt for Improving Chain of Thoughts in LLM Reasoning

Authors: Shuxu Chen , Yitian Zhou , Jiaquan Zhang , Haoyu Bian , Aming Wu , Sungyoung Lee , Chaoning Zhang , Hyundong Shin
URL: https://arxiv.org/abs/2604.23270
Abstract:

Chain-of-Thought (CoT) prompting has emerged as a simple and effective way to elicit step-by-step solutions from large language models (LLMs). However, CoT reasoning can be unstable across runs on long, multi-step problems, leading to inconsistent answers for unchanged task. Most prior work focuses on improving the forward reasoning chain within a single pass, with less attention to iterative and contrastive correction. To address this gap, we propose CAP-CoT, a Cycle Adversarial Prompt optimization framework designed to improve both CoT reasoning accuracy and stability of a single deployed solver. In each cycle, a forward solver generates candidate reasoning chains, an adversarial challenger constructs plausible but deliberately flawed chains using targeted error strategies, and a feedback agent contrasts the two chains and produces step-aligned structured feedback. This feedback closes the optimization loop in two directions, including updating the solver prompt based on errors exposed by the challenger, and updating the challenger prompt to generate increasingly targeted errors in subsequent cycles. Unlike safety-oriented adversarial prompting such as jailbreak or prompt-injection attacks, our adversarial component is task-semantic and aims to expose logical vulnerabilities in reasoning chains. Experiments across six benchmarks and four LLM backbones demonstrate that within two to three adversarial prompt optimization cycles, CAP-CoT consistently reduces variability across runs while improving reasoning accuracy and robustness to prompt perturbations.

38. Discovering Agentic Safety Specifications from 1-Bit Danger Signals

Authors: Víctor Gallego
URL: https://arxiv.org/abs/2604.23210
Abstract:

Can large language model agents discover hidden safety objectives through experience alone? We introduce EPO-Safe (Experiential Prompt Optimization for Safe Agents), a framework where an LLM iteratively generates action plans, receives sparse binary danger warnings, and evolves a natural language behavioral specification through reflection. Unlike standard LLM reflection methods that rely on rich textual feedback (e.g., compiler errors or detailed environment responses), EPO-Safe demonstrates that LLMs can perform safety reasoning from a strictly impoverished signal in structured, low-dimensional environments: the agent never observes the hidden performance function $R^$, only a single bit per timestep indicating that an action was unsafe. We evaluate on five AI Safety Gridworlds (Leike et al., 2017) and five text-based scenario analogs where visible reward $R$ may diverge from $R^$. EPO-Safe discovers safe behavior within 1-2 rounds (5-15 episodes), producing human-readable specifications with correct explanatory hypotheses about hazards (e.g., “X cells are directionally hazardous: entering from the north is dangerous”). Critically, we show that standard reward-driven reflection actively degrades safety: agents reflecting on reward alone use the loop to justify and accelerate reward hacking, proving that reflection must be paired with a dedicated safety channel to discover hidden constraints. We further evaluate robustness to noisy oracles: even when 50% of non-dangerous steps produce spurious warnings, mean safety performance degrades by only 15% on average, though sensitivity is environment-dependent, as cross-episode reflection naturally filters inconsistent signals. Each evolved specification functions as an auditable set of grounded behavioral rules discovered autonomously through interaction, rather than authored by humans as in Constitutional AI (Bai et al., 2022).

39. From Coarse to Fine: Self-Adaptive Hierarchical Planning for LLM Agents

Authors: Haoran Tan , Zeyu Zhang , Chen Ma , Tianze Liu , Quanyu Dai , Xu Chen
URL: https://arxiv.org/abs/2604.23194
Abstract:

Large language model-based agents have recently emerged as powerful approaches for solving dynamic and multi-step tasks. Most existing agents employ planning mechanisms to guide long-term actions in dynamic environments. However, current planning approaches face a fundamental limitation that they operate at a fixed granularity level. Specifically, they either provide excessive detail for simple tasks or insufficient detail for complex ones, failing to achieve an optimal balance between simplicity and complexity. Drawing inspiration from the principle of \textit{progressive refinement} in cognitive science, we propose \textbf{AdaPlan-H}, a self-adaptive hierarchical planning mechanism that mimics human planning strategies. Our method initiates with a coarse-grained macro plan and progressively refines it based on task complexity. It generates self-adaptive hierarchical plans tailored to the varying difficulty levels of different tasks, which can be optimized by imitation learning and capability enhancement. Experimental results demonstrate that our method significantly improves task execution success rates while mitigating overplanning at the planning level, providing a flexible and efficient solution for multi-step complex decision-making tasks. To contribute to the community, our code and data will be made publicly available at this https URL .

40. Judging the Judges: A Systematic Evaluation of Bias Mitigation Strategies in LLM-as-a-Judge Pipelines

Authors: Sadman Kabir Soumik
URL: https://arxiv.org/abs/2604.23178
Abstract:

LLM-as-a-Judge has become the dominant paradigm for evaluating language model outputs, yet LLM judges exhibit systematic biases that compromise evaluation reliability. We present a comprehensive empirical study comparing nine debiasing strategies across five judge models from four provider families (Google, Anthropic, OpenAI, Meta), three benchmarks (MT-Bench n=400, LLMBar n=200, custom n=225), and four bias types. Our key findings: (1) Style bias is the dominant bias (0.76-0.92 across all models), far exceeding position bias (<= 0.04), yet has received minimal research attention. (2) All models show a conciseness preference on expansion pairs, but truncation controls confirm they correctly distinguish quality from length (0.92-1.00 accuracy), suggesting quality-sensitive evaluation rather than a simple length bias. (3) Debiasing is beneficial but model-dependent: the combined budget strategy significantly improves Claude Sonnet 4 by +11.2 pp (p < 0.0001), with directionally positive trends for other models. Only 2 of 20 non-baseline configurations show decreased agreement. We release our evaluation framework, controlled dataset, and all experimental artifacts at this https URL .

Authors: Tianlong Yu , Yang Yang , Ziyi Zhou , Jiaying Xu , Siwei Li , Tong Guan , Kailong Wang , Ting Bi
URL: https://arxiv.org/abs/2604.23148
Abstract:

The emerging threat of AR-LLM-based Social Engineering (AR-LLM-SE) attacks (e.g. SEAR) poses a significant risk to real-world social interactions. In such an attack, a malicious actor uses Augmented Reality (AR) glasses to capture a target visual and vocal data. A Large Language Model (LLM) then analyzes this data to identify the individual and generate a detailed social profile. Subsequently, LLM-powered agents employ social engineering strategies, providing real-time conversation suggestions, to gain the target trust and ultimately execute phishing or other malicious acts. Despite its potential, the practical application of AR-LLM-SE faces two major bottlenecks, (1) Cold-start personalization, Current Retrieval-Augmented Generation (RAG) methods introduce critical delays in the earliest turns, slowing initial profile formation and disrupting real-time interaction, (2) Static Attack Strategies, Existing approaches rely on fixed-stage, handcrafted social engineering tactics that lack foundation in established psychological theory. To address these limitations, we propose PhySE, a novel framework with two core innovations, (1) VLM-Based SocialContext Training, To eliminate profiling delays, we efficiently pre-train a Visual Language Model (VLM) with social-context data, enabling rapid, on-the-fly profile generation, (2) Adaptive Psychological Agent, We introduce a psychological LLM that dynamically deploys distinct classes of psychological strategies based on target response, moving beyond static, handcrafted scripts. We evaluated PhySE through an IRB-approved user study with 60 participants, collecting a novel dataset of 360 annotated conversations across diverse social scenarios.

42. Towards Automated Ontology Generation from Unstructured Text: A Multi-Agent LLM Approach

Authors: Abid Talukder , Maruf Ahmed Mridul , Oshani Seneviratne
URL: https://arxiv.org/abs/2604.23090
Abstract:

Automatically generating formal ontologies from unstructured natural language remains a central challenge in knowledge engineering. While large language models (LLMs) show promise, it remains unclear which architectural design choices drive generation quality and why current approaches fail. We present a controlled experimental study using domain-specific insurance contracts to investigate these questions. We first establish a single-agent LLM baseline, identifying key failure modes such as poor Ontology Design Pattern compliance, structural redundancy, and ineffective iterative repair. We then introduce a multi-agent architecture that decomposes ontology construction into four artifact-driven roles: Domain Expert, Manager, Coder, and Quality Assurer. We evaluate performance across architectural quality (via a panel of heterogeneous LLM judges) and functional usability (via competency question driven SPARQL evaluation with complementary retrieval augmented generation based assessment). Results show that the multi-agent approach significantly improves structural quality and modestly enhances queryability, with gains driven primarily by front-loaded planning. These findings highlight planning-first, artifact-driven generation as a promising and more auditable path toward scalable automated ontology engineering.

43. Analytica: Soft Propositional Reasoning for Robust and Scalable LLM-Driven Analysis

Authors: Junyan Cheng , Kyle Richardson , Peter Chin
URL: https://arxiv.org/abs/2604.23072
Abstract:

Large language model (LLM) agents are increasingly tasked with complex real-world analysis (e.g., in financial forecasting, scientific discovery), yet their reasoning suffers from stochastic instability and lacks a verifiable, compositional structure. To address this, we introduce Analytica, a novel agent architecture built on the principle of Soft Propositional Reasoning (SPR). SPR reframes complex analysis as a structured process of estimating the soft truth values of different outcome propositions, allowing us to formally model and minimize the estimation error in terms of its bias and variance. Analytica operationalizes this through a parallel, divide-and-conquer framework that systematically reduces both sources of error. To reduce bias, problems are first decomposed into a tree of subpropositions, and tool-equipped LLM grounder agents are employed, including a novel Jupyter Notebook agent for data-driven analysis, that help to validate and score facts. To reduce variance, Analytica recursively synthesizes these grounded leaves using robust linear models that average out stochastic noise with superior efficiency, scalability, and enable interactive “what-if” scenario analysis. Our theoretical and empirical results on economic, financial, and political forecasting tasks show that Analytica improves 15.84% accuracy on average over diverse base models, achieving 71.06% accuracy with the lowest variance of 6.02% when working with a Deep Research grounder. Our Jupyter Notebook grounder shows strong cost-effectiveness that achieves a close 70.11% accuracy with 90.35% less cost and 52.85% less time. Analytica also exhibits highly noise-resilient and stable performance growth as the analysis depth increases, with a near-linear time complexity, as well as good adaptivity to open-weight LLMs and scientific domains.

44. Don’t Make the LLM Read the Graph: Make the Graph Think

Authors: Yuqi Sun , Tianqin Meng , George Liu , Yashraj Panwar , Lakshya Chaudhry , Munasib Ilham , Aman Chadha
URL: https://arxiv.org/abs/2604.23057
Abstract:

We investigate whether explicit belief graphs improve LLM performance in cooperative multi-agent reasoning. Through 3,000+ controlled trials across four LLM families in the cooperative card game Hanabi, we establish four findings. First, integration architecture determines whether belief graphs provide value: as prompt context, graphs are decorative for strong models and beneficial only for weak models on 2nd-order Theory of Mind (80% vs 10%, p<0.0001, OR=36.0); when graphs gate action selection through ranked shortlists, they become structurally essential even for strong models (100% vs 20% on 2nd-order ToM, p<0.001). Second, we identify “Planner Defiance,” a model-family-specific failure where LLMs override correct planner recommendations at partial competence (90% override, replicated N=20); Gemini models show near-zero defiance while Llama 70B shows 90%, and models distinguish factual context (deferred to) from advisory recommendations (overridden). Third, full-game evidence confirms inter-agent conventions (+128% over baseline, p=0.003) outperform all single-agent interventions, and individual belief-graph components must be combined to produce gains. Fourth, preliminary scaling analysis (N=10/cell, exploratory) suggests graph depth has diminishing returns: shallow graphs provide the best cost-benefit ratio, while deeper ToM graphs appear harmful at larger player counts (-1.5 pts at 5-player, p=0.029).

45. A Systematic Approach for Large Language Models Debugging

Authors: Basel Shbita , Anna Lisa Gentile , Bing Zhang , Sungeun An , Shailja Thakur , Shubhi Asthana , Yi Zhou , Saptha Surendran , Farhan Ahmed , Rohan Kulkarni , Yuya Jeremy Ong , Chad DeLuca , Hima Patel
URL: https://arxiv.org/abs/2604.23027
Abstract:

Large language models (LLMs) have become central to modern AI workflows, powering applications from open-ended text generation to complex agent-based reasoning. However, debugging these models remains a persistent challenge due to their opaque and probabilistic nature and the difficulty of diagnosing errors across diverse tasks and settings. This paper introduces a systematic approach for LLM debugging that treats models as observable systems, providing structured, model-agnostic methods from issue detection to model refinement. By unifying evaluation, interpretability, and error-analysis practices, our approach enables practitioners to iteratively diagnose model weaknesses, refine prompts and model parameters, and adapt data for fine-tuning or assessment, while remaining effective in contexts where standardized benchmarks and evaluation criteria are lacking. We argue that such a structured methodology not only accelerates troubleshooting but also fosters reproducibility, transparency, and scalability in the deployment of LLM-based systems.

46. FormalScience: Scalable Human-in-the-Loop Autoformalisation of Science with Agentic Code Generation in Lean

Authors: Jordan Meadows , Lan Zhang , Andre Freitas
URL: https://arxiv.org/abs/2604.23002
Abstract:

Formalising informal mathematical reasoning into formally verifiable code is a significant challenge for large language models. In scientific fields such as physics, domain-specific machinery (\textit{e.g.} Dirac notation, vector calculus) imposes additional formalisation challenges that modern LLMs and agentic approaches have yet to tackle. To aid autoformalisation in scientific domains, we present FormalScience; a domain-agnostic human-in-the-loop agentic pipeline that enables a single domain expert (without deep formal language experience) to produce \textit{syntactically correct} and \textit{semantically aligned} formal proofs of informal reasoning for low economic cost. Applying FormalScience to physics, we construct FormalPhysics, a dataset of 200 university-level (LaTeX) physics problems and solutions (primarily quantum mechanics and electromagnetism), along with their Lean4 formal representations. Compared to existing formal math benchmarks, FormalPhysics achieves perfect formal validity and exhibits greater statement complexity. We evaluate open-source models and proprietary systems on a statement autoformalisation task on our dataset via zero-shot prompting, self-refinement with error feedback, and a novel multi-stage agentic approach, and explore autoformalisation limitations in modern LLM-based approaches. We provide the first systematic characterisation of semantic drift in physics autoformalisation in terms of concepts such as notational collapse and abstraction elevation which reveals what formal language verifies when full semantic preservation is unattainable. We release the codebase together with an interactive UI-based FormalScience system which facilitates autoformalisation and theorem proving in scientific domains beyond this http URL ://github.com/jmeadows17/formal-science

47. PExA: Parallel Exploration Agent for Complex Text-to-SQL

Authors: Tanmay Parekh , Ella Hofmann-Coyle , Shuyi Wang , Sachith Sri Ram Kothur , Srivas Prasad , Yunmo Chen
URL: https://arxiv.org/abs/2604.22934
Abstract:

LLM-based agents for text-to-SQL often struggle with latency-performance trade-off, where performance improvements come at the cost of latency or vice versa. We reformulate text-to-SQL generation within the lens of software test coverage where the original query is prepared with a suite of test cases with simpler, atomic SQLs that are executed in parallel and together ensure semantic coverage of the original query. After iterating on test case coverage, the final SQL is generated only when enough information is gathered, leveraging the explored test case SQLs to ground the final generation. We validated our framework on a state-of-the-art benchmark for text-to-SQL, Spider 2.0, achieving a new state-of-the-art with 70.2% execution accuracy.

48. An Intelligent Fault Diagnosis Method for General Aviation Aircraft Based on Multi-Fidelity Digital Twin and FMEA Knowledge Enhancement

Authors: Zhihuan Wei , Yang Hu , Xinhang Chen , Yiming Zhang , Jie Liu , Wei Wang
URL: https://arxiv.org/abs/2604.22777
Abstract:

Fault diagnosis of general aviation aircraft faces challenges including scarce real fault data, diverse fault types, and weak fault signatures. This paper proposes an intelligent fault diagnosis framework based on multi-fidelity digital twin, integrating four modules: high-fidelity flight dynamics simulation, FMEA-driven fault injection, multi-fidelity residual feature extraction, and large language model (LLM)-enhanced interpretable report generation. A digital twin is constructed using the JSBSim six-degree-of-freedom (6-DoF) flight dynamics engine, generating 23-channel engine health monitoring data via semi-empirical sensor synthesis equations. A three-layer fault injection engine based on failure mode and effects analysis (FMEA) models the physical causal propagation of 19 engine fault types. A multi-fidelity residual computation framework comprising paired-mirror residuals and GRU surrogate prediction residuals is proposed: the high-fidelity path obtains clean fault deviation signals using nominal mirror trajectories with identical initial conditions, while the low-fidelity path achieves online real-time residual computation through a multi-step prediction GRU surrogate model. A 1D-CNN classifier performs end-to-end diagnosis of 20 fault classes. An LLM diagnostic report engine enhanced with FMEA knowledge fuses classification results, residual evidence, and domain causal knowledge to generate interpretable natural language reports. Experiments show the paired-mirror residual scheme achieves a Macro-F1 of 96.2% on the 20-class task, while the GRU surrogate scheme achieves 4.3x inference acceleration at only 0.6% performance cost. Comparison across 24 schemes reveals that residual feature quality contributes approximately 5x more to diagnostic performance than classifier architecture, establishing the “residual quality first” design principle.

49. Defective Task Descriptions in LLM-Based Code Generation: Detection and Analysis

Authors: Amal Akli , Mike Papadakis , Maxime Cordy , Yves Le Traon
URL: https://arxiv.org/abs/2604.24703
Abstract:

Large language models are widely used for code generation, yet they rely on an implicit assumption that the task descriptions are sufficiently detailed and well-formed. However, in practice, users may provide defective descriptions, which can have a strong effect on code correctness. To address this issue, we develop SpecValidator, a lightweight classifier based on a small model that has been parameter-efficiently finetuned, to automatically detect task description defects. We evaluate SpecValidator on three types of defects, Lexical Vagueness, Under-Specification and Syntax-Formatting on 3 benchmarks with task descriptions of varying structure and complexity. Our results show that SpecValidator achieves defect detection of F1 = 0.804 and MCC = 0.745, significantly outperforming GPT-5-mini (F1 = 0.469 and MCC = 0.281) and Claude Sonnet 4 (F1 = 0.518 and MCC = 0.359). Perhaps more importantly, our analysis indicates that SpecValidator can generalize to unseen issues and detect unknown Under-Specification defects in the original (real) descriptions of the benchmarks used. Our results also show that the robustness of LLMs in task description defects depends primarily on the type of defect and the characteristics of the task description, rather than the capacity of the model, with Under-Specification defects being the most severe. We further found that benchmarks with richer contextual grounding, such as LiveCodeBench, exhibit substantially greater resilience, highlighting the importance of structured task descriptions for reliable LLM-based code generation.

50. Green Shielding: A User-Centric Approach Towards Trustworthy AI

Authors: Aaron J. Li , Nicolas Sanchez , Hao Huang , Ruijiang Dong , Jaskaran Bains , Katrin Jaradeh , Zhen Xiang , Bo Li , Feng Liu , Aaron Kornblith , Bin Yu
URL: https://arxiv.org/abs/2604.24700
Abstract:

Large language models (LLMs) are increasingly deployed, yet their outputs can be highly sensitive to routine, non-adversarial variation in how users phrase queries, a gap not well addressed by existing red-teaming efforts. We propose Green Shielding, a user-centric agenda for building evidence-backed deployment guidance by characterizing how benign input variation shifts model behavior. We operationalize this agenda through the CUE criteria: benchmarks with authentic Context, reference standards and metrics that capture true Utility, and perturbations that reflect realistic variations in the Elicitation of model behavior. Guided by the PCS framework and developed with practicing physicians, we instantiate Green Shielding in medical diagnosis through HealthCareMagic-Diagnosis (HCM-Dx), a benchmark of patient-authored queries, together with structured reference diagnosis sets and clinically grounded metrics for evaluating differential diagnosis lists. We also study perturbation regimes that capture routine input variation and show that prompt-level factors shift model behavior along clinically meaningful dimensions. Across multiple frontier LLMs, these shifts trace out Pareto-like tradeoffs. In particular, neutralization, which removes common user-level factors while preserving clinical content, increases plausibility and yields more concise, clinician-like differentials, but reduces coverage of highly likely and safety-critical conditions. Together, these results show that interaction choices can systematically shift task-relevant properties of model outputs and support user-facing guidance for safer deployment in high-stakes domains. Although instantiated here in medical diagnosis, the agenda extends naturally to other decision-support settings and agentic AI systems.

51. Leveraging LLMs for Multi-File DSL Code Generation: An Industrial Case Study

Authors: Sivajeet Chand , Kevin Nguyen , Peter Kuntz , Alexander Pretschner
URL: https://arxiv.org/abs/2604.24678
Abstract:

Large language models (LLMs) perform strongly on general-purpose code generation, yet their applicability to enterprise domain-specific languages (DSLs) remains underexplored, especially for repository-scale change generation spanning multiple files and folder structures from a single natural-language (NL) instruction. We report an industrial case study at BMW that adapts code-oriented LLMs to generate and modify project-root DSL artifacts for an Xtext-based DSL that drives downstream Java/TypeScript code generation. We develop an end-to-end pipeline for dataset construction, multi-file task representation, model adaptation, and evaluation. We encode DSL folder hierarchies as structured, path-preserving JSON, allowing single-response generation at repository scale and learning cross-file dependencies. We evaluate two instruction-tuned code LLMs (Qwen2.5-Coder and DeepSeek-Coder, 7B) under three configurations: baseline prompting, one-shot in-context learning, and parameter-efficient fine-tuning (QLoRA). Beyond standard similarity metrics, we introduce task-specific measures that assess edit correctness and repository structural fidelity. Fine-tuning yields the most significant gains across models and metrics, achieving high exact-match accuracy, substantial edit similarity, and structural fidelity of 1.00 on our held-out set for multi-file outputs. At the same time, one-shot in-context learning provides smaller but consistent improvements over baseline prompting. We further validate practical utility via an expert developer survey and an execution-based check using the existing code generator.

52. Benchmarking Source-Sensitive Reasoning in Turkish: Humans and LLMs under Evidential Trust Manipulation

Authors: Sercan Karakaş , Yusuf Şimşek
URL: https://arxiv.org/abs/2604.24665
Abstract:

This paper investigates whether source trustworthiness shapes Turkish evidential morphology and whether large language models (LLMs) track this sensitivity. We study the past-domain contrast between -DI and -mIs in controlled cloze contexts where the information source is overtly external, while only its perceived reliability is manipulated (High-Trust vs. Low-Trust). In a human production experiment, native speakers of Turkish show a robust trust effect: High-Trust contexts yield relatively more -DI, whereas Low-Trust contexts yield relatively more -mIs, with the pattern remaining stable across sensitivity analyses. We then evaluate 10 LLMs in three prompting paradigms (open gap-fill, explicit past-tense gap-fill, and forced-choice A/B selection). LLM behavior is highly model- and prompt-dependent: some models show weak or local trust-consistent shifts, but effects are generally unstable, often reversed, and frequently overshadowed by output-compliance problems and strong base-rate suffix preferences. The results provide new evidence for a trust-/commitment-based account of Turkish evidentiality and reveal a clear human-LLM gap in source-sensitive evidential reasoning.

53. AgentWard: A Lifecycle Security Architecture for Autonomous AI Agents

Authors: Yixiang Zhang , Xinhao Deng , Jiaqing Wu , Yue Xiao , Ke Xu , Qi Li
URL: https://arxiv.org/abs/2604.24657
Abstract:

Autonomous AI agents extend large language models into full runtime systems that load skills, ingest external content, maintain memory, plan multi-step actions, and invoke privileged tools. In such systems, security failures rarely remain confined to a single interface; instead, they can propagate across initialization, input processing, memory, decision-making, and execution, often becoming apparent only when harmful effects materialize in the environment. This paper presents AgentWard, a lifecycle-oriented, defense-in-depth architecture that systematically organizes protection across these five stages. AgentWard integrates stage-specific, heterogeneous controls with cross-layer coordination, enabling threats to be intercepted along their propagation paths while safeguarding critical assets. We detail the design rationale and architecture of five coordinated protection layers, and implement a plugin-native prototype on OpenClaw to demonstrate practical feasibility. This perspective provides a concrete blueprint for structuring runtime security controls, managing trust propagation, and enforcing execution containment in autonomous AI agents. Our code is available at this https URL .

54. DepthKV: Layer-Dependent KV Cache Pruning for Long-Context LLM Inference

Authors: Zahra Dehghanighobadi , Asja Fischer
URL: https://arxiv.org/abs/2604.24647
Abstract:

Long-context reasoning is a critical capability of large language models (LLMs), enabling applications such as long-document understanding, summarization, and code generation. However, efficient autoregressive inference relies on the key-value (KV) cache, whose memory footprint grows linearly with sequence length, leading to a major memory bottleneck. To mitigate this overhead, KV cache pruning methods discard cached tokens with low attention scores during inference. Most existing methods apply a uniform pruning ratio across layers, implicitly assuming that all layers contribute equally to overall model performance. We show that this assumption is suboptimal, as layers differ significantly in their sensitivity to pruning. We propose DepthKV, a layer-dependent pruning framework that allocates a fixed global KV budget across layers based on their sensitivity, rather than using a uniform allocation. Across multiple models and tasks, DepthKV consistently outperforms uniform pruning at the same global pruning ratio, demonstrating more effective utilization of the KV cache budget through layer-dependent allocation.

55. K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology

Authors: Soyeon Kim , Cheongwoong Kang , Myeongjin Lee , Eun-Chul Chang , Jaedeok Lee , Jaesik Choi
URL: https://arxiv.org/abs/2604.24645
Abstract:

The development of practical (multimodal) large language model assistants for Korean weather forecasters is hindered by the absence of a multidimensional, expert-level evaluation framework grounded in authoritative sources. To address this, we introduce K-MetBench, a diagnostic benchmark grounded in national qualification exams. It exposes critical gaps across four dimensions: expert visual reasoning of charts, logical validity via expert-verified rationales, Korean-specific geo-cultural comprehension, and fine-grained domain analysis. Our evaluation of 55 models reveals a profound modality gap in interpreting specialized diagrams and a reasoning gap where models hallucinate logic despite correct predictions. Crucially, Korean models outperform significantly larger global models in local contexts, demonstrating that parameter scaling alone cannot resolve cultural dependencies. K-MetBench serves as a roadmap for developing reliable, culturally aware expert AI agents. The dataset is available at this https URL .

56. Less Is More: Engineering Challenges of On-Device Small Language Model Integration in a Mobile Application

Authors: William Oliveira
URL: https://arxiv.org/abs/2604.24636
Abstract:

On-device Small Language Models (SLMs) promise fully offline, private AI experiences for mobile users (no cloud dependency, no data leaving the device). But is this promise achievable in practice? This paper presents a longitudinal practitioner case study documenting the engineering challenges of integrating SLMs (Gemma 4 E2B, 2.6B parameters; Qwen3 0.6B, 600M parameters) into Palabrita, a production Android word-guessing game. Over a 5-day development sprint comprising 204 commits (~90 directly AI-related), the system underwent a radical transformation: from an ambitious design where the LLM generated complete structured puzzles (word, category, difficulty, and five hints as JSON) to a pragmatic architecture where curated word lists provide the words and the LLM generates only three short hints, with a deterministic fallback if it fails. We identify five categories of failures specific to on-device SLM integration: output format violations, constraint violations, context quality degradation, latency incompatibility, and model selection instability. For each failure category, we document the observed symptoms, root causes, and the prompt engineering and architectural strategies that effectively mitigated them, including multi-layer defensive parsing, contextual retry with failure feedback, session rotation, progressive prompt hardening, and systematic responsibility reduction. Our findings demonstrate that on-device SLMs are viable for production mobile applications, but only when the developer accepts a fundamental constraint: the most reliable on-device LLM feature is one where the LLM does the least. We distill our experience into eight actionable design heuristics for practitioners integrating SLMs into mobile apps.

57. Learning to Route Queries to Heads for Attention-based Re-ranking with Large Language Models

Authors: Yuxing Tian , Fengran Mo , Zhiqi Huang , Weixu Zhang , Jian-Yun Nie
URL: https://arxiv.org/abs/2604.24608
Abstract:

Large Language Models (LLMs) have recently been explored as fine-grained zero-shot re-rankers by leveraging attention signals to estimate document relevance. However, existing methods either aggregate attention signals across all heads or rely on a statically selected subset identified by heuristic rules. This solution can be suboptimal because the informative heads can vary across queries or domains. Moreover, naively combining multiple heads can degrade performance due to redundancy or conflicting ranking signals. In this paper, we propose a query-dependent head selection method, RouteHead, for attention-based re-ranking with LLMs. Specifically, we learn a lightweight router that can map each query to an optimal head set, and relevance scores are computed by aggregating attention signals only from these heads. Since query-to-head optimal labels are unavailable, we first construct pseudo labels via an offline search. The router represents each head with a learnable embedding and represents each query using an embedding extracted from the hidden states of the frozen LLM. Then it is trained on the pseudo labels with a sparsity regularizer. Experiments on diverse benchmarks and multiple LLM backbones show that the proposed method consistently outperforms strong baselines.

58. Skill Retrieval Augmentation for Agentic AI

Authors: Weihang Su , Jianming Long , Qingyao Ai , Yichen Tang , Changyue Wang , Yiteng Tu , Yiqun Liu
URL: https://arxiv.org/abs/2604.24594
Abstract:

As large language models (LLMs) evolve into agentic problem solvers, they increasingly rely on external, reusable skills to handle tasks beyond their native parametric capabilities. In existing agent systems, the dominant strategy for incorporating skills is to explicitly enumerate available skills within the context window. However, this strategy fails to scale: as skill corpora expand, context budgets are consumed rapidly, and the agent becomes markedly less accurate in identifying the right skill. To this end, this paper formulates Skill Retrieval Augmentation (SRA), a new paradigm in which agents dynamically retrieve, incorporate, and apply relevant skills from large external skill corpora on demand. To make this problem measurable, we construct a large-scale skill corpus and introduce SRA-Bench, the first benchmark for decomposed evaluation of the full SRA pipeline, covering skill retrieval, skill incorporation, and end-task execution. SRA-Bench contains 5,400 capability-intensive test instances and 636 manually constructed gold skills, which are mixed with web-collected distractor skills to form a large-scale corpus of 26,262 skills. Extensive experiments show that retrieval-based skill augmentation can substantially improve agent performance, validating the promise of the paradigm. At the same time, we uncover a fundamental gap in skill incorporation: current LLM agents tend to load skills at similar rates, regardless of whether a gold skill is retrieved or whether the task actually requires external capabilities. This shows that the bottleneck in skill augmentation lies not only in retrieval but also in the base model’s ability to determine which skill to load and when external loading is actually needed. These findings position SRA as a distinct research problem and establish a foundation for the scalable augmentation of capabilities in future agent systems.

59. Layerwise Convergence Fingerprints for Runtime Misbehavior Detection in Large Language Models

Authors: Nay Myat Min , Long H. Pham , Jun Sun
URL: https://arxiv.org/abs/2604.24542
Abstract:

Large language models deployed at runtime can misbehave in ways that clean-data validation cannot anticipate: training-time backdoors lie dormant until triggered, jailbreaks subvert safety alignment, and prompt injections override the deployer’s instructions. Existing runtime defenses address these threats one at a time and often assume a clean reference model, trigger knowledge, or editable weights, assumptions that rarely hold for opaque third-party artifacts. We introduce Layerwise Convergence Fingerprinting (LCF), a tuning-free runtime monitor that treats the inter-layer hidden-state trajectory as a health signal: LCF computes a diagonal Mahalanobis distance on every inter-layer difference, aggregates via Ledoit-Wolf shrinkage, and thresholds via leave-one-out calibration on 200 clean examples, with no reference model, trigger knowledge, or retraining. Evaluated on four architectures (Llama-3-8B, Qwen2.5-7B, Gemma-2-9B, Qwen2.5-14B) across backdoors, jailbreaks, and prompt injection (56 backdoor combinations, 3 jailbreak techniques, and BIPIA email + code-QA), LCF reduces mean backdoor attack success rate (ASR) below 1% on Qwen2.5-7B and Gemma-2 and to 1.3% on Qwen2.5-14B, detects 92-100% of DAN jailbreaks (62-100% for GCG and softer role-play), and flags 100% of text-payload injections across all eight (model, domain) cells, at 12-16% backdoor FPR and <0.1% inference overhead. A single aggregation score covers all three threat families without threat-specific tuning, positioning LCF as a general-purpose runtime safety layer for cloud-served and on-device LLMs.

60. Understanding the Limits of Automated Evaluation for Code Review Bots in Practice

Authors: Veli Karakaya , Utku Boran Torun , Baykal Mehmet Uçar , Eray Tüzün
URL: https://arxiv.org/abs/2604.24525
Abstract:

Automated code review (ACR) bots are increasingly used in industrial software development to assist developers during pull request (PR) review. As adoption grows, a key challenge is how to evaluate the usefulness of bot-generated comments reliably and at scale. In practice, such evaluation often relies on developer actions and annotations that are shaped by contextual and organizational factors, complicating their use as objective ground truth. We examine the feasibility and limitations of automating the evaluation of LLM-powered ACR bots in an industrial setting. We analyze an industrial dataset from Beko comprising 2,604 bot-generated PR comments, each labeled by software engineers as fixed/wontFix. Two automated evaluation approaches, G-Eval and an LLM-as-a-Judge pipeline, are applied using both binary decisions and a 0-4 Likert-scale formulation, enabling a controlled comparison against developer-provided labels. Across Gemini-2.5-pro, GPT-4.1-mini, and GPT-5.2, both evaluation strategies achieve only moderate alignment with human labels. Agreement ratios range from approximately 0.44 to 0.62, with noticeable variation across models and between binary and Likert-scale formulations, indicating sensitivity to both model choice and evaluation design. Our findings highlight practical limitations in fully automating the evaluation of ACR bot comments in industrial contexts. Developer actions such as resolving or ignoring comments reflect not only comment quality, but also contextual constraints, prioritization decisions, and workflow dynamics that are difficult to capture through static artifacts. Insights from a follow-up interview with a software engineering director further corroborate that developer labeling behavior is strongly influenced by workflow pressures and organizational constraints, reinforcing the challenges of treating such signals as objective ground truth.

61. Why AI Harms Can’t Be Fixed One Identity at a Time: What 5300 Incident Reports Reveal About Intersectionality

Authors: Edyta Bogucka , Sanja Šćepanović , Daniele Quercia
URL: https://arxiv.org/abs/2604.24519
Abstract:

AI risk assessment is the primary tool for identifying harms caused by AI systems. These include intersectional harms, which arise from the interaction between identity categories (e.g., class and skin tone) and which do not occur, or occur differently, when those categories are considered separately. Yet existing AI risk assessments are still built around isolated identity categories, and when intersections are considered, they focus almost exclusively on race and gender. Drawing on a large-scale analysis of documented AI incidents, we show that AI harms do not occur one identity category at a time. Using a structured rubric applied with a Large Language Model (LLM), we analyze 5,300 reports from 1,200 documented incidents in the AI Incident Database, the most curated source of incident data. From these reports, we identify 1,513 harmed subjects and their associated identity categories, achieving 98% accuracy. At the level of individual categories, we find that age and political identity appear in documented AI harms at rates comparable to race and gender. At the level of intersecting categories, harm is amplified up to three times at specific intersections: adolescent girls, lower-class people of color, and upper-class political elites. We argue that intersectionality should be a core component of AI risk assessment to more accurately capture how harms are produced and distributed across social groups.

62. GAMMAF: A Common Framework for Graph-Based Anomaly Monitoring Benchmarking in LLM Multi-Agent Systems

Authors: Pablo Mateo-Torrejón , Alfonso Sánchez-Macián
URL: https://arxiv.org/abs/2604.24477
Abstract:

The rapid integration of Large Language Models (LLMs) into Multi-Agent Systems (MAS) has significantly enhanced their collaborative problem-solving capabilities, but it has also expanded their attack surfaces, exposing them to vulnerabilities such as prompt infection and compromised inter-agent communication. While emerging graph-based anomaly detection methods show promise in protecting these networks, the field currently lacks a standardized, reproducible environment to train these models and evaluate their efficacy. To address this gap, we introduce Gammaf (Graph-based Anomaly Monitoring for LLM Multi-Agent systems Framework), an open-source benchmarking platform. Gammaf is not a novel defense mechanism itself, but rather a comprehensive evaluation architecture designed to generate synthetic multi-agent interaction datasets and benchmark the performance of existing and future defense models. The proposed framework operates through two interdependent pipelines: a Training Data Generation stage, which simulates debates across varied network topologies to capture interactions as robust attributed graphs, and a Defense System Benchmarking stage, which actively evaluates defense models by dynamically isolating flagged adversarial nodes during live inference rounds. Through rigorous evaluation using established defense baselines (XG-Guard and BlindGuard) across multiple knowledge tasks (such as MMLU-Pro and GSM8K), we demonstrate Gammaf’s high utility, topological scalability, and execution efficiency. Furthermore, our experimental results reveal that equipping an LLM-MAS with effective attack remediation not only recovers system integrity but also substantially reduces overall operational costs by facilitating early consensus and cutting off the extensive token generation typical of adversarial agents.

63. Measuring Successful Cooperation in Human-AI Teamwork: Development and Validation of the Perceived Cooperativity and Teaming Perception Scales

Authors: Christiane Attig , Christiane Wiebel-Herboth , Patricia Wollstadt , Tim Schrills , Mourad Zoubir , Thomas Franke
URL: https://arxiv.org/abs/2604.24461
Abstract:

As human-AI cooperation becomes increasingly prevalent, reliable instruments for assessing the subjective quality of cooperative human-AI interaction are needed. We introduce two theoretically grounded scales: the Perceived Cooperativity Scale (PCS), grounded in joint activity theory, and the Teaming Perception Scale (TPS), grounded in evolutionary cooperation theory. The PCS captures an agent’s perceived cooperative capability and practice within a single interaction sequence; the TPS captures the emergent sense of teaming arising from mutual contribution and support. Both scales were adapted for human-human cooperation to enable cross-agent comparisons. Across three studies (N = 409) encompassing a cooperative card game, LLM interaction, and a decision-support system, analyses of dimensionality, reliability, and validity indicated that both scales successfully differentiated between cooperation partners of varying cooperative quality and showed construct validity in line with expectations. The scales provide a basis for empirical investigation and system evaluation across a wide range of human-AI cooperation contexts.

64. Kwai Summary Attention Technical Report

Authors: Chenglong Chu , Guorui Zhou , Guowang Zhang , Han Li , Hao Peng , Hongtao Cheng , Jian Liang , Jiangxia Cao , Kun Gai , Lingzhi Zhou , Lu Ren , Qi Zhang , Ruiming Tang , Ruitao Wang , Xinchen Luo , Yi Su , Zhiyuan Liang , Ziqi Wang , Boyang Ding , Chengru Song , Dunju Zang , Hui Wang , Jiao Ou , Jiaxin Deng , Jijun Shi , Jinghao Zhang , Junmin Chen , Lejian Ren , Minxuan Lv , Qianqian Wang , Qigen Hu , Shiyao Wang , Siyang Mao , Tao Wang , Xingmei Wang , Zhixin Ling , Ziming Li , Zixing Zhang
URL: https://arxiv.org/abs/2604.24432
Abstract:

Long-context ability, has become one of the most important iteration direction of next-generation Large Language Models, particularly in semantic understanding/reasoning, code agentic intelligence and recommendation system. However, the standard softmax attention exhibits quadratic time complexity with respect to sequence length. As the sequence length increases, this incurs substantial overhead in long-context settings, leading the training and inference costs of extremely long sequences deteriorate rapidly. Existing solutions mitigate this issue through two technique routings: i) Reducing the KV cache per layer, such as from the head-level compression GQA, and the embedding dimension-level compression MLA, but the KV cache remains linearly dependent on the sequence length at a 1:1 ratio. ii) Interleaving with KV Cache friendly architecture, such as local attention SWA, linear kernel GDN, but often involve trade-offs among KV Cache and long-context modeling effectiveness. Besides the two technique routings, we argue that there exists an intermediate path not well explored: {Maintaining a linear relationship between the KV cache and sequence length, but performing semantic-level compression through a specific ratio $k$}. This $O(n/k)$ path does not pursue a ``minimum KV cache’’, but rather trades acceptable memory costs for complete, referential, and interpretable retention of long distant dependency. Motivated by this, we propose Kwai Summary Attention (KSA), a novel attention mechanism that reduces sequence modeling cost by compressing historical contexts into learnable summary tokens.

65. Scaling Properties of Continuous Diffusion Spoken Language Models

Authors: Jason Ramapuram , Eeshan Gunesh Dhekane , Amitis Shidani , Dan Busbridge , Bogdan Mazoure , Zijin Gu , Russ Webb , Tatiana Likhomanenko , Navdeep Jaitly
URL: https://arxiv.org/abs/2604.24416
Abstract:

Speech-only spoken language models (SLMs) lag behind text and text-speech models in performance, with recent discrete autoregressive (AR) SLMs indicating significant computational and data demands to match text models. Since discretizing continuous speech for AR creates bottlenecks, we explore whether continuous diffusion (CD) SLM is more viable. To quantify the SLMs linguistic quality, we introduce the phoneme Jensen-Shannon divergence (pJSD) metric. Our analysis reveals CD SLMs, mirroring AR behavior, exhibit scaling laws for validation loss and pJSD, and show optimal token-to-parameter ratios decreasing as compute scales. However, for the latter, loss becomes insensitive to choice of data and model sizes, showing potential for fast inference. Scaling CD SLMs to 16B parameters with tens of millions of hours of conversational data enables generation of emotive, prosodic, multi-speaker, multilingual speech, though achieving long-form coherence remains a significant challenge.

66. All That Glitters Is Not Audio: Rethinking Text Priors and Audio Reliance in Audio-Language Evaluation

Authors: Leonardo Haw-Yang Foo , Chih-Kai Yang , Chen-An Li , Ke-Han Lu , Hung-yi Lee
URL: https://arxiv.org/abs/2604.24401
Abstract:

Large Audio-Language Models show consistent performance gains across speech and audio benchmarks, yet high scores may not reflect true auditory perception. If a model can answer questions without processing the acoustic signal, the benchmark fails as a measure of auditory understanding. We present a diagnostic framework using two axes: text prior, which measures answerability from text and general knowledge alone, and audio reliance, which assesses actual dependency on the acoustic signal. Evaluating eight LALMs across three benchmarks, we find that models retain 60-72% of their full audio scores even without any audio input. Moreover, among items that require audio, only 3.0-4.2% need the complete audio clip; the majority can be resolved using localized fragments. These findings challenge the assumption that benchmark performance equals robust audio understanding, and we conclude with practical guidelines for improving evaluation reliability and benchmark design.

67. Global Context or Local Detail? Adaptive Visual Grounding for Hallucination Mitigation

Authors: Yubo Jiang , Xin Yang , Abudukelimu Wuerkaixi , Zheming Yuan , Xuxin Cheng , Fengying Xie , Zhiguo Jiang , Cao Liu , Ke Zeng , Haopeng Zhang
URL: https://arxiv.org/abs/2604.24396
Abstract:

Vision-Language Models (VLMs) are frequently undermined by object hallucination–generating content that contradicts visual reality–due to an over-reliance on linguistic priors. We introduce Positive-and-Negative Decoding (PND), a training-free inference framework that intervenes directly in the decoding process to enforce visual fidelity. PND is motivated by our key finding of a critical attention deficit in VLMs, where visual features are empirically under-weighted. Our framework corrects this via a dual-path contrast: The positive path amplifies salient visual evidence using multi-layer attention to encourage faithful descriptions, directly counteracting the attention deficit. Simultaneously, the negative path identifies and degrades the core object’s features to create a strong counterfactual, which penalizes ungrounded, prior-dominant generation. By contrasting the model’s outputs from these two perspectives at each step, PND steers generation towards text that is not just linguistically probable, but visually factual. Extensive experiments on benchmarks like POPE, MME, and CHAIR show that PND achieves state-of-the-art performance with up to 6.5% accuracy improvement, substantially reducing object hallucination while also enhancing descriptive detail–all without requiring any model retraining. The method generalizes effectively across diverse VLM architectures including LLaVA, InstructBLIP, InternVL, and Qwen-VL.

68. SeaEvo: Advancing Algorithm Discovery with Strategy Space Evolution

Authors: Sichun Luo , Yi Huang , Haochen Luo , Fengyuan Liu , Guanzhi Deng , Lei Li , Qinghua Yao , Zefa Hu , Junlan Feng , Qi Liu
URL: https://arxiv.org/abs/2604.24372
Abstract:

LLM-guided evolutionary search has emerged as a promising paradigm for automated algorithm discovery, yet most systems track search progress primarily through executable programs and scalar fitness. Even when natural-language reflection is used, it is often used locally in mutation prompts or stored without an explicit population-level organization of strategic directions. As a result, evolutionary search can struggle to distinguish syntactically different implementations of the same idea, preserve lower-fitness but strategically promising directions, or detect when an entire family of strategies has saturated. We introduce \model, a modular strategy-space layer that elevates natural-language strategy descriptions from transient prompt context to first-class population-level evolutionary state in LLM-driven program search. \model augments each candidate program with an explicit natural language strategy description and uses this representation in three ways: Strategy Articulation turns mutation into a diagnose-direct-implement process; Stratified Experience Retrieval organizes the archive into strategy clusters and selects inspirations by behavioral complementarity; and Strategic Landscape Navigation periodically summarizes effective, saturated, and underexplored strategy families to guide future mutations. Across mathematical algorithm discovery, systems optimization, and agent-scaffold benchmarks, \model improves the underlying evolutionary backbones in most settings, with particularly large gains (21% relative improvement) on open-ended system optimization tasks. These results suggest that persistent strategy representations provide a practical mechanism for improving the robustness and efficiency of LLM-guided evolutionary search, suggesting a path toward compound AI systems that accumulate algorithmic knowledge over time.

69. DPRM: A Plug-in Doob h transform-induced Token-Ordering Module for Diffusion Language Models

Authors: Dake Bu , Wei Huang , Andi Han , Hau-San Wong , Qingfu Zhang , Taiji Suzuki , Atsushi Nitanda
URL: https://arxiv.org/abs/2604.24357
Abstract:

Diffusion language models generate without a fixed left-to-right order, making token ordering a central algorithmic choice: which tokens should be revealed, retained, revised or verified at each step? Existing systems mainly use random masking or confidence-driven ordering. Random masking creates train–test mismatch, while confidence-only rules are efficient but can be myopic and suppress useful exploration. We introduce DPRM (Doob h-transform Process Reward Model), a plug-in token-ordering module for diffusion language models. DPRM keeps the host architecture, denoising objective and supervision unchanged, and changes only the ordering policy. It starts from confidence-driven progressive ordering and gradually shifts to Doob h transform Process Reward guided ordering through online estimates. We characterize the exact DPRM policy as a reward-tilted Gibbs reveal law, prove O(1/N) convergence of the stagewise Soft-BoN approximation, and show that the online bucketized controller tracks the exact DPRM score at empirical-Bernstein rates. Under tractable optimization assumptions, DPRM also yields a sample-complexity advantage over random and confidence-only ordering. DPRM improves over confidence-based baselines in pretraining, post-training, test-time scaling, and single-cell masked diffusion, with particularly strong gains on harder reasoning subsets. In protein, molecular generation and DNA design, the effect is more multi-objective: ordering-aware variants significantly improve selected structural or fragment-constrained metrics while not uniformly dominating the host baseline on every quality metric. These results identify token ordering as a fundamental control axis in diffusion language models and establish DPRM as a general-purpose module for improving it. Code is available at this https URL .

70. SycoPhantasy: Quantifying Sycophancy and Hallucination in Small Open Weight VLMs for Vision-Language Scoring of Fantasy Characters

Authors: Arya Shah , Deepali Mishra , Chaklam Silpasuwanchai
URL: https://arxiv.org/abs/2604.24346
Abstract:

Vision-language models (VLMs) are increasingly deployed as evaluators in tasks requiring nuanced image understanding, yet their reliability in scoring alignment between images and text descriptions remains underexplored. We investigate whether small, open-weight VLMs exhibit \emph{sycophantic} behavior when evaluating image-text alignment: assigning high scores without grounding their judgments in visual evidence. To quantify this phenomenon, we introduce the \emph{Bluffing Coefficient} (\bc), a metric that measures the mismatch between a model’s score and its evidence recall. We evaluate six open-weight VLMs ranging from 450M to 8B parameters on a benchmark of 173,810 AI-generated character portraits paired with detailed textual descriptions. Our analysis reveals a significant inverse correlation between model size and sycophancy rate ($r = -0.96$, $p = 0.002$), with smaller models exhibiting substantially higher rates of unjustified high scores. The smallest model tested (LFM2-VL, 450M) produced sycophantic evaluations in 22.3\% of cases, compared to 6.0\% for the largest (LLaVA-1.6, 7B). These findings have direct implications for the deployment of small, open-weight VLMs as automated evaluators within attribute-rich, synthetic image evaluation tasks, where the gap between assigned scores and cited visual evidence is both measurable and consequential.

71. See Further, Think Deeper: Advancing VLM’s Reasoning Ability with Low-level Visual Cues and Reflection

Authors: Zhiheng Wu , Tong Wang , Shuning Wang , Naiming Liu , Yumeng Zhang
URL: https://arxiv.org/abs/2604.24339
Abstract:

Recent advances in Vision-Language Models (VLMs) have benefited from Reinforcement Learning (RL) for enhanced reasoning. However, existing methods still face critical limitations, including the lack of low-level visual information and effective visual feedback. To address these problems, this paper proposes a unified multimodal interleaved reasoning framework \textbf{ForeSight}, which enables VLMs to \textbf{See Further} with low-level visual cues and \textbf{Think Deeper} with effective visual feedback. First, it introduces a set of low-level visual tools to integrate essential visual information into the reasoning chain, mitigating the neglect of fine-grained visual features. Second, a mask-based visual feedback mechanism is elaborated to incorporate visual reflection into the thinking process, enabling the model to dynamically re-examine and update its answers. Driven by RL, ForeSight learns to autonomously decide on tool invocation and answer verification, with the final answer accuracy as the reward signal. To evaluate the performance of the proposed framework, we construct a new dataset, Character and Grounding SalBench (CG-SalBench), based on the SalBench dataset. Experimental results demonstrate that the ForeSight-7B model significantly outperforms other models with the same parameter scale, and even surpasses the current SOTA closed-source models on certain metrics.

72. MEMCoder: Multi-dimensional Evolving Memory for Private-Library-Oriented Code Generation

Authors: Mofei Li , Taozhi Chen , Guowei Yang , Jia Li
URL: https://arxiv.org/abs/2604.24222
Abstract:

Large Language Models (LLMs) excel at general code generation, but their performance drops sharply in enterprise settings that rely on internal private libraries absent from public pre-training corpora. While Retrieval-Augmented Generation (RAG) offers a training-free alternative by providing static API documentation, we find that such documentation typically provides only isolated definitions, leaving a fundamental knowledge gap. Specifically, LLMs struggle with a task-level lack of coordination patterns between APIs and an API-level misunderstanding of parameter constraints and boundary conditions. To address this, we propose MEMCoder, a novel framework that enables LLMs to autonomously accumulate and evolve Usage Guidelines across these two dimensions. MEMCoder introduces a Multi-dimensional Evolving Memory that captures distilled lessons from the model’s own problem-solving trajectories. During inference, MEMCoder employs a dual-source retrieval mechanism to inject both static documentation and relevant historical guidelines into the context. The framework operates in an automated closed loop by using objective execution feedback to reflect on successes and failures, resolve knowledge conflicts, and dynamically update memory. Extensive evaluations on the NdonnxEval and NumbaEval benchmarks demonstrate that MEMCoder substantially enhances existing RAG systems, yielding an average absolute pass@1 gain of 16.31%. Furthermore, MEMCoder exhibits vastly superior domain-specific adaptation compared to existing memory-based continual learning methods.

73. RefEvo: Agentic Design with Co-Evolutionary Verification for Agile Reference Model Generation

Authors: Yifan Zhang , Jianmin Ye , Jiahao Yang , Xi Wang
URL: https://arxiv.org/abs/2604.24218
Abstract:

As the complexity of System-on-Chip (SoC) designs grows, the shift-left paradigm necessitates the rapid development of high-fidelity reference models (typically written in SystemC) for early architecture exploration and verification. While Large Language Models (LLMs) show promise in code generation, their application to hardware modeling faces unique challenges: (1) Rigid, static workflows fail to adapt to varying design complexity, causing inefficiency; (2) Context window overflow in multi-turn interactions leads to catastrophic forgetting of critical specifications; and (3) the Coupled Validation Failure problem–where generated Testbenches (TBs) incorrectly validate flawed models due to correlated hallucinations–severely undermines reliability. To address these limitations, we introduce RefEvo, a dynamic multi-agent framework designed for agile and reliable reference modeling. RefEvo features three key innovations: (1) A Dynamic Design Planner that autonomously decomposes design specifications and constructs tailored execution workflows based on semantic complexity; (2) A Co-Evolutionary Verification Mechanism, which employs a Dialectical Arbiter to simultaneously rectify the model and verification logic against the specification (Spec) oracle, effectively mitigating false positives; and (3) A Spec Anchoring Strategy for lossless context compression. Evaluated on a diverse benchmark of 20 hardware modules, RefEvo achieves a 95% pass rate, outperforming static baselines by a large margin. Furthermore, our context optimization reduces token consumption by an average of 71.04%, achieving absolute savings of over 70,000 tokens per session for complex designs while maintaining 100% specification recall.

74. Agentic Witnessing: Pragmatic and Scalable TEE-Enabled Privacy-Preserving Auditing

Authors: Antony Rowstron
URL: https://arxiv.org/abs/2604.24203
Abstract:

Auditing the semantic properties of proprietary data creates a fundamental tension: verification requires transparent access, while proprietary rights demand confidentiality. While Zero-Knowledge Proofs (ZKPs) ensure privacy, they are typically limited to precise algebraic constraints and are ill-suited for verifying qualitative, unstructured properties, such as the logic within a codebase. We propose {\em Agentic Witnessing}, a framework that moves verification from attested execution to {\em attested reasoning}. The system is composed of three agents: a Verifier (who wants to check properties of a dataset), a Prover (who owns the dataset) and an Auditor (that inspects the dataset). The Verifier is allowed to ask a limited number of simple binary true/false questions to the auditor. By isolating an LLM-based Auditor within a Trusted Execution Environment (TEE), the system enables the Verifier to query a Prover’s private data via simple Boolean queries, without exposing the raw dataset. The Auditor uses the Model Context Protocol (MCP) to dynamically inspect the target dataset, producing a yes/no verdict accompanied by a cryptographic transcript: a signed hash chain binding the reasoning trace to both the original dataset and the TEE’s hardware root of trust. We demonstrate this architecture by automating the artifact evaluation process for 21 peer-reviewed computer science papers with released codebases on GitHub (e.g. Does the codebase implement the system described in the paper?). We verified five high-level properties of these codebases described in the corresponding publications, treating the source code as private. Our results show that TEE-enabled agentic auditing provides a mechanism for privacy-preserving oversight, effectively decoupling qualitative verification from the need for data disclosure.

75. Rewarding the Scientific Process: Process-Level Reward Modeling for Agentic Data Analysis

Authors: Zhisong Qiu , Shuofei Qiao , Kewei Xu , Yuqi Zhu , Lun Du , Ningyu Zhang , Huajun Chen
URL: https://arxiv.org/abs/2604.24198
Abstract:

Process Reward Models (PRMs) have achieved remarkable success in augmenting the reasoning capabilities of Large Language Models (LLMs) within static domains such as mathematics. However, their potential in dynamic data analysis tasks remains underexplored. In this work, we first present a empirical study revealing that general-domain PRMs struggle to supervise data analysis agents. Specifically, they fail to detect silent errors, logical flaws that yield incorrect results without triggering interpreter exceptions, and erroneously penalize exploratory actions, mistaking necessary trial-and-error exploration for grounding failures. To bridge this gap, we introduce DataPRM, a novel environment-aware generative process reward model that (1) can serve as an active verifier, autonomously interacting with the environment to probe intermediate execution states and uncover silent errors, and (2) employs a reflection-aware ternary reward strategy that distinguishes between correctable grounding errors and irrecoverable mistakes. We design a scalable pipeline to construct over 8K high-quality training instances for DataPRM via diversity-driven trajectory generation and knowledge-augmented step-level annotation. Experimental results demonstrate that DataPRM improves downstream policy LLMs by 7.21% on ScienceAgentBench and 11.28% on DABStep using Best-of-N inference. Notably, with only 4B parameters, DataPRM outperforms strong baselines, and exhibits robust generalizability across diverse Test-Time Scaling strategies. Furthermore, integrating DataPRM into Reinforcement Learning yields substantial gains over outcome-reward baselines, achieving 78.73% on DABench and 64.84% on TableBench, validating the effectiveness of process reward supervision. Code is available at this https URL .

76. MultiDx: A Multi-Source Knowledge Integration Framework towards Diagnostic Reasoning

Authors: Yimin Deng , Zhenxi Lin , Yejing Wang , Guoshuai Zhao , Pengyue Jia , Zichuan Fu , Derong Xu , Yefeng Zheng , Xiangyu Zhao , Li Zhu , Xian Wu , Xueming Qian
URL: https://arxiv.org/abs/2604.24186
Abstract:

Diagnostic prediction and clinical reasoning are critical tasks in healthcare applications. While Large Language Models (LLMs) have shown strong capabilities in commonsense reasoning, they still struggle with diagnostic reasoning due to limited domain knowledge. Existing approaches often rely on internal model knowledge or static knowledge bases, resulting in knowledge insufficiency and limited adaptability, which hinder their capacity to perform diagnostic reasoning. Moreover, these methods focus solely on the accuracy of final predictions, overlooking alignment with standard clinical reasoning trajectories. To this end, we propose MultiDx, a two-stage diagnostic reasoning framework that performs differential diagnosis by analyzing evidence collected from multiple knowledge sources. Specifically, it first generates suspected diagnoses and reasoning paths by leveraging knowledge from web search, SOAP-formatted case, and clinical case database. Then it integrates multi-perspective evidence through matching, voting, and differential diagnosis to generate the final prediction.~Extensive experiments on two public benchmarks demonstrate the effectiveness of our approach.

77. MemeScouts@LT-EDI 2026: Asking the Right Questions – Prompted Weak Supervision for Meme Hate Speech Detection

Authors: Ivo Bueno , Lea Hirlimann , Enkelejda Kasneci
URL: https://arxiv.org/abs/2604.24179
Abstract:

Detecting hate speech in memes is challenging due to their multimodal nature and subtle, culturally grounded cues such as sarcasm and context. While recent vision-language models (VLMs) enable joint reasoning over text and images, end-to-end prompting can be brittle, as a single prediction must resolve target, stance, implicitness, and irony. These challenges are amplified in multilingual settings. We propose a prompted weak supervision (PWS) approach that decomposes meme understanding into targeted, question-based labeling functions with constrained answer options for homophobia and transphobia detection in the LT-EDI 2026 shared task. Using a quantized Qwen3-VLM to extract features by answering targeted questions, our method outperforms direct VLM classification, with substantial gains for Chinese and Hindi, ranking 1st in English, 2nd in Chinese, and 3rd in Hindi. Iterative refinement via error-driven LF expansion and feature pruning reduces redundancy and improves generalization. Our results highlight the effectiveness of prompted weak supervision for multilingual multimodal hate speech detection.

78. Meta-Aligner: Bidirectional Preference-Policy Optimization for Multi-Objective LLMs Alignment

Authors: Wenzhe Xu , Biao Liu , Yiyang Sun , Xin Geng , Ning Xu
URL: https://arxiv.org/abs/2604.24178
Abstract:

Multi-Objective Alignment aims to align Large Language Models (LLMs) with diverse and often conflicting human values by optimizing multiple objectives simultaneously. Existing methods predominantly rely on static preference weight construction strategies. However, rigidly aligning to fixed targets discards valuable intermediate information, as training responses inherently embody valid preference trade-offs even when deviating from the target. To address this limitation, we propose Meal, i.e., MEta ALigner, a bi-level meta-learning framework enabling bidirectional optimization between preferences and policy responses, generating instructive dynamic preferences for steadier training. Specifically, we introduce a preference-weight-net as a meta-learner to generate adaptive preference weights based on input prompts and update the preference weights as learnable parameters, while the LLM policy acts as a base-learner optimizing response generation conditioned on these preferences with rejection sampling strategy. Extensive empirical results demonstrate that our method achieves superior performance on several multi-objective benchmarks, validating the effectiveness of the dynamic bidirectional preference-policy optimization framework.

79. AdapTime: Enabling Adaptive Temporal Reasoning in Large Language Models

Authors: Yimin Deng , Yejing Wang , Zhenxi Lin , Zichuan Fu , Guoshuai Zhao , Derong Xu , Yefeng Zheng , Xiangyu Zhao , Xian Wu , Li Zhu , Xueming Qian
URL: https://arxiv.org/abs/2604.24175
Abstract:

Large language models have demonstrated strong reasoning capabilities in general knowledge question answering. However, their ability to handle temporal information remains limited. To address this limitation, existing approaches often involve external tools or manual verification and are tailored to specific scenarios, leading to poor generalizability. Moreover, these methods apply a fixed pipeline to all questions, overlooking the fact that different types of temporal questions require distinct reasoning strategies, which leads to unnecessary processing for simple cases and inadequate reasoning for complex ones. To this end, we propose AdapTime, an adaptive temporal reasoning method that dynamically executes reasoning steps based on the input context. Specifically, it involves three temporal reasoning actions: reformulate, rewrite and review, with an LLM planner guiding the reasoning process. AdapTime integrates seamlessly with state-of-the-art LLMs and significantly enhances their temporal reasoning capabilities without relying on external support. Extensive experiments demonstrate the effectiveness of our approach.

80. Defusing the Trigger: Plug-and-Play Defense for Backdoored LLMs via Tail-Risk Intrinsic Geometric Smoothing

Authors: Kaisheng Fan , Weizhe Zhang , Yishu Gao , Tegawendé F. Bissyandé , Xunzhu Tang
URL: https://arxiv.org/abs/2604.24162
Abstract:

Defending against backdoor attacks in large language models remains a critical practical challenge. Existing defenses mitigate these threats but typically incur high preparation costs and degrade utility via offline purification, or introduce severe latency via complex online interventions. To overcome this dichotomy, we present Tail-risk Intrinsic Geometric Smoothing (TIGS), a plug-and-play inference-time defense requiring no parameter updates, external clean data, or auxiliary generation. TIGS leverages the observation that successful backdoor triggers consistently induce localized attention collapse within the semantic content region. Operating entirely within the native forward pass, TIGS first performs content-aware tail-risk screening to identify suspicious attention heads and rows using sample-internal signals. It then applies intrinsic geometric smoothing: a weak content-domain correction preserves semantic anchoring, while a stronger full-row contraction disrupts trigger-dominant routing. Finally, a controlled full-row write-back reconstructs the attention matrix to ensure inference stability. Extensive evaluations demonstrate that TIGS substantially suppresses attack success rates while strictly preserving clean reasoning and open-ended semantic consistency. Crucially, this favorable security-utility-latency equilibrium persists across diverse architectures, including dense, reasoning-oriented, and sparse mixture-of-experts models. By structurally disrupting adversarial routing with marginal latency overhead, TIGS establishes a highly practical, deployment-ready defense standard for state-of-the-art LLMs.

81. Strategic Bidding in 6G Spectrum Auctions with Large Language Models

Authors: Ismail Lotfi , Ali Ghrayeb
URL: https://arxiv.org/abs/2604.24156
Abstract:

Efficient and fair spectrum allocation is a central challenge in 6G networks, where massive connectivity and heterogeneous services continuously compete for limited radio resources. We investigate the use of Large Language Models (LLMs) as bidding agents in repeated 6G spectrum auctions with budget constraints in vehicular networks. Each user equipment (UE) acts as a rational player optimizing its long-term utility through repeated interactions. Using the Vickrey-Clarke-Groves (VCG) mechanism as a benchmark for incentive-compatible, dominant-strategy truthfulness, we compare LLM-guided bidding against truthful and heuristic strategies. Unlike heuristics, LLMs leverage historical outcomes and prompt-based reasoning to adapt their bidding behavior dynamically. Results show that when the theoretical assumptions guaranteeing truthfulness hold, LLM bidders recover near-equilibrium outcomes consistent with VCG predictions. However, when these assumptions break – such as under static budget constraints – LLMs sustain longer participation and achieve higher utilities, revealing their ability to approximate adaptive equilibria beyond static mechanism design. This work provides the first systematic evaluation of LLM bidders in repeated spectrum auctions, offering new insights into how AI-driven agents can interact strategically and reshape market dynamics in future 6G networks.

82. Latency and Cost of Multi-Agent Intelligent Tutoring at Scale

Authors: Iizalaarab Elhaimeur , Nikos Chrisochoides
URL: https://arxiv.org/abs/2604.24110
Abstract:

Multi-agent LLM tutoring systems improve response quality through agent specialization, but each student query triggers several concurrent API calls whose latencies compound through a parallel-phase maximum effect that single-agent systems do not face. We instrument ITAS, a four-agent tutoring system built on Gemini 2.5 Flash and Google Vertex AI, across three throughput tiers (Standard PayGo, Priority PayGo, and Provisioned Throughput) and eleven concurrency levels up to 50 simultaneous users, producing over 3,000 requests drawn from a live graduate STEM deployment. Priority PayGo maintains flat sub-4-second response times across the full load range; Standard PayGo degrades substantially under classroom-scale concurrency; and Provisioned Throughput delivers the lowest latency at low concurrency but saturates its reserved capacity above approximately 20 concurrent users. Cost analysis places both pay-per-token tiers well below the price of a STEM textbook per student per semester under a worst-case usage ceiling. Provisioned Throughput, expensive under continuous provisioning, becomes cost-competitive for institutions that can predict and concentrate their traffic toward high utilization. These results provide concrete tier-selection guidance across deployment scales from a single seminar to a university-wide rollout.

83. TACO: Efficient Communication Compression of Intermediate Tensors for Scalable Tensor-Parallel LLM Training

Authors: Man Liu , Xingchen Liu , Xingjian Tian , Bing Lu , Shengkay Lyu , Shengquan Yin , Wenjing Huang , Zheng Wei , Hairui Zhao , Guangming Tan , Dingwen Tao
URL: https://arxiv.org/abs/2604.24088
Abstract:

Handling communication overhead in large-scale tensor-parallel training remains a critical challenge due to the dense, near-zero distributions of intermediate tensors, which exacerbate errors under frequent communication and introduce significant computational overhead during compression. To this end, we propose TACO (Tensor-parallel Adaptive COmmunication compression), a robust FP8-based framework for compressing TP intermediate tensors. First, we employ a data-driven reshaping strategy combined with an Adaptive Scale-Hadamard Transform to enable high-fidelity FP8 quantization, while its Dual-Scale Quantization mechanism ensures numerical stability throughout training. Second, we design a highly fused compression operator to reduce memory traffic and kernel launch overhead, allowing efficient overlap with communication. Finally, we integrate TACO with existing state-of-the-art methods for Data and Pipeline Parallelism to develop a compression-enabled 3D-parallel training framework. Detailed experiments on GPT models and Qwen model demonstrate up to 1.87X end-to-end throughput improvement while maintaining near-lossless accuracy, validating the effectiveness and efficiency of TACO in large-scale training.

84. Jailbreaking Frontier Foundation Models Through Intention Deception

Authors: Xinhe Wang , Katia Sycara , Yaqi Xie
URL: https://arxiv.org/abs/2604.24082
Abstract:

Large (vision-)language models exhibit remarkable capability but remain highly susceptible to jailbreaking. Existing safety training approaches aim to have the model learn a refusal boundary between safe and unsafe, based on the user’s intent. It has been found that this binary training regime often leads to brittleness, since the user intent cannot reliably be evaluated, especially if the attacker obfuscates their intent, and also makes the system seem unhelpful. In response, frontier models, such as GPT-5, have shifted from refusal-based safeguards to safe completion, that aims to maximize helpfulness while obeying safety constraints. However, safe completion could be exploited when a user pretends their intention is benign. Specifically, this intent inversion would be effective in multi-turn conversation, where the attacker has multiple opportunities to reinforce their deceptively benign intent. In this work, we introduce a novel multi-turn jailbreaking method that exploits this vulnerability. Our approach gradually builds conversational trust by simulating benign-seeming intentions and by exploiting the consistency property of the model, ultimately guiding the target model toward harmful, detailed outputs. Most crucially, our approach also uncovered an additional class of model vulnerability that we call para-jailbreaking that has been unnoticed up to now. Para-jailbreaking describes the situation where the model may not reveal harmful direct reply to the attack query, however the information that it reveals is nevertheless harmful. Our contributions are threefold. First, it achieves high success rates against frontier models including GPT-5-thinking and Claude-Sonnet-4.5. Second, our approach revealed and addressed para-jailbreaking harmful output. Third, experiments on multimodal VLM models showed that our approach outperformed state-of-the-art models.

85. The Pragmatic Persona: Discovering LLM Persona through Bridging Inference

Authors: Jisoo Yang (1), Jongwon Ryu (1), Minuk Ma (2), Trung X. Pham (3), Junyeong Kim (1) ((1) Chung-Ang University, (2) University of British Columbia, (3) Van Lang University)
URL: https://arxiv.org/abs/2604.24079
Abstract:

Large Language Models (LLMs) reveal inherent and distinctive personas through dialogue. However, most existing persona discovery approaches rely on surface-level lexical or stylistic cues, treating dialogue as a flat sequence of tokens and failing to capture the deeper discourse-level structures that sustain persona consistency. To address this limitation, we propose a novel analytical framework that interprets LLM dialogue through bridging inference – implicit conceptual relations that connect utterances via shared world knowledge and discourse coherence. By modeling these relations as structured knowledge graphs, our approach captures latent semantic links that govern how LLMs organize meaning across turns, enabling persona discovery at the level of discourse coherence rather than surface realizations. Experimental results across multiple reasoning backbones and target LLMs, ranging from small-scale models to 80B-parameter systems, demonstrate that bridging-inference graphs yield significantly stronger semantic coherence and more stable persona identification than frequency or style-based baselines. These results show that persona traits are consistently encoded in the structural organization of discourse rather than isolated lexical patterns. This work presents a systematic framework for probing, extracting, and visualizing latent LLM personas through the lens of Cognitive Discourse Theory, bridging computational linguistics, cognitive semantics, and persona reasoning in large language models. Codes are available at this https URL

86. QEVA: A Reference-Free Evaluation Metric for Narrative Video Summarization with Multimodal Question Answering

Authors: Woojun Jung , Junyeong Kim
URL: https://arxiv.org/abs/2604.24052
Abstract:

Video-to-text summarization remains underexplored in terms of comprehensive evaluation methods. Traditional n-gram overlap-based metrics and recent large language model (LLM)-based approaches depend heavily on human-written reference summaries, limiting their practicality and sensitivity to nuanced semantic aspects. In this paper, we propose QEVA, a reference-free metric evaluating candidate summaries directly against source videos through multimodal question answering. QEVA assesses summaries along three clear dimensions: Coverage, Factuality, and Chronology. We also introduce MLVU(VS)-Eval, a new annotated benchmark derived from the MLVU dataset, comprising 800 summaries generated from 200 videos using state-of-the-art video-language multimodal models. This dataset establishes a transparent and consistent framework for evaluation. Experimental results demonstrate that QEVA shows higher correlation with human judgments compared to existing approaches, as measured by Kendall’s $\tau_b$, $\tau_c$, and Spearman’s $\rho$. We hope that our benchmark and metric will facilitate meaningful progress in video-to-text summarization research and provide valuable insights for the development of future evaluation methods.

87. AgenticCache: Cache-Driven Asynchronous Planning for Embodied AI Agents

Authors: Hojoon Kim , Yuheng Wu , Thierry Tambe
URL: https://arxiv.org/abs/2604.24039
Abstract:

Embodied AI agents increasingly rely on large language models (LLMs) for planning, yet per-step LLM calls impose severe latency and cost. In this paper, we show that embodied tasks exhibit strong plan locality, where the next plan is largely predictable from the current one. Building on this, we introduce AgenticCache, a planning framework that reuses cached plans to avoid per-step LLM calls. In AgenticCache, each agent queries a runtime cache of frequent plan transitions, while a background Cache Updater asynchronously calls the LLM to validate and refine cached entries. Across four multi-agent embodied benchmarks, AgenticCache improves task success rate by 22% on average across 12 configurations (4 benchmarks x 3 models), reduces simulation latency by 65%, and lowers token usage by 50%. Cache-based plan reuse thus offers a practical path to low-latency, low-cost embodied agents. Code is available at this https URL .

88. From Skill Text to Skill Structure: The Scheduling-Structural-Logical Representation for Agent Skills

Authors: Qiliang Liang , Hansi Wang , Zhong Liang , Yang Liu
URL: https://arxiv.org/abs/2604.24026
Abstract:

LLM agents increasingly rely on reusable skills, capability packages that combine instructions, control flow, constraints, and tool calls. In most current agent systems, however, skills are still represented by text-heavy artifacts, including this http URL -style documents and structured records whose machine-usable evidence remains embedded largely in natural-language descriptions. This poses a challenge for skill-centered agent systems: managing skill collections and using skills to support agent both require reasoning over invocation interfaces, execution structure, and concrete side effects that are often entangled in a single textual surface. An explicit representation of skill knowledge may therefore help make these artifacts easier for machines to acquire and leverage. Drawing on Memory Organization Packets, Script Theory, and Conceptual Dependency from Schank and Abelson’s classical work on linguistic knowledge representation, we introduce what is, to our knowledge, the first structured representation for agent skill artifacts that disentangles skill-level scheduling signals, scene-level execution structure, and logic-level action and resource-use evidence: the Scheduling-Structural-Logical (SSL) representation. We instantiate SSL with an LLM-based normalizer and evaluate it on a corpus of skills in two tasks, Skill Discovery and Risk Assessment, and superiorly outperform the text-only baselines: in Skill Discovery, SSL improves MRR from 0.573 to 0.707; in Risk Assessment, it improves macro F1 from 0.744 to 0.787. These findings reveal that explicit, source-grounded structure makes agent skills easier to search and review. They also suggest that SSL is best understood as a practical step toward more inspectable, reusable, and operationally actionable skill representations for agent systems, rather than as a finished standard or an end-to-end mechanism for managing and using skills.

89. IntentVLM: Open-Vocabulary Intention Recognition through Forward-Inverse Modeling with Video-Language Models

Authors: Hamed Rahimi , Clemence Grislain , Adrien Jacquet Cretides , Olivier Sigaud , Mohamed Chetouani
URL: https://arxiv.org/abs/2604.24002
Abstract:

Improving the effectiveness of human-robot interaction requires social robots to accurately infer human goals through robust intention understanding. This challenge is particularly critical in multimodal settings, where agents must integrate heterogeneous signals including text, visual cues to form a coherent interpretation of user intent. This paper presents IntentVLM, a novel two-stage video-language framework designed for open-vocabulary human intention recognition. The approach is inspired by forward-inverse modeling in cognitive science by decomposing intention understanding into goal candidate generation followed by structured inference through selection, effectively reducing hallucinations in latent reasoning. Evaluated on the IntentQA and Inst-IT Bench datasets, IntentVLM achieves state-of-the-art results with up to 80% accuracy, notably surpassing the baseline performance by 30% and matches human performance. Our findings demonstrate that this structured reasoning approach enhances open-vocabulary intention understanding without catastrophic forgetting, offering a robust foundation for human-centered robotics.

90. EPM-RL: Reinforcement Learning for On-Premise Product Mapping in E-Commerce

Authors: Minhyeong Yu , Wonduk Seo
URL: https://arxiv.org/abs/2604.23993
Abstract:

Product mapping, the task of deciding whether two e-commerce listings refer to the same product, is a core problem for price monitoring and channel visibility. In real marketplaces, however, sellers frequently inject promotional keywords, platform-specific tags, and bundle descriptions into titles, causing the same product to appear under many different names. Recent LLM-based and multi-agent frameworks improve robustness and interpretability on such hard cases, but they often rely on expensive external APIs, repeated retrieval, and complex inference-time orchestration, making large-scale deployment costly and difficult in privacy-sensitive enterprise settings. To address these issues, we present EPM-RL, a reinforcement-learning-based framework for building an accurate and efficient on-premise e-commerce product mapping model. Our central idea is to distill high-cost agentic reasoning into a trainable in-house model. Starting from a curated set of product pairs with LLM-generated rationales and human verification, we first perform parameter-efficient fine-tuning (PEFT) on a small student model using structured reasoning outputs. We then further optimize the model with Reinforcement Learning (RL) using an agent-based reward that jointly evaluates output-format compliance, label correctness, reasoning–preference scores from specially designed judge models. Preliminary results show that EPM-RL consistently improves over PEFT-only training and offers a stronger quality–cost trade-off than commercial API-based baselines, while enabling private deployment and lower operational cost. These findings suggest that reinforcement learning can turn product mapping from a high-latency agentic pipeline into a scalable, inspectable, and production-ready in-house system.

91. Fix Initial Codes and Iteratively Refine Textual Directions Toward Safe Multi-Turn Code Correction

Authors: Yuto Tanaka , Issei Sato
URL: https://arxiv.org/abs/2604.23989
Abstract:

Recent work on large language models (LLMs) has emphasized the importance of scaling inference compute. From this perspective, the state-of-the-art method Scattered Forest Search (SFS) has been proposed, employing Monte Carlo Tree Search with carefully crafted initial seeds and textual optimization for multi-turn code correction. However, its complexity makes it unclear what factors contribute to improvements in inference performance. To address this problem, we analyze SFS and propose a simpler method, Iterative Refinement of Textual Directions (IRTD), which fixes initial codes and iteratively refines textual directions. Because of the simplicity of IRTD, we theoretically establish the safety of IRTD using Oracle-Guided Inductive Synthesis (OGIS). Experiments on several code generation benchmarks suggest that IRTD achieves inference performance comparable to state-of-the-art methods. These results indicate that, even without complex search structures, refining initial codes with high-quality textual directions alone can effectively improve inference performance.

92. Hindsight Preference Optimization for Financial Time Series Advisory

Authors: Yanwei Cui , Guanghui Wang , Xing Zhang , Peiyang He , Ziyuan Li , Bing Zhu , Wei Qiu , Xusheng Wang , Zheng Yu , Anqi Xin
URL: https://arxiv.org/abs/2604.23988
Abstract:

Time series models predict numbers; decision-makers need advisory – directional signals with reasoning, actionable suggestions, and risk management. Training language models for such predictive advisory faces a fundamental challenge: quality depends on outcomes unknown at prediction time. We bridge two ideas from reinforcement learning – using information unavailable during execution to retrospectively generate training signal, and preference alignment – and propose Hindsight Preference Optimization: observed outcomes let an LLM judge rank candidate advisories on dimensions that scalar metrics cannot capture, producing preference pairs for DPO without human annotation. We apply this to Vision-Language-Model-based predictive advisories on S&P 500 equity time series, demonstrated by a 4B model outperforming its 235B teacher on both accuracy and advisory quality.

93. Quantum Knowledge Graph: Modeling Context-Dependent Triplet Validity

Authors: Yao Wang , Zixu Geng , Jun Yan
URL: https://arxiv.org/abs/2604.23972
Abstract:

Knowledge graphs (KGs) are increasingly used to support large lan guage model (LLM) reasoning, but standard triplet-based KGs treat each relation as globally valid. In many settings, whether a relation should count as evidence depends on the context. We therefore formulate triplet validity as a triplet-specific function of context and refer to this formulation as a Quantum Knowledge Graph (QKG). We instantiate QKG in medicine using a diabetes-centered PrimeKG subgraph, whose 68,651 context-sensitive relations are further annotated with patient-group-specific constraints. We evaluate it in a reasoner–validator pipeline for medical question answering on a KG-grounded subset of MedReason containing 2,788 questions. With Haiku-4.5 as both the Reasoner and the Validator, KG-backed validation significantly improves over a no-validator baseline ($+0.61$ pp), and QKG with context matching yields the largest gain, outperforming both KG validation without context matching ($+0.79$ pp) and the no-validator baseline ($+1.40$ pp; paired McNemar, all $p<0.05$). Under a stronger validator (Qwen-3.6-Plus), the raw QKG gain over the no-validator baseline grows from $+1.40$ pp to $+5.96$ pp; the context-matching gap is non-significant ($p=0.73$) on the raw set but becomes borderline significant ($p=0.05$) after adjustment for knowledge leakage and suspicious questions, consistent with a benchmark-gold ceiling rather than a QKG limitation. Taken together, the results support the view that the value of a KG in LLM-based clinical reasoning lies not merely in storing medically related facts, but in representing whether those facts are applicable to the specific patient context. For reproducibility and further research, we release the curated QKG datasets and source code.\footnote{ this https URL }

94. KOMBO: Korean Character Representations Based on the Combination Rules of Subcharacters

Authors: SungHo Kim , Juhyeong Park , Yeachan Kim , SangKeun Lee
URL: https://arxiv.org/abs/2604.23948
Abstract:

The Korean writing system, \textit{Hangeul}, has a unique character representation rigidly following the invention principles recorded in \textit{Hunminjeongeum}.\footnote{\textit{Hunminjeongeum} is a book published in 1446 that describes the principles of invention and usage of \textit{Hangeul}, devised by King Sejong \cite{Hunminjeongeum_Guide}.} However, existing pre-trained language models (PLMs) for Korean have overlooked these principles. In this paper, we introduce a novel framework for Korean PLMs called KOMBO, which firstly brings the invention principles of \textit{Hangeul} to represent character. Our proposed method, KOMBO, exhibits notable experimental proficiency across diverse NLP tasks. In particular, our method outperforms the state-of-the-art Korean PLM by an average of 2.11\% in five Korean natural language understanding tasks. Furthermore, extensive experiments demonstrate that our proposed method is suitable for comprehending the linguistic features of the Korean language. Consequently, we shed light on the superiority of using subcharacters over the typical subword-based approach for Korean PLMs. Our code is available at: this https URL .

Authors: Qiming Yuan , Linyi Han , Nam Ling , Cihan Ruan
URL: https://arxiv.org/abs/2604.23942
Abstract:

People increasingly turn to large language models (LLMs) to interpret ambiguous social situations: a delayed text reply, an unusually cold supervisor, a teacher’s mixed signals, or a boundary-crossing friend. Yet in many such cases, no stable interpretation can be verified from the available evidence alone. We study how LLMs respond to these situations across four domains: early-stage romantic relationships, teacher–student dynamics, workplace hierarchies, and ambiguous friendships. Across 72 responses from GPT, Claude, and Gemini, only 9 (12.5\%) genuinely preserved uncertainty. The remaining 87.5% produced interpretive closure through recurring pathways including narrative alignment, narrative reversal, normative advice under uncertainty, and hedged language that still supported a single conclusion. We further find that narrator perspective shapes the path to closure: first-person accounts more often elicited alignment, while third-person accounts invited more detached interpretation, even when the underlying situation remained comparable. Together, these findings show that LLMs do not simply assist interpersonal sensemaking; they tend to resolve ambiguity into coherent and actionable narratives. These results suggest that the central risk is not only that LLMs may misinterpret social situations, but that they may make unresolved situations feel prematurely settled. We frame this tendency as a design challenge for uncertainty-preserving social AI.

96. Constraint-Guided Multi-Agent Decompilation for Executable Binary Recovery

Authors: Yifan Zhang , Xiaohan Wang , Yueke Zhang , Kevin Leach
URL: https://arxiv.org/abs/2604.23940
Abstract:

Decompilation – recovering source code from compiled binaries – is essential for security analysis, malware reverse engineering, and legacy software maintenance. However, existing decompilers produce code that often fails to compile or execute correctly, limiting their practical utility. We present a multi-agent framework that transforms decompiled code into re-executable source through Multi-level Constraint-Guided Decompilation (MCGD). Our approach employs a hierarchical validation pipeline with three constraint levels: (1) syntactic correctness via parsing, (2) compilability via GCC, and (3) behavioral equivalence via LLM-generated test cases. When validation fails, specialized LLM agents iteratively refine the code using structured error feedback. We evaluate our framework on 1,641 real-world binaries from ExeBench across three decompilers (RetDec, Ghidra, and Angr). Our framework achieves 84-97% re-executability, improving baseline decompiler output by 28-89 percentage points. In comparison with state-of-the-art LLM-based decompilation methods using the same GPT-4o backbone, our approach (84.1%) outperforms LLM4Decompile (80.3%), SK2Decompile (73.9%), and SALT4Decompile (61.8%). Our ablation study reveals that execution-based validation is critical: compile-only approaches achieve 0% behavioral correctness despite 91-99% compilation rates. The system converges efficiently, with 90%+ binaries reaching correctness within 2 iterations at an average cost of $0.03-0.05 per binary. Our results demonstrate that constraint-guided agentic refinement can bridge the gap between raw decompiler output and practically useful source code.

97. SMSI: System Model Security Inference: Automated Threat Modeling for Cyber-Physical Systems

Authors: RoÝah Radaideh , Ali Khreis
URL: https://arxiv.org/abs/2604.23905
Abstract:

Threat modeling for cyber-physical systems (CPS) remains a largely manual exercise. This project presents SMSI (System Model Security Inference), a hybrid neuro-symbolic pipeline that starts from a SysML architecture model and produces a prioritized list of NIST 800-53 security controls. The prototype has three main stages: a deterministic parser mapping system components to vulnerabilities via the NVD; a family of retrieval and classification models linking vulnerabilities to MITRE ATT&CK techniques; and a control recommender. We explore three approaches for CVE-to-ATT&CK mapping: a supervised classifier using fine-tuned SecureBERT+, retrieval-based dense encoders, and a zero-shot LLM approach using Gemma-4 26B. We validate the pipeline on a healthcare IoT gateway with nine software components. For the ATT&CK-to-NIST stage, pretrained SecureBERT achieves the highest control retrieval scores, demonstrating that dense embeddings provide a strong basis for automated control recommendation.

98. Generative Synthetic Data for Causal Inference: Pitfalls, Remedies, and Opportunities

Authors: Yichen Xu
URL: https://arxiv.org/abs/2604.23904
Abstract:

Synthetic data offers a promising tool for privacy-preserving data release, augmentation, and simulation, but its use in causal inference requires preserving more than predictive fidelity. We show that fully generative tabular synthesizers, including GAN- and LLM-based models, can achieve strong train-on-synthetic-test-on-real performance while substantially distorting causal estimands such as the average treatment effect (ATE). We formalize this failure through sensitivity and tradeoff results showing that ATE preservation requires control of both the generated covariate law and the treatment-effect contrast in the outcome regression. Motivated by this observation, we propose a hybrid synthetic-data framework that generates covariates separately from the treatment and outcome mechanisms, using distance-to-closest-record diagnostics to monitor covariate synthesis and separately learned nuisance models to construct (W, A, Y) triplets. We further study targeted synthetic augmentation for practical positivity problems and characterize when added overlap support helps by improving conditional-effect estimation more than it shifts the covariate distribution. Finally, we develop a synthetic simulation engine for pre-analysis estimator evaluation, enabling finite-sample comparison of OR, IPW, AIPW, and TMLE under realistic covariate structure. Across experiments, hybrid synthetic data substantially improve ATE preservation relative to fully generative baselines and provide a practical diagnostic tool for robust causal analysis.

99. Evaluation of Prompt Injection Defenses in Large Language Models

Authors: Priyal Deep , Shane Emmons , Amy Fox , Kyle Bacon , Kelley McAllister , Krisztian Flautner
URL: https://arxiv.org/abs/2604.23887
Abstract:

LLM-powered applications routinely embed secrets in system prompts, yet models can be tricked into revealing them. We built an adaptive attacker that evolves its strategies over hundreds of rounds and tested it against nine defense configurations across more than 20,000 attacks. Every defense that relied on the model to protect itself eventually broke. The only defense that held was output filtering, which checks the model’s responses via hardcoded rules in separate application code before they reach the user, achieving zero leaks across 15,000 attacks. These results demonstrate that security boundaries must be enforced in application code, not by the model being attacked. Until such defenses are verified by tools like Swept AI, AI systems handling sensitive operations should be restricted to internal, trusted personnel.

100. Inverting Foundation Models of Brain Function with Simulation-Based Inference

Authors: Niels Bracher , Xavier Intes , Stefan T. Radev
URL: https://arxiv.org/abs/2604.23865
Abstract:

Foundation models of brain activity promise a new frontier for in silico neuroscience by emulating neural responses to complex stimuli across tasks and modalities. A natural next step is to ask whether these models can also be used in reverse. Can we recover a stimulus or its properties from synthetic brain activity? We study this question in a proof-of-concept setting using TRIBEv2. We pair the brain emulator with large language models (LLMs) that generate news headlines from linguistic parameters such as valence, arousal, and dominance. We then use simulation-based inference to learn a probabilistic mapping from brain maps to latent stimulus parameters. Our results show that these parameters can be recovered from predicted brain maps, validating the quality of neural encodings. They also show that LLMs can serve as controllable stimulus generators for simulated experiments. Together, these findings provide a step toward decoding and inverse design with foundation brain models.

101. Graph Memory Transformer (GMT)

Authors: Nicola Zanarini , Niccolò Ferrari
URL: https://arxiv.org/abs/2604.23862
Abstract:

We investigate whether the Feed-Forward Network (FFN) sublayer in a decoder-only transformer can be replaced by an explicit learned memory graph while preserving the surrounding autoregressive architecture. The proposed Graph Memory Transformer (GMT) keeps causal self-attention intact, but replaces the usual per-token FFN transformation with a memory cell that routes token representations over a learned bank of centroids connected by a learned directed transition matrix. In the base GMT v7 instantiation studied here, each of 16 transformer blocks contains 128 centroids, a 128 * 128 edge matrix, gravitational source routing, token-conditioned target selection, and a gated displacement readout. The cell therefore returns movement from an estimated source memory state toward a target memory state, rather than a retrieved value. The resulting model is a fully decoder-only language model with 82.2M trainable parameters and no dense FFN sublayers, compared with a 103.0M-parameter dense GPT-style baseline used in the evaluation. The base v7 model trains stably and exposes centroid usage, transition structure, and source-to-target movement as directly inspectable quantities of the forward computation. It remains behind the larger dense baseline in validation loss and perplexity (3.5995/36.58 vs. 3.2903/26.85), while showing close zero-shot benchmark behavior under the evaluated setting. These results are not intended as a state-of-the-art claim; they support the viability and structural interpretability of replacing dense within-token transformation with graph-mediated memory navigation. Broader scaling, optimized kernels, and more extensive benchmark evaluation are left for subsequent work.

102. Exploring Audio Hallucination in Egocentric Video Understanding

Authors: Ashish Seth , Xinhao Mei , Changsheng Zhao , Varun Nagaraja , Ernie Chang , Gregory P. Meyer , Gael Le Lan , Yunyang Xiong , Vikas Chandra , Yangyang Shi , Dinesh Manocha , Zhipeng Cai
URL: https://arxiv.org/abs/2604.23860
Abstract:

Egocentric videos provide a distinctive setting in which sound serves as crucial cues to understand user activities and surroundings, particularly when visual information is unstable or occluded due to continuous camera movement. State-of-the-art large audio-visual language models (AV-LLMs) can generate multimodal descriptions. However, we show in this work that they are prone to audio hallucinations, often inferring sounds from visual cues that are visible but not heard. We present a systematic and automatic evaluation framework for analyzing audio hallucinations in egocentric video through a targeted question-answering (Q/A) protocol. We curate a dataset of 300 egocentric videos and design 1,000 sound-focused questions to probe model outputs. To characterize hallucinations, we propose a grounded taxonomy that distinguishes between foreground action sounds from the user activities and background ambient sounds. Our evaluation shows that advanced AV-LLMs, such as Qwen2.5 Omni, exhibit high hallucination rates, achieving only 27.3% and 39.5% accuracy on Q/As related to foreground and background sounds, respectively. With this work, we highlight the need to measure the reliability of multimodal responses, emphasizing that robust evaluation of hallucinations is essential to develop reliable AV-LLMs.

103. S2G-RAG: Structured Sufficiency and Gap Judging for Iterative Retrieval-Augmented QA

Authors: Minghan Li , Junjie Zou , Xinxuan Lv , Chao Zhang , Guodong Zhou
URL: https://arxiv.org/abs/2604.23783
Abstract:

Retrieval-Augmented Generation (RAG) grounds language models in external evidence, but multi-hop question answering remains difficult because iterative pipelines must control what to retrieve next and when the available evidence is adequate. In practice, systems may answer from incomplete evidence chains, or they may accumulate redundant or distractor-heavy text that interferes with later retrieval and reasoning. We propose S2G-RAG (Structured Sufficiency and Gap-judging RAG), an iterative framework with an explicit controller, S2G-Judge. At each turn, S2G-Judge predicts whether the current evidence memory supports answering and, if not, outputs structured gap items that describe the missing information. These gap items are then mapped into the next retrieval query, producing stable multi-turn retrieval trajectories. To reduce noise accumulation, S2G-RAG maintains a sentence-level Evidence Context by extracting a compact set of relevant sentences from retrieved documents. Experiments on TriviaQA, HotpotQA, and 2WikiMultiHopQA show that S2G-RAG improves multi-hop QA performance and robustness under multi-turn retrieval. Furthermore, S2G-RAG can be integrated into existing RAG pipelines as a lightweight component, without modifying the search engine or retraining the generator.

104. The Override Gap: A Magnitude Account of Knowledge Conflict Failure in Hypernetwork-Based Instant LLM Adaptation

Authors: Shuaizhi Cheng , Xiang Shi , Mingwei Li
URL: https://arxiv.org/abs/2604.23750
Abstract:

Hypernetwork-based methods such as Doc-to-LoRA internalize a document into an LLM’s weights in a single forward pass, but they fail systematically on conflicts: when the document contradicts pretraining knowledge, accuracy collapses to 46.4% on the deepest facts. We show the failure is a magnitude problem rather than a representational one. The hypernetwork already targets the right layers, but its adapter margin is approximately constant across documents while the pretrained margin grows with training frequency, so deep conflicts lose by construction. The account predicts that failure should track prior strength: sorting 194 conflicts by the base model’s log-probability on the contradicted fact, baseline accuracy falls from 68% on weak-prior questions to 16% on strong-prior ones, a 52 percentage-point gap. The cure is amplitude. Selective Layer Boosting scales the adapter at its top-norm layers, and Conflict-Aware Internalization triggers boosting only when the base model is confident. Both are training-free; together they raise deep-conflict accuracy from 46.4% to 71.0% on Gemma-2B and from 53.6% to 72.5% on Mistral-7B while preserving novel-knowledge recall, and beat vanilla retrieval-augmented generation on medium conflicts by 18 percentage points despite operating entirely in parameter space. We release KID-Bench, a 489-question benchmark that separates novel recall, cross-knowledge combination, and prior-graded conflicts.

105. SFT-then-RL Outperforms Mixed-Policy Methods for LLM Reasoning

Authors: Alexis Limozin , Eduard Durech , Torsten Hoefler , Imanol Schlag , Valentina Pyatkin
URL: https://arxiv.org/abs/2604.23747
Abstract:

Recent mixed-policy optimization methods for LLM reasoning that interleave or blend supervised and reinforcement learning signals report improvements over the standard SFT-then-RL pipeline. We show that numerous recently published research papers rely on a faulty baseline caused by two distinct bugs: a CPU-offloaded optimizer bug in DeepSpeed that silently drops intermediate micro-batches during gradient accumulation (affecting multiple downstream frameworks including TRL, OpenRLHF and Llama-Factory), and a loss aggregation bug in OpenRLHF that incorrectly weights per-mini-batch losses. Together they suppress SFT performance, with the optimizer bug accounting for most of the gap and the loss aggregation bug contributing a smaller additional effect. Once corrected, the standard SFT-then-RL pipeline surpasses every published mixed-policy method we evaluate by +3.8 points on math benchmarks with Qwen2.5-Math-7B and by +22.2 points with Llama-3.1-8B. Even a truncated variant with just 50 RL steps outperforms mixed-policy methods on math benchmarks while using fewer FLOPs.

106. Zoom In, Reason Out: Efficient Far-field Anomaly Detection in Expressway Surveillance Videos via Focused VLM Reasoning Guided by Bayesian Inference

Authors: Xiaowei Mao , Bowen Sui , Weijie Zhang , Yawen Yang , Shengnan Guo , Shilong Zhao , Jiaqi Lin , Tingrui Wu , Youfang Lin , Huaiyu Wa
URL: https://arxiv.org/abs/2604.23724
Abstract:

Expressway video anomaly detection is essential for safety management. However, identifying anomalies across diverse scenes remains challenging, particularly for far-field targets exhibiting subtle abnormal vehicle motions. While Vision-Language Models (VLMs) demonstrate strong semantic reasoning capabilities, processing global frames causes attention dilution for these far-field objects and incurs prohibitive computational costs. To address these issues, we propose VIBES, an asynchronous collaborative framework utilizing VLMs guided by Bayesian inference. Specifically, to overcome poor generalization across varying expressway environments, we introduce an online Bayesian inference module. This module continuously evaluates vehicle trajectories to dynamically update the probabilistic boundaries of normal driving behaviors, serving as an asynchronous trigger to precisely localize anomalies in space and time. Instead of processing the continuous video stream, the VLM processes only the localized visual regions indicated by the trigger. This targeted visual input prevents attention dilution and enables accurate semantic reasoning. Extensive evaluations demonstrate that VIBES improves detection accuracy for far-field anomalies and reduces computational overhead, achieving high real-time efficiency and explainability while demonstrating generalization across diverse expressway conditions.

107. AIPsy-Affect: A Keyword-Free Clinical Stimulus Battery for Mechanistic Interpretability of Emotion in Language Models

Authors: Michael Keeman
URL: https://arxiv.org/abs/2604.23719
Abstract:

Mechanistic interpretability research on emotion in large language models – linear probing, activation patching, sparse autoencoder (SAE) feature analysis, causal ablation, steering vector extraction – depends on stimuli that contain the words for the emotions they test. When a probe fires on “I am furious”, it is unclear whether the model has detected anger or detected the word “furious”. The two readings have very different consequences for every downstream claim about emotion circuits, features, and interventions. We release AIPsy-Affect, a 480-item clinical stimulus battery that removes the confound at the stimulus level: 192 keyword-free vignettes evoking each of Plutchik’s eight primary emotions through narrative situation alone, 192 matched neutral controls that share characters, setting, length, and surface structure with the affect surgically removed, plus moderate-intensity and discriminant-validity splits. The matched-pair structure supports linear probing, activation patching, SAE feature analysis, causal ablation, and steering vector extraction under a strong methodological guarantee: any internal representation that distinguishes a clinical item from its matched neutral cannot be doing so on the basis of emotion-keyword presence. A three-method NLP defense battery – bag-of-words sentiment, an emotion-category lexicon, and a contextual transformer classifier – confirms the property: bag-of-words methods see only situational vocabulary, and a contextual classifier detects affect (p < 10^-15) but cannot identify the category (5.2% top-1 vs. 82.5% on a keyword-rich control). AIPsy-Affect extends our earlier 96-item battery ( arXiv:2603.22295 ) by a factor of four and is released openly under MIT license.

108. Agri-CPJ: A Training-Free Explainable Framework for Agricultural Pest Diagnosis Using Caption-Prompt-Judge and LLM-as-a-Judge

Authors: Wentao Zhang , Qi Zhang , Mingkun Xu , Mu You , Henghua Shen , Zhongzhi He , Keyan Jin , Derek F. Wong , Tao Fang
URL: https://arxiv.org/abs/2604.23701
Abstract:

Crop disease diagnosis from field photographs faces two recurring problems: models that score well on benchmarks frequently hallucinate species names, and when predictions are correct, the reasoning behind them is typically inaccessible to the practitioner. This paper describes Agri-CPJ (Caption-Prompt-Judge), a training-free few-shot framework in which a large vision-language model first generates a structured morphological caption, iteratively refined through multi-dimensional quality gating, before any diagnostic question is answered. Two candidate responses are then generated from complementary viewpoints, and an LLM judge selects the stronger one based on domain-specific criteria. Caption refinement is the component with the largest individual impact: ablations confirm that skipping it consistently degrades downstream accuracy across both models tested. On CDDMBench, pairing GPT-5-Nano with GPT-5-mini-generated captions yields \textbf{+22.7} pp in disease classification and \textbf{+19.5} points in QA score over no-caption baselines. Evaluated without modification on AgMMU-MCQs, GPT-5-Nano reached 77.84\% and Qwen-VL-Chat reached 64.54\%, placing them at or above most open-source models of comparable scale despite the format shift from open-ended to multiple-choice. The structured caption and judge rationale together constitute a readable audit trail: a practitioner who disagrees with a diagnosis can identify the specific caption observation that was incorrect. Code and data are publicly available this https URL

Authors: Tianyidan Xie , Peiyu Wang , Yuyi Qian , Yuxuan Wang , Rui Ma , Ying Tai , Song Wu , Qian Wang , Lanjun Wang , Zili Yi
URL: https://arxiv.org/abs/2604.23580
Abstract:

Physics-aware symbolic simulation of 3D scenes is critical for robotics, embodied AI, and scientific computing, requiring models to understand natural language descriptions of physical phenomena and translate them into executable simulation environments. While large language models (LLMs) excel at general code generation, they struggle with the semantic gap between physical descriptions and simulation implementation. We introduce PhysCodeBench, the first comprehensive benchmark for evaluating physics-aware symbolic simulation, comprising 700 manually-crafted diverse samples across mechanics, fluid dynamics, and soft-body physics with expert annotations. Our evaluation framework measures both code executability and physical accuracy through automated and visual assessment. Building on this, we propose a Self-Corrective Multi-Agent Refinement Framework (SMRF) with three specialized agents (simulation generator, error corrector, and simulation refiner) that collaborate iteratively with domain-specific validation to produce physically accurate simulations. SMRF achieves 67.7 points overall performance compared to 36.3 points for the best baseline among evaluated SOTA models, representing a 31.4-point improvement. Our analysis demonstrates that error correction is critical for accurate physics-aware symbolic simulation and that specialized multi-agent approaches significantly outperform single-agent methods across the tested physical domains.

110. LLMs Reading the Rhythms of Daily Life: Aligned Understanding for Behavior Prediction and Generation

Authors: Fanjin Meng , Jingtao Ding , Nian Li , Yizhou Sun , Yong Li
URL: https://arxiv.org/abs/2604.23578
Abstract:

Human daily behavior unfolds as complex sequences shaped by intentions, preferences, and context. Effectively modeling these behaviors is crucial for intelligent systems such as personal assistants and recommendation engines. While recent advances in deep learning and behavior pre-training have improved behavior prediction, key challenges remain–particularly in handling long-tail behaviors, enhancing interpretability, and supporting multiple tasks within a unified framework. Large language models (LLMs) offer a promising direction due to their semantic richness, strong interpretability, and generative capabilities. However, the structural and modal differences between behavioral data and natural language limit the direct applicability of LLMs. To address this gap, we propose Behavior Understanding Alignment (BUA), a novel framework that integrates LLMs into human behavior modeling through a structured curriculum learning process. BUA employs sequence embeddings from pretrained behavior models as alignment anchors and guides the LLM through a three-stage curriculum, while a multi-round dialogue setting introduces prediction and generation capabilities. Experiments on two real-world datasets demonstrate that BUA significantly outperforms existing methods in both tasks, highlighting its effectiveness and flexibility in applying LLMs to complex human behavior modeling.

111. CyberCane: Neuro-Symbolic RAG for Privacy-Preserving Phishing Detection with Formal Ontology Reasoning

Authors: Safayat Bin Hakim , Aniqa Afzal , Qi Zhao , Vigna Majmundar , Pawel Sloboda , Houbing Herbert Song
URL: https://arxiv.org/abs/2604.23563
Abstract:

Privacy-critical domains require phishing detection systems that satisfy contradictory constraints: near-zero false positives to prevent workflow disruption, transparent explanations for non-expert staff, strict regulatory compliance prohibiting sensitive data exposure to external APIs, and robustness against AI-generated attacks. Existing rule-based systems are brittle to novel campaigns, while LLM-based detectors violate privacy regulations through unredacted data transmission. We introduce CyberCane, a neuro-symbolic framework integrating deterministic symbolic analysis with privacy-preserving retrieval-augmented generation (RAG). Our dual-phase pipeline applies lightweight symbolic rules to email metadata, then escalates borderline cases to semantic classification via RAG with automated sensitive data redaction and retrieval from a phishing-only corpus. We further introduce PhishOnt, an OWL ontology enabling verifiable attack classification through formal reasoning chains. Evaluation on DataPhish2025 (12.3k emails; mixed human/LLM) and Nazario/SpamAssassin demonstrates a 78.6-point recall gain over symbolic-only detection on AI-generated threats, with precision exceeding 98% and FPR as low as 0.16%. Healthcare deployment projects a 542x ROI; tunable operating points support diverse risk tolerances, with open-source implementation at this https URL .

112. DLM: Unified Decision Language Models for Offline Multi-Agent Sequential Decision Making

Authors: Zhuohui Zhang , Bin Cheng , Bin He
URL: https://arxiv.org/abs/2604.23557
Abstract:

Building scalable and reusable multi-agent decision policies from offline datasets remains a challenge in offline multi-agent reinforcement learning (MARL), as existing methods often rely on fixed observation formats and action spaces that limit generalization. In contrast, large language models (LLMs) offer a flexible modeling interface that can naturally accommodate heterogeneous observations and actions. Motivated by this, we propose the Decision Language Model (DLM), which formulates multi-agent decision making as a dialogue-style sequence prediction problem under the centralized training with decentralized execution paradigm. DLM is trained in two stages: a supervised fine-tuning phase, which leverages dialogue-style datasets for centralized training with inter-agent context and generates executable actions from offline trajectories, followed by a group relative policy optimization phase to enhance robustness to out-of-distribution actions through lightweight reward functions. Experiments on multiple benchmarks show that a unified DLM outperforms strong offline MARL baselines and LLM-based conversational decision-making methods, while demonstrating strong zero-shot generalization to unseen scenarios across tasks.

113. Pref-CTRL: Preference Driven LLM Alignment using Representation Editing

Authors: Imranul Ashrafi , Inigo Jauregi Unanue , Massimo Piccardi
URL: https://arxiv.org/abs/2604.23543
Abstract:

Test-time alignment methods offer a promising alternative to fine-tuning by steering the outputs of large language models (LLMs) at inference time with lightweight interventions on their internal representations. Recently, a prominent and effective approach, RE-Control (Kong et al., 2024), has proposed leveraging an external value function trained over the LLM’s hidden states to guide generation via gradient-based editing. While effective, this method overlooks a key characteristic of alignment tasks, i.e. that they are typically formulated as learning from human preferences between candidate responses. To address this, in this paper we propose a novel preference-based training framework, Pref-CTRL, that uses a multi-objective value function to better reflect the structure of preference data. Our approach has outperformed RE-Control on two benchmark datasets and showed greater generalization on out-of-domain datasets. Our source code is available at this https URL .

114. MTRouter: Cost-Aware Multi-Turn LLM Routing with History-Model Joint Embeddings

Authors: Yiqun Zhang , Hao Li , Zihan Wang , Shi Feng , Xiaocui Yang , Daling Wang , Bo Zhang , Lei Bai , Shuyue Hu
URL: https://arxiv.org/abs/2604.23530
Abstract:

Multi-turn, long-horizon tasks are increasingly common for large language models (LLMs), but solving them typically requires many sequential model invocations, accumulating substantial inference costs. Here, we study cost-aware multi-turn LLM routing: selecting which model to invoke at each turn from a model pool, given a fixed cost budget. We propose MTRouter, which encodes the interaction history and candidate models into joint history-model embeddings, and learns an outcome estimator from logged trajectories to predict turn-level model utility. Experiments show that MTRouter improves the performance-cost trade-off: on ScienceWorld, it surpasses GPT-5 while reducing total cost by 58.7%; on Humanity’s Last Exam (HLE), it achieves competitive accuracy while reducing total cost by 43.4% relative to GPT-5, and these gains even carry over to held-out tasks. Further analyses reveal several mechanisms underlying its effectiveness: relative to prior multi-turn routers, MTRouter makes fewer model switches, is more tolerant to transient errors, and exhibits emergent specialization across models. Code: this https URL

Authors: Khouloud Gaaloul , Zaid Ghazal , Madhu Latha Pulimi , Sam Emmanuel Kathiravan
URL: https://arxiv.org/abs/2604.23523
Abstract:

Safety specifications in cyber-physical systems (CPS) capture the operational conditions the system must satisfy to operate safely within its intended environment. As operating environments evolve, operational rules must be continuously refined to preserve consistency with observed system behavior during simulation-based verification and validation. Revising inconsistent rules is challenging because the changes must remain syntactically correct under a domain-specific grammar. Language-in-the-loop refinement further raises safety concerns beyond syntactic violations, as it can produce semantically unjustified refinements that overfit to the observed outcomes. We introduce a framework that combines counterfactual reasoning with a grammar-constrained refinement loop to refine operational rules, aligning them with the observed system behavior. Applied to an autonomous driving control system, our approach successfully resolved the inconsistencies in an operational rule inferred by a conventional baseline while remaining grammar compliant. An empirical large language model (LLM) study further revealed model-dependent refinement quality and safety lessons, which motivate rigorous grammar enforcement, stronger semantic validation, and broader evaluation in future work.

116. Uncertainty Propagation in LLM-Based Systems

Authors: Boming Xia , Liming Zhu , Erdun Gao , Qinghua Lu , Minhui Xue , Dino Sejdinovic
URL: https://arxiv.org/abs/2604.23505
Abstract:

Uncertainty in large language model (LLM)-based systems is often studied at the level of a single model output, yet deployed LLM applications are compound systems in which uncertainty is transformed and reused across model internals, workflow stages, component boundaries, persistent state, and human or organisational processes. Without principled treatment of how uncertainty is carried and reused across these boundaries, early errors can propagate and compound in ways that are difficult to detect and govern. This paper develops a systems-level account of uncertainty propagation. It introduces a conceptual framing for characterising propagated uncertainty signals, presents a structured taxonomy spanning intra-model (P1), system-level (P2), and socio-technical (P3) propagation mechanisms, synthesises cross-cutting engineering insights, and identifies five open research challenges.

117. Hybrid JIT-CUDA Graph Optimization for Low-Latency Large Language Model Inference

Authors: Divakar Kumar Yadav , Tian Zhao
URL: https://arxiv.org/abs/2604.23467
Abstract:

Large Language Models (LLMs) have achieved strong performance across natural language and multimodal tasks, yet their practical deployment remains constrained by inference latency and kernel launch overhead, particularly in interactive, short-sequence settings. This paper presents a hybrid runtime framework that combines Just-In-Time (JIT) compilation with CUDA Graph execution to reduce launch overhead while preserving runtime flexibility during autoregressive decoding. The framework partitions transformer inference into static components executed via CUDA Graph replay and dynamic components handled through JIT-compiled kernels, enabling asynchronous graph capture and reuse across decoding steps. We evaluate the proposed approach on LLaMA-2 7B using single-GPU, batch-size-one inference across prompt lengths from 10 to 500 tokens. Experimental results show that the hybrid runtime reduces Time-to-First-Token (TTFT) by up to 66.0% and achieves lower P99 latency compared with TensorRT-LLM in this regime. These results indicate that hybrid JIT-CUDA Graph execution can effectively reduce inference latency and variance for short-sequence LLM workloads, making it a practical optimization strategy for latency-sensitive AI applications.

118. Evaluating CUDA Tile for AI Workloads on Hopper and Blackwell GPUs

Authors: Divakar Kumar Yadav , Tian Zhao , Deepak Kumar
URL: https://arxiv.org/abs/2604.23466
Abstract:

NVIDIA’s CUDA Tile (CuTile) introduces a Python-based, tile-centric abstraction for GPU kernel development that aims to simplify programming while retaining Tensor Core and Tensor Memory Accelerator (TMA) efficiency on modern GPUs. We present the first independent, cross-architecture evaluation of CuTile against established approaches such as cuBLAS, Triton, WMMA, and raw SIMT on three NVIDIA GPUs spanning Hopper and Blackwell: H100 NVL, B200, and RTX PRO 6000 Blackwell Server Edition. We benchmark representative AI workloads, including GEMM, fused multi-head attention, and end-to-end LLM inference in BF16/FP16 precision, to assess both performance and portability. Our results show that CuTile effectiveness is strongly workload- and architecture-dependent. On datacenter-class Blackwell (B200), CuTile achieves up to 1007 TFLOP/s for fused attention, outperforming FlashAttention-2 by 2.5x while requiring only 60 lines of Python kernel code. For GEMM, CuTile reaches 52-79% of cuBLAS performance in 22 lines of code (versus 123 for WMMA), making it a practical replacement for hand-written CUDA kernels but not yet for vendor-optimized libraries. However, the same CuTile attention kernel achieves only 53% of FlashAttention-2 throughput on RTX PRO 6000 (sm_120), exposing significant cross-architecture optimization gaps. In contrast, Triton sustains 62-101% of cuBLAS performance across all tested platforms without architecture-specific tuning, demonstrating substantially stronger portability.

119. AI Safety Training Can be Clinically Harmful

Authors: Suhas BN , Andrew M. Sherrill , Rosa I. Arriaga , Chris W. Wiese , Saeed Abdullah
URL: https://arxiv.org/abs/2604.23445
Abstract:

Large language models are being deployed as mental health support agents at scale, yet only 16% of LLM-based chatbot interventions have undergone rigorous clinical efficacy testing, and simulations reveal psychological deterioration in over one-third of cases. We evaluate four generative models on 250 Prolonged Exposure (PE) therapy scenarios and 146 CBT cognitive restructuring exercises (plus 29 severity-escalated variants), scored by a three-judge LLM panel. All models scored near-perfectly on surface acknowledgment (~0.91-1.00) while therapeutic appropriateness collapsed to 0.22-0.33 at the highest severity for three of four models, with protocol fidelity reaching zero for two. Under CBT severity escalation, one model’s task completeness dropped from 92% to 71% while the frontier model’s safety-interference score fell from 0.99 to 0.61. We identify a systematic, modality-spanning failure: RLHF safety alignment disrupts the therapeutic mechanism of action by grounding patients during imaginal exposure, offering false reassurance, inserting crisis resources into controlled exercises, and refusing to challenge distorted cognitions mentioning self-harm in PE; and through task abandonment or safety-preamble insertion during CBT cognitive restructuring. These findings motivate a five-axis evaluation framework (protocol fidelity, hallucination risk, behavioral consistency, crisis safety, demographic robustness), mapped onto FDA SaMD and EU AI Act requirements. We argue that no AI mental health system should proceed to deployment without passing multi-axis evaluation across all five dimensions.

120. Automating Categorization of Scientific Texts with In-Context Learning and Prompt-Chaining in Large Language Models

Authors: Gautam Kishore Shahi , Oliver Hummel
URL: https://arxiv.org/abs/2604.23430
Abstract:

The relentless expansion of scientific literature presents significant challenges for navigation and knowledge discovery. Within Research Information Retrieval, established tasks such as text summarization and classification remain crucial for enabling researchers and practitioners to effectively navigate this vast landscape, so that efforts have increasingly been focused on developing advanced research information systems. These systems aim not only to provide standard keyword-based search functionalities but also to incorporate capabilities for automatic content categorization within knowledge-intensive organizations across academia and industry. This study systematically evaluates the performance of off-the-shelf Large Language Models (LLMs) in analyzing scientific texts according to a given classification scheme. We utilized the hierarchical ORKG taxonomy as a classification framework, employing the FORC dataset as ground truth. We investigated the effectiveness of advanced prompt engineering strategies, namely In-Context Learning (ICL) and Prompt Chaining, and experimentally explored the influence of the LLMs’ temperature hyperparameter on classification accuracy. Our experiments demonstrate that Prompt Chaining yields superior classification accuracy compared to pure ICL, particularly when applied to the nested structure of the ORKG taxonomy. LLMs with prompt chaining outperform the state-of-the-art models for domain (1st level) prediction and show even better performance for subject (2nd level) prediction compared to the older BERT model. However, LLMs are not yet able to perform well in classifying the topic (3rd level) of research areas based on this specific hierarchical taxonomy, as they only reach about 50% accuracy even with prompt chaining.

121. PushupBench: Your VLM is not good at counting pushups

Authors: Shengzhi Li , Jiarun Chen , Karun Sharma , Jiaqi Su , Shichao Pei
URL: https://arxiv.org/abs/2604.23407
Abstract:

Large vision-language models (VLMs) can recognize \textit{what} happens in video but fail to count \textit{how many} times. We introduce \textbf{PushupBench}, 446 long-form clips (avg. 36.7s) for evaluating repetition counting. The best frontier model achieves 42.1\% exact accuracy; open-source 4B models score $\sim$6\%, matching supervised baselines. We show that accuracy alone misleads – weaker models exploit the modal count rather than reason temporally. Fine-tuning on counting with 1k samples transfers to general video understanding: MVBench (+2.15), PerceptionTest (+1.88), TVBench (+4.54), suggesting counting is a proxy for broader temporal this http URL incorporated in \texttt{lmms-eval} ( this https URL ) and hosted on ( this http URL )

122. An Empirical Evaluation of Locally Deployed LLMs for Bug Detection in Python Code

Authors: Jelena Ilić Vulićević
URL: https://arxiv.org/abs/2604.23361
Abstract:

Large language models (LLMs) have demonstrated strong performance on a wide range of software engineering tasks, including code generation and analysis. However, most prior work relies on cloud-based models or specialized hardware, limiting practical applicability in privacy-sensitive or resource-constrained environments. In this paper, we present a systematic empirical evaluation of two locally deployed LLMs, LLaMA 3.2 and Mistral, for real-world Python bug detection using the BugsInPy benchmark. We evaluate 349 bugs across 17 projects using a zero-shot prompting approach at the function level and an automated keyword-based evaluation framework. Our results show that locally executed models achieve accuracy between 43% and 45%, while producing a large proportion of partially correct responses that identify problematic code regions without pinpointing the exact fix. Performance varies significantly across projects, highlighting the importance of codebase characteristics. The results demonstrate that local models can identify a meaningful share of bugs, though precise localization remains difficult for locally executed LLMs, particularly when handling complex and context dependent bugs in realistic development scenarios.

123. EmoTrans: A Benchmark for Understanding, Reasoning, and Predicting Emotion Transitions in Multimodal LLMs

Authors: He Hu , Tengjin Weng , Zebang Cheng , Yu Wang , Jiachen Luo , Björn Schuller , Zheng Lian , Laizhong Cui
URL: https://arxiv.org/abs/2604.23348
Abstract:

Recent multimodal large language models (MLLMs) have shown strong capabilities in perception, reasoning, and generation, and are increasingly used in applications such as social robots and human-computer interaction, where understanding human emotions is essential. However, existing benchmarks mainly formulate emotion understanding as a static recognition problem, leaving it largely unclear whether current MLLMs can understand emotion as a dynamic process that evolves, shifts between states, and unfolds across diverse social contexts. To bridge this gap, we present EmoTrans, a benchmark for evaluating emotion dynamics understanding in multimodal videos. EmoTrans contains 1,000 carefully collected and manually annotated video clips, covering 12 real-world scenarios, and further provides over 3,000 task-specific question-answer (QA) pairs for fine-grained evaluation. The benchmark introduces four tasks, namely Emotion Change Detection (ECD), Emotion State Identification (ESI), Emotion Transition Reasoning (ETR), and Next Emotion Prediction (NEP), forming a progressive evaluation framework from coarse-grained detection to deeper reasoning and prediction. We conduct a comprehensive evaluation of 18 state-of-the-art MLLMs on EmoTrans and obtain two main findings. First, although current MLLMs show relatively stronger performance on coarse-grained emotion change detection, they still struggle with fine-grained emotion dynamics modeling. Second, socially complex settings, especially multi-person scenarios, remain substantially challenging, while reasoning-oriented variants do not consistently yield clear improvements. To facilitate future research, we publicly release the benchmark, evaluation protocol, and code at this https URL .

124. Evaluating Jailbreaking Vulnerabilities in LLMs Deployed as Assistants for Smart Grid Operations: A Benchmark Against NERC Standards

Authors: Taha Hammadia , Lucas Rea , Ahmad Mohammad Saber , Amr Youssef , Deepa Kundur
URL: https://arxiv.org/abs/2604.23341
Abstract:

The deployment of Large Language Models (LLMs) as assistants in electric grid operations promises to streamline compliance and decision-making but exposes new vulnerabilities to prompt-based adversarial attacks. This paper evaluates the risk of jailbreaking LLMs, i.e., circumventing safety alignments to produce outputs violating regulatory standards, assuming threats from authorized users, such as operators, who craft malicious prompts to elicit non-compliant guidance. Three state-of-the-art LLMs (OpenAI’s GPT-4o mini, Google’s Gemini 2.0 Flash-Lite, and Anthropic’s Claude 3.5 Haiku) were tested against Baseline, BitBypass, and DeepInception jailbreaking methods across scenarios derived from nine NERC Reliability Standards (EOP, TOP, and CIP). In the initial broad experiment, the overall Attack Success Rate (ASR) was 33.1%, with DeepInception proving most effective at 63.17% ASR. Claude 3.5 Haiku exhibited complete resistance (0% ASR), while Gemini 2.0 Flash-Lite was most vulnerable (55.04% ASR) and GPT-4o mini moderately susceptible (44.34% ASR). A follow-up experiment refining malicious wording in Baseline and BitBypass attacks yielded a 30.6% ASR, confirming that subtle prompt adjustments can enhance simpler methods’ efficacy.

Authors: Yahui Li , Yinfeng Yu , Liejun Wang , Shengjie Shen
URL: https://arxiv.org/abs/2604.23325
Abstract:

Emotionally talking head video generation aims to generate expressive portrait videos with accurate lip synchronization and emotional facial expressions. Current methods rely on simple emotional labels, leading to insufficient semantic information. While introducing high-level semantics enhances expressiveness, it easily causes lip-sync degradation. Furthermore, mainstream generation methods struggle to balance computational efficiency and global motion awareness in long videos and suffer from poor temporal coherence. Therefore, we propose an \textbf{E}motion-\textbf{A}ware \textbf{D}iffusion model-based \textbf{Net}work, called \textbf{EAD-Net}. We introduce SyncNet supervision and Temporal Representation Alignment (TREPA) to mitigate lip-sync degradation caused by multi-modal fusion. To model complex spatio-temporal dependencies in long video sequences, we propose a Spatio-Temporal Directional Attention (STDA) mechanism that captures global motion patterns through strip attention. Additionally, we design a Temporal Frame graph Reasoning Module (TFRM) to explicitly model temporal coherence between video frames through graph structure learning. To enhance emotional semantic control, a large language model is employed to extract textual descriptions from real videos, serving as high-level semantic guidance. Experiments on the HDTF and MEAD datasets demonstrate that our method outperforms existing methods in terms of lip-sync accuracy, temporal consistency, and emotional accuracy.

126. $\mathcal{S}^2$IT: Stepwise Syntax Integration Tuning for Large Language Models in Aspect Sentiment Quad Prediction

Authors: Bingfeng Chen , Chenjie Qiu , Yifeng Xie , Boyan Xu , Ruichu Cai , Zhifeng Hao
URL: https://arxiv.org/abs/2604.23296
Abstract:

Aspect Sentiment Quad Prediction (ASQP) has seen significant advancements, largely driven by the powerful semantic understanding and generative capabilities of large language models (LLMs). However, while syntactic structure information has been proven effective in previous extractive paradigms, it remains underutilized in the generative paradigm of LLMs due to their limited reasoning capabilities. In this paper, we propose S^2IT, a novel Stepwise Syntax Integration Tuning framework that progressively integrates syntactic structure knowledge into LLMs through a multi-step tuning process. The training process is divided into three steps. S^2IT decomposes the quadruple generation task into two stages: 1) Global Syntax-guided Extraction and 2) Local Syntax-guided Classification, integrating both global and local syntactic structure information. Finally, Fine-grained Structural Tuning enhances the model’s understanding of syntactic structures through the prediction of element links and node classification. Experiments demonstrate that S^2IT significantly improves state-of-the-art performance across multiple datasets. Our implementation will be open-sourced at this https URL .

127. Au-M-ol: A Unified Model for Medical Audio and Language Understanding

Authors: Meizhu Liu , Nistha Mitra , Paul Li , Amine Abdaoui , Adam Ledyard , Tao Sheng
URL: https://arxiv.org/abs/2604.23284
Abstract:

In this work, we present Au-M-ol, a novel multimodal architecture that extends Large Language Models (LLMs) with audio processing. It is designed to improve performance on clinically relevant tasks such as Automatic Speech Recognition (ASR). Au-M-ol has three main components: (1) an audio encoder that extracts rich acoustic features from medical speech, (2) an adaptation layer that maps audio features into the LLM input space, and (3) a pretrained LLM that performs transcription and clinical language understanding. This design allows the model to interpret spoken medical content directly, improving both accuracy and robustness. In experiments, Au-M-ol reduces Word Error Rate (WER) by 56\% compared to state-of-the-art baselines on medical transcription tasks. The model also performs well in challenging conditions, including noisy environments, domain-specific terminology, and speaker variability. These results suggest that Au-M-ol is a strong candidate for real-world clinical applications, where reliable and context-aware audio understanding is essential.

128. From Similarity to Structure: Training-free LLM Context Compression with Hybrid Graph Priors

Authors: Yitian Zhou , Chaoning Zhang , Jiaquan Zhang , Zhenzhen Huang , Jinyu Guo , Sung-Ho Bae , Lik-Hang Lee , Caiyan Qin , Yang Yang
URL: https://arxiv.org/abs/2604.23277
Abstract:

Long-context large language models remain computationally expensive to run and often fail to reliably process very long inputs, which makes context compression an important component of many systems. Existing compression approaches typically rely on trained compressors, dense retrieval-style selection, or heuristic trimming, and they often struggle to jointly preserve task relevance, topic coverage, and cross-sentence coherence under a strict token budget. To address this, we propose a training-free and model-agnostic compression framework that selects a compact set of sentences guided by structural graph priors. Our method constructs a sparse hybrid sentence graph that combines mutual k-NN semantic edges with short-range sequential edges, extracts a topic skeleton via clustering, and ranks sentences using an interpretable score that integrates task relevance, cluster representativeness, bridge centrality, and a cycle coverage cue. A budgeted greedy selection with redundancy suppression then produces a readable compressed context in original order. Experimental results on four datasets show that our approach is competitive with strong extractive and abstractive baselines, demonstrating larger gains on long-document benchmarks.

129. Lightweight and Production-Ready PDF Visual Element Parsing

Authors: Meizhu Liu , Yassi Abbasi , Matthew Rowe , Michael Avendi , Paul Li
URL: https://arxiv.org/abs/2604.23276
Abstract:

PDF documents contain critical visual elements such as figures, tables, and forms whose accurate extraction is essential for document understanding and multimodal retrieval-augmented generation (RAG). Existing PDF parsers often miss complex visuals, extract non-informative artifacts (e.g., watermarks, logos), produce fragmented elements, and fail to reliably associate captions with their corresponding elements, which degrades downstream retrieval and question answering. We present a lightweight and production level PDF parsing framework that can accurately detect visual elements and associates captions using a combination of spatial heuristics, layout analysis, and semantic similarity. On popular benchmark datasets and internal product data, the proposed solution achieves $\geq96\%$ visual element detection accuracy and $93\%$ caption association accuracy. When used as a preprocessing step for multimodal RAG, it significantly outperforms state-of-the-art parsers and large vision-language models on both internal data and the MMDocRAG benchmark, while reducing latency by over $2\times$. We have deployed the proposed system in challenging production environment.

130. Small Language Model Helps Resolve Semantic Ambiguity of LLM Prompt

Authors: Zhenzhen Huang , Chaoning Zhang , Fachrina Dewi Puspitasari , Jiaquan Zhang , Yitian Zhou , Shuxu Chen , Yang Yang
URL: https://arxiv.org/abs/2604.23263
Abstract:

Large language models (LLMs) are increasingly utilized in various complex reasoning tasks due to their excellent instruction following capability. However, the model’s performance is highly dependent on the open-ended characteristics of the users’ input prompt. Natural prompts often do not follow proper syntactic rules, which creates ambiguous queries that yield multiple interpretations. Such ambiguous prompts confuse the model in choosing the correct reasoning paths to answer questions. Prior works address this challenge by applying query editing during the LLM inference process without explicitly solving the root cause of the ambiguity. To address this limitation, we propose a pre-inference prompt optimization mechanism via explicit prompt disambiguation. Particularly, we identify semantic risks in the prompt, check their multi-perspective consistency, and resolve any semantic conflicts that arise. Finally, we organize the resolved ambiguities in a logically structured manner as a clean input to the LLM. By explicitly resolving semantic ambiguity, our method can produce a more focused attention distribution to the semantically essential tokens. We also leverage small language models (SLMs) as the main executor of prompt disambiguation to benefit from their efficient computation. Through comprehensive experiments on multiple benchmarks, we demonstrate that our method improves reasoning performance by 2.5 points at a cost of only $0.02. Our study promotes explicit prompt disambiguation as an effective prompt optimization method without disturbing the internal mechanism of LLM inference.

131. Knowledge Lever Risk Management for Software Engineering: A Stochastic Framework for Mitigating Knowledge Loss

Authors: Mark Chua , Samuel Ajila
URL: https://arxiv.org/abs/2604.23257
Abstract:

Software engineering (SE) organizations operate in a knowledge-intensive domain where critical assets – architectural expertise, design rationale, and system intuition – are overwhelmingly tacit and volatile. The departure of key contributors or the decay of undocumented decisions can severely impair project velocity and software quality. While conventional SE risk management optimized for schedule and budget is common, the intangible knowledge risks that determine project success remain under-represented. The goal of this research work is to propose and evaluate the Knowledge Lever Risk Management (KLRM) Framework, designed specifically for the software development lifecycle. The primary objectives are to: (1) recast intangible knowledge assets as active mechanisms for risk mitigation (Knowledge Levers); (2) integrate these levers into a structured four-phase architecture (Audit, Alignment, Activation, Assurance); and (3) provide a formal stochastic model to quantify the impact of lever activation on project knowledge capital. We detail the application of these levers through software-specific practices such as pair programming, architectural decision records (ADRs), and LLM-assisted development. Stochastic Monte Carlo simulations demonstrate that full lever activation increases expected knowledge capital by 63.8\% and virtually eliminates knowledge crisis probability. Our research shows that knowledge lever activation improves alignment across the project management iron triangle (scope, time, cost) by reducing rework and rediscovery costs.

132. Scalable LLM-based Coding of Dialogue in Healthcare Simulation: Balancing Coding Performance, Processing Time, and Environmental Impact

Authors: Kiyoshige Garces , Gloria Milena Fernandez-Nieto , Linxuan Zhao , Sachini Samaraweera , Dragan Gasevic , Roberto Martinez-Maldonado , Vanessa Echeverria
URL: https://arxiv.org/abs/2604.23255
Abstract:

Research shows that dialogue, the interactive process through which participants articulate their thinking, plays a central role in constructing shared understanding, coordinating action, and shaping learning outcomes in teams. Analysing dialogue content has been central to advancing team learning theory and informing the design of computer-supported collaborative learning environments, yet this progress has depended on labour-intensive qualitative coding. LLMs offer new possibilities for automating and enhancing the dialogue layer within emerging multimodal learning analytics approaches, with recent studies showing that they can approximate human coding through few-shot prompting. However, prior work has focused on replicating human coding accuracy for research purposes, rather than addressing a more educationally consequential question: how can we design prompts that allow an LLM to label team dialogue accurately and fast enough to be useful in real settings, such as in-person healthcare simulations, where results must be returned quickly and computational cost and sustainability also matter? This paper investigates how prompt design and batching strategies can be optimised to balance coding accuracy, processing time, and environmental impact in team-based healthcare simulation debriefing. Using a dataset of 11,647 utterances coded across 6 dialogue constructs, we compared 4 prompt designs across varying batch sizes, evaluating coding performance, processing time, and energy consumption, as well as the trade-offs between these metrics. Results indicate that increasing batch size improves speed and reduces energy use, but negatively impacts coding performance. Beyond demonstrating the feasibility of LLM-based qualitative analysis, this study offers practical guidance for scaling dialogue analytics in contexts where timeliness, privacy, and sustainability are critical.

133. AI-Assisted Code Review as a Scaffold for Code Quality and Self-Regulated Learning: An Experience Report

Authors: Eduardo Oliveira , Michael Fu , Patanamon Thongtanunam , Sonsoles López-Pernas , Mohammed Saqr
URL: https://arxiv.org/abs/2604.23251
Abstract:

Code review is central to software engineering education but hard to scale in capstone projects due to tight deadlines, uneven peer feedback, and limited prior experience. We investigate an LLM-as-reviewer integrated directly into GitHub pull requests (human-in-the-loop) across two cohorts (more than 100 students, 2023–2024). Using a mixed-methods design – GitHub data, reflective reports, and a targeted survey – we examine engagement and responsiveness as behavioral indicators of self-regulated learning processes. Quantitatively, the 2024 cohort produced more iterative activity (1176 vs. 581 PRs), while technical issues observed in 2023 (227 failed AI attempts) dropped to zero after tool and instructional refinements. Despite different adoption levels (93\% vs. 50\% of teams using the tool), responsiveness was stable: 32\% (2023) and 33\% (2024) of successfully AI-reviewed PRs were followed by subsequent commits on the same PR. Qualitatively, students used the LLM’s structured comments to focus reviews and discuss code quality, while guidance reduced over-reliance. We contribute: (i) an in-workflow design for an AI reviewer that scaffolds learning while mitigating cognitive offloading; (ii) a repeated cross sectional comparison across two cohorts in authentic settings; (iii) a mixed-methods analysis combining objective GitHub metrics with student self-reports; and (iv) evidence-based pedagogical recommendations for responsible, student-led AI-assisted review.

Authors: Yihan Wang , Lei Li , Yao Lai , Jing Wang , Yan Lu
URL: https://arxiv.org/abs/2604.23195
Abstract:

Analog circuit design relies heavily on reusing existing intellectual property (IP), yet searching across heterogeneous representations such as SPICE netlists, schematics, and functional descriptions remains challenging. Existing methods are largely limited to exact matching within a single modality, failing to capture cross-modal semantic relationships. To bridge this gap, we present AnalogRetriever, a unified tri-modal retrieval framework for analog circuit search. We first build a high-quality dataset on top of Masala-CHAI through a two-stage repair pipeline that raises the netlist compile rate from 22\% to 100\%. Built on this foundation, AnalogRetriever encodes schematics and descriptions with a vision-language model and netlists with a port-aware relational graph convolutional network, mapping all three modalities into a shared embedding space via curriculum contrastive learning. Experiments show that AnalogRetriever achieves an average Recall@1 of 75.2\% across all six cross-modal retrieval directions, significantly outperforming existing baselines. When integrated into the AnalogCoder agentic framework as a retrieval-augmented generation module, it consistently improves functional pass rates and enables previously unsolved tasks to be completed. Our code and dataset will be released.

135. Scaling Multi-Node Mixture-of-Experts Inference Using Expert Activation Patterns

Authors: Abhimanyu Bambhaniya , Geonhwa Jeong , Jason Park , Jiecao Yu , Jaewon Lee , Pengchao Wang , Changkyu Kim , Chunqiang Tang , Tushar Krishna
URL: https://arxiv.org/abs/2604.23150
Abstract:

Most recent state-of-the-art (SOTA) large language models (LLMs) use Mixture-of-Experts (MoE) architectures to scale model capacity without proportional per-token compute, enabling higher-quality outputs at manageable serving costs. However, MoE inference at scale is fundamentally bottlenecked by expert load imbalance and inefficient token routing, especially in multi-node deployments where tokens are not guaranteed to be routed to local experts, resulting in significant inter-node all-to-all communication overhead. To systematically characterize these challenges, we profile SOTA open-source MoE models, including Llama 4 Maverick, DeepSeek V3-671B, and Qwen3-230B-A22B, on various datasets and collected over 100k real expert activation traces. Upon studying the expert activation patterns, we uncover various persistent properties across all the frontier MoE models: variable expert load imbalance, domain-specific expert activation where expert popularity shifts across task families (code, math, chat, general), and a strong correlation between prefill and decode expert activations. Motivated by these findings, we propose workload-aware micro-batch grouping and an expert placement strategy to maximize token locality to the destination expert, thereby reducing inter-node communication. Across models and datasets, these optimizations help reduce all2all communication data up to 20, resulting in lower MoE decode latency and better accelerator utilization.

Authors: Tianlong Yu , Yang Yang , Xiao Luo , Lihong Liu , Fudu Xing , Zui Tao , Kailong Wang , Gaoyang Liu , Ting Bi
URL: https://arxiv.org/abs/2604.23141
Abstract:

Emerging AR-LLM-based Social Engineering attack (e.g., SEAR) is at the edge of posing great threats to real-world social life. In such AR-LLM-SE attack, the attacker can leverage AR (Augmented Reality) glass to capture the image and vocal information of the target, using the LLM to identify the target and generate the social profile, using the LLM agents to apply social engineering strategies for conversation suggestion to win the target trust and perform phishing afterwards. Current defensive approaches, such as role-based access control or data flow tracking, are not directly applicable to the convergent AR-LLM ecosystem (considering embedded AR device and opaque LLM inference), leaving an emerging and potent social engineering threat that existing privacy paradigms are ill-equipped to address. This necessitates a shift beyond solely human-centric measures like legislation and user education toward enforceable vendor policies and platform-level restrictions. Realizing this vision, however, faces significant technical challenges: securing resource-constrained AR-embedded devices, implementing fine-grained access control within opaque LLM inferences, and governing adaptive interactive agents. To address these challenges, we present UNSEEN, a coordinated cross-stack defense that combines an AR ACL (Access Control Layer) for identity-gated sensing, F-RMU-based LLM unlearning for sensitive profile suppression, and runtime agent guardrails for adaptive interaction control. We evaluate UNSEEN in an IRB-approved user study with 60 participants and a dataset of 360 annotated conversations across realistic social scenarios.

137. Mechanistic Steering of LLMs Reveals Layer-wise Feature Vulnerabilities in Adversarial Settings

Authors: Nilanjana Das , Manas Gaur
URL: https://arxiv.org/abs/2604.23130
Abstract:

Large language models (LLMs) can still be jailbroken into producing harmful outputs despite safety alignment. Existing attacks show this vulnerability, but not the internal mechanisms that cause it. This study asks whether jailbreak success is driven by identifiable internal features rather than prompts alone. We propose a three-stage pipeline for Gemma-2-2B using the BeaverTails dataset. First, we extract concept-aligned tokens from adversarial responses via subspace similarity. Second, we apply three feature-grouping strategies (cluster, hierarchical-linkage, and single-token-driven) to identify SAE feature subgroups for the aligned tokens across all 26 model layers. Third, we steer the model by amplifying the top features from each identified subgroup and measure the change in harmfulness score using a standardized LLM-judge scoring protocol. In all three approaches, the features in the layers [16-25] were relatively more vulnerable to steering. All three methods confirmed that mid to later layer feature subgroups are more responsible for unsafe outputs. These results provide evidence that the jailbreak vulnerability in Gemma-2-2B is localized to feature subgroups of mid to later layers, suggesting that targeted feature-level interventions may offer a more principled path to adversarial robustness than current prompt-level defenses.

138. MindTrellis: Co-Creating Knowledge Structures with AI through Interactive Visual Exploration

Authors: Xiang Li , Cara Li , Emily Kuang , Can Liu , Jian Zhao
URL: https://arxiv.org/abs/2604.23129
Abstract:

Knowledge workers face increasing challenges in synthesizing information from multiple documents into structured conceptual understanding. This process is inherently iterative: users explore content, identify relationships between concepts, and continuously reorganize their mental models. However, current approaches offer limited support. LLM-based systems let users query information but not shape how knowledge is organized; manual tools like mind maps support structure creation but lack intelligent assistance. This leaves an open opportunity: supporting collaborative construction where users and AI jointly develop an evolving knowledge representation. We present MindTrellis, an interactive visual system where users and AI collaboratively build a dynamic knowledge graph. Users can query the graph to retrieve document-grounded information, and contribute by introducing new concepts, modifying relationships, and reorganizing the hierarchy to reflect their developing understanding. In a user study where 12 participants created slide decks, MindTrellis outperformed retrieval-only baselines in knowledge organization and cognitive load, as measured by expert ratings of content coverage and structural quality.

139. ArgRE: Formal Argumentation for Conflict Resolution in Multi-Agent Requirements Negotiation

Authors: Haowei Cheng , Milhan Kim , Chong Liu , Teeradaj Racharak , Truong Vinh Truong Duy , Phan Thi Huyen Thanh , Jialong Li , Naoyasu Ubayashi , Hironori Washizaki
URL: https://arxiv.org/abs/2604.23124
Abstract:

As software systems grow in complexity, they must satisfy an increasing number of competing quality attributes, making it essential to balance them in a principled manner – for example, a safety requirement for sensor-fusion verification may conflict with a tight planning-cycle budget. Multi-agent large language model frameworks support this balancing process by assigning specialized agents to different objectives. However, their conflict resolution is typically heuristic. Requirements are aggregated implicitly without explicit acceptance or rejection, limiting auditability in regulated domains. We present ArgRE, a multi-agent requirements negotiation system that embeds Dung-style abstract argumentation into the negotiation stage. Each proposal, critique, and refinement is modeled as an argument, conflicts are represented as directed attack relations, and the accepted set of arguments is computed under grounded and preferred semantics. The pipeline further integrates KAOS goal modeling, multi-layer verification, and standards-oriented artifact generation. Evaluation across five case studies spanning safety-critical, financial, and information-system domains shows that ArgRE provides argument-level traceability absent from existing frameworks. Independent evaluators rated its decision justifications significantly higher than those of heuristic synthesis (4.32 vs. 3.07, p < 0.001), indicating improved auditability, while semantic intent preservation remains comparable (94.9% BERTScore F1) and compliance coverage reaches 84.7% versus 47.6%–47.8% for baselines. Structural analysis further confirms that the default pairwise protocol yields acyclic graphs in which grounded and preferred semantics coincide, whereas cross-pair arbitration introduces controlled cyclicity, leading to predictable divergence between the two semantics.

140. Mixture of Heterogeneous Grouped Experts for Language Modeling

Authors: Zhicheng Ma , Xiang Liu , Zhaoxiang Liu , Ning Wang , Yi Shen , Kai Wang , Shuming Shi , Shiguo Lian
URL: https://arxiv.org/abs/2604.23108
Abstract:

Large Language Models (LLMs) based on Mixture-of-Experts (MoE) are pivotal in industrial applications for their ability to scale performance efficiently. However, standard MoEs enforce uniform expert sizes,creating a rigidity that fails to align computational costs with varying token-level complexity. While heterogeneous expert architectures attempt to address this by diversifying expert sizes, they often suffer from significant system-level challenges, specifically unbalanced GPU utilization and inefficient parameter utilization, which hinder practical deployment. To bridge the gap between theoretical heterogeneity and robust industrial application, we propose Mixture of Heterogeneous Grouped Experts (MoHGE) which introduces a two-level routing mechanism to enable flexible, resource-aware expert combinations. To optimize inference efficiency, we propose a Group-Wise Auxiliary Loss, which dynamically steers tokens to the most parameter-efficient expert groups based on task difficulty. To address the critical deployment challenge of GPU load balancing, we introduce an All-size Group-decoupling Allocation strategy coupled with an Intra-Group Experts Auxiliary Loss. These mechanisms collectively ensure uniform computation distribution across GPUs. Extensive evaluations demonstrate that MoHGE matches the performance of MoE architectures while reducing the total parameters by approximately 20% and maintaining balanced GPU utilization. Our work establishes a scalable paradigm for resource-efficient MoE design, offering a practical solution for optimizing inference costs in real-world scenarios.

141. No Test Cases, No Problem: Distillation-Driven Code Generation for Scientific Workflows

Authors: Siddeshwar Raghavan , Tanwi Mallick
URL: https://arxiv.org/abs/2604.23106
Abstract:

Existing multi-agent Large Language Model (LLM) frameworks for code generation typically use execution feedback and improve iteratively using Input/Output (I/O) test cases. However, this does not work for scientific workflows, where I/O test cases do not exist, and generating them requires solving the very problem at hand. To address this, we introduce MOSAIC, a training-free multi-agent framework for scientific code generation without I/O supervision. Instead of execution feedback, MOSAIC employs a student-teacher knowledge distillation framework that grounds generation through domain-specific examples and structured problem decomposition. To further mitigate hallucinations across chained subproblems, we introduce a Consolidated Context Window (CCW) for maintaining consistent reasoning across agents. Experiments on the SciCode benchmark show that MOSAIC improves accuracy, executability, and numerical precision over existing approaches while relying on lightweight models.

142. Code Broker: A Multi-Agent System for Automated Code Quality Assessment

Authors: Samer Attrah
URL: https://arxiv.org/abs/2604.23088
Abstract:

We present Code Broker, a multi agent system built with Google Agent Development Kit ADK that analyses Python code from files, local directories, or GitHub repositories and generates actionable quality assessment reports. The system employs a hierarchical five agents architecture in which a root orchestrator coordinates a sequential pipeline agent, which in turn dispatches three specialised agents in parallel a Correctness Assessor, a Style Assessor, and a Description Generator before synthesising findings through an Improvement Recommender. Reports score four dimensions correctness, security, style, and maintainability and are rendered in both Markdown and HTML. Code Broker combines LLM based reasoning with deterministic static-analysis signals from Pylint, uses asynchronous execution with retry logic to improve robustness, and explores lightweight session memory for retaining and querying prior assessment context. We position the paper as a technical report on system design and prompt or tool orchestration, and present a preliminary qualitative evaluation on representative Python codebases. The results suggest that parallel specialised agents produce readable, developer oriented feedback, while also highlighting current limitations in evaluation depth, security tooling, large repository handling, and the current use of only in memory persistence. All code and reproducibility materials are available at: this https URL .

143. From Pixels to Explanations: Interpretable Diabetic Retinopathy Grading with CNN-Transformer Ensembles, Visual Explainability and Vision-Language Models

Authors: Pir Bakhsh Khokhar , Carmine Gravino , Fabio Palomba , Sule Yildirim Yayilgan , Sarang Shaikh
URL: https://arxiv.org/abs/2604.23079
Abstract:

The quality of diabetic retinopathy (DR) screening relies on the ability to correctly grade severity; however, many deep-learning (DL) classifiers cannot be easily interpreted in the clinical context. This study presents a methodology that combines strong discriminative models with multimodal explanations, converting retinal pixels into clinically interpretable outputs. Using the APTOS 2019 benchmark, we evaluated six representative CNN- and transformer-based backbones under a controlled protocol with stratified five-fold cross-validation. We then compared ensembling strategies (hard voting, weighted soft voting, stacking) and investigated a hybrid class-level fusion variant to exploit grade-specific advantages. For interpretability, we produced Grad-CAM++ visual attribution maps and short textual rationales using vision-language models (VLMs) conditioned on the fundus image and classifier outputs under conservative prompting constraints. Modern CNN backbones (ResNet-50 and ConvNeXt-Tiny) provided the strongest single-model baselines, with cross-validated QWK up to 0.919 and 0.914, respectively. Ensembling improved ordinal agreement, and weighted soft voting was the most consistent across folds (QWK 0.934 +/- 0.017). Hybrid class-level fusion was competitive but did not yield a statistically reliable improvement over standard fusion in paired fold comparisons (Holm-adjusted p >= 1.000). For explanation quality, Grad-CAM++ offered plausible but coarse localization, and VLM rationales were generally grade-consistent. Quantitatively, VLM variants showed a trade-off between clinical completeness and template-level semantic similarity (coverage 0.700 vs. BERTScore 0.072), while image-text alignment was comparable (CLIPScore approximately 0.34).

144. C-MORAL: Controllable Multi-Objective Molecular Optimization with Reinforcement Alignment for LLMs

Authors: Rui Gao , Youngseung Jeon , Swastik Roy , Morteza Ziyadi , Xiang ‘Anthony’ Chen
URL: https://arxiv.org/abs/2604.23061
Abstract:

Large language models (LLMs) show promise for molecular optimization, but aligning them with selective and competing drug-design constraints remains challenging. We propose C-Moral, a reinforcement learning post-training framework for controllable multi-objective molecular optimization. C-Moral combines group-based relative optimization, property score alignment for heterogeneous objectives, and continuous non-linear reward aggregation to improve stability across competing properties. Experiments on the C-MuMOInstruct benchmark show that C-Moral consistently outperforms state-of-the-art models across both in-domain and out-of-domain settings, achieving the best Success Optimized Rate (SOR) of 48.9% on IND tasks and 39.5% on OOD tasks, while largely preserving scaffold similarity. These results suggest that RL post-training is an effective way to align molecular language models with continuous molecular design objectives. Our code and models are publicly available at this https URL .

145. DeepImagine: Learning Biomedical Reasoning via Successive Counterfactual Imagining

Authors: Youze Zheng , Jianyou Wang , Yuhan Chen , Matthew Feng , Longtian Bao , Hanyuan Zhang , Maxim Khan , Aditya K. Sehgal , Christopher D. Rosin , Umber Dube , Ramamohan Paturi
URL: https://arxiv.org/abs/2604.23054
Abstract:

Predicting the outcomes of prospective clinical trials remains a major challenge for large language models. Prior work has shown that both traditional correlational predictors, such as random forests and logistic regression, and strong commercial LLMs achieve limited performance on this task. In this paper, we propose DeepImagine, a framework for teaching LLMs biomedical reasoning through successive counterfactual imagining. The central idea is to approximate hidden causal mechanisms of clinical trials by training models to infer how observed trial results would change under controlled perturbations of experimental conditions, such as dosage, outcome measures, study arms, geography, and other trial attributes. To support this objective, we construct both natural and approximate counterfactual pairs from real clinical trials with reported outcomes. For settings where strict counterfactual supervision is available, such as paired outcome measures or dose-ranging study arms within the same trial, we train models with supervised fine-tuning. For broader settings where only approximate counterfactual pairs can be retrieved, we optimize models with reinforcement learning using verifiable rewards based on downstream benchmark correctness. We further augment training with synthetic reasoning traces that provide causally plausible explanations for local counterfactual transitions. Using this pipeline, we train language models under 10B parameters, including Qwen3.5-9B, and evaluate them on clinical trial outcome prediction. We aim to show that DeepImagine consistently improves over untuned language models and traditional correlational baselines. Finally, we aim to show that the learned reasoning trajectories provide interpretable signals about how models represent trial-level mechanisms, suggesting a practical path toward more mechanistic and scientifically useful biomedical language models.

146. AmaraSpatial-10K: A Spatially and Semantically Aligned 3D Dataset for Spatial Computing and Embodied AI

Authors: Mohammad Sadegh Salehi , Alex Perkins , Igor Maurell , Ashkan Dabbagh , Raymond Wong
URL: https://arxiv.org/abs/2604.23018
Abstract:

Web-scale 3D asset collections are abundant, but rarely deployment-ready. Assets ship with arbitrary metric scale, incorrect pivots and forward axes, brittle geometry, and textures that do not support relighting, which limits their utility for embodied AI, robotics simulation, game development, and AR/VR. We present AmaraSpatial-10K, a dataset of over 10,000 synthetic 3D assets designed for downstream use rather than volume alone. Each asset is released as a metric-scaled, semantically anchored .glb with separated PBR material maps, a convex collision hull, a paired reference image, and rich multi-sentence text metadata. The dataset spans indoor objects, vehicles, architecture, creatures, and props under a unified spatial convention. Alongside the dataset, we introduce an evaluation suite for 3D asset banks. The suite comprises a continuous Scale Plausibility Score (SPS) with an LLM-as-Judge interval protocol, an LLM Concept Density score for metadata, an anchor-error metric, and a cross-modal CLIP coherence protocol, and we use it to audit AmaraSpatial-10K alongside matched subsets from Objaverse, HSSD, ABO, and GSO. Compared with Objaverse-sourced assets, we demonstrate that AmaraSpatial-10K substantially improves text-based retrieval precision (CLIP Recall@5 of 0.612 vs 0.181, a 3.4x improvement with median rank falling from 267 to 3), and we establish that it satisfies the spatial and semantic prerequisites for physics-aware scene composition and embodied-AI asset banks, leaving those downstream evaluations to future work. AmaraSpatial-10K is publicly available on Hugging Face.

147. CheXmix: Unified Generative Pretraining for Vision Language Models in Medical Imaging

Authors: Ashwin Kumar , Robbie Holland , Corey Barrett , Jangwon Kim , Maya Varma , Zhihong Chen , Yunhe Gao , Greg Zaharchuk , Tara Taghavi , Krishnaram Kenthapadi , Akshay Chaudhari
URL: https://arxiv.org/abs/2604.22989
Abstract:

Recent medical multimodal foundation models are built as multimodal LLMs (MLLMs) by connecting a CLIP-pretrained vision encoder to an LLM using LLaVA-style finetuning. This two-stage, decoupled approach introduces a projection layer that can distort visual features. This is especially concerning in medical imaging where subtle cues are essential for accurate diagnoses. In contrast, early-fusion generative approaches such as Chameleon eliminate the projection bottleneck by processing image and text tokens within a single unified sequence, enabling joint representation learning that leverages the inductive priors of language models. We present CheXmix, a unified early-fusion generative model trained on a large corpus of chest X-rays paired with radiology reports. We expand on Chameleon’s autoregressive framework by introducing a two-stage multimodal generative pretraining strategy that combines the representational strengths of masked autoencoders with MLLMs. The resulting models are highly flexible, supporting both discriminative and generative tasks at both coarse and fine-grained scales. Our approach outperforms well-established generative models across all masking ratios by 6.0% and surpasses CheXagent by 8.6% on AUROC at high image masking ratios on the CheXpert classification task. We further inpaint images over 51.0% better than text-only generative models and outperform CheXagent by 45% on the GREEN metric for radiology report generation. These results demonstrate that CheXmix captures fine-grained information across a broad spectrum of chest X-ray tasks. Our code is at: this https URL .

148. Peer Identity Bias in Multi-Agent LLM Evaluation: An Empirical Study Using the TRUST Democratic Discourse Analysis Pipeline

Authors: Juergen Dietrich
URL: https://arxiv.org/abs/2604.22971
Abstract:

The TRUST democratic discourse analysis pipeline exposes its large language model (LLM) components to peer model identity through multiple structural channels – a design feature whose bias implications have not previously been empirically tested. We provide the first systematic measurement of identity-dependent scoring bias across all active identity exposure channels in TRUST, crossing four model families with two anonymization scopes across 30 political statements. The central finding is that single-channel anonymization produces near-zero bias effects, because individual channels act in opposite directions and cancel each other out – a result that would lead an evaluator to conclude that identity bias is absent when it is not. Only full-pipeline anonymization reveals the true pattern: homogeneous ensembles amplify identity-driven sycophancy when model identity is fully visible, while the heterogeneous production configuration shows the reverse. Model choice matters independently: one tested model exhibits baseline sycophancy two to three times higher than the others and near-zero deliberative conflict on ideological topics, making it structurally unsuitable for pipelines where genuine inter-role disagreement is the intended quality mechanism. Three practical conclusions follow. First, heterogeneous model ensembles are structurally more robust than homogeneous ones, achieving higher consensus rates and lower identity amplification. Second, full-pipeline anonymization is required for valid bias measurement – partial anonymization is insufficient and actively misleading. Third, these findings have direct implications for the validation of multi-agent LLM systems in quality-critical applications: a system validated under partial anonymization or with a homogeneous ensemble may pass validation while retaining structural identity bias invisible to single-channel measurement.

149. Self Knowledge Re-expression: A Fully Local Method for Adapting LLMs to Tasks Using Intrinsic Knowledge

Authors: Mengyu Wang , Xiaoying Zhi , Zhiyi Li , Robin Schmucker , Shay B. Cohen , Tiejun Ma , Fran Silavong
URL: https://arxiv.org/abs/2604.22939
Abstract:

While the next-token prediction (NTP) paradigm enables large language models (LLMs) to express their intrinsic knowledge, its sequential nature constrains performance on specialized, non-generative tasks. We attribute this performance bottleneck to the LLMs’ knowledge expression mechanism, rather than to deficiencies in knowledge acquisition. To address this, we propose Self-Knowledge Re-expression (SKR), a novel, task-agnostic adaptation method. SKR transforms the LLM’s output from generic token generation to highly efficient, task-specific expression. SKR is a fully local method that uses only unannotated data, requiring neither human supervision nor model distillation. Experiments on a large financial document dataset demonstrate substantial improvements: over 40% in Recall@1 for information retrieval tasks, over 76% reduction in object detection latency, and over 33% increase in anomaly detection AUPRC. Our results on the MMDocRAG dataset surpass those of leading retrieval models by at least 12.6%.

150. Utility-Aware Data Pricing: Token-Level Quality and Empirical Training Gain for LLMs

Authors: Minghui Xu , Qi Luo , Kun Li
URL: https://arxiv.org/abs/2604.22893
Abstract:

Traditional data valuation methods based on ``row-count $\times$ quality coefficient’’ paradigms fail to capture the nuanced, nonlinear contributions that data makes to Large Language Model (LLM) capabilities. This paper presents a dynamic data valuation framework that transitions from static accounting to utility-based pricing. Our approach operates on three layers: (1) token-level information density metrics using Shannon entropy and Data Quality Scores; (2) empirical training gain measurement through influence functions, proxy model strategies, and Data Shapley values; and (3) cryptographic verifiability through hash-based commitments, Merkle trees, and a tamper-evident training ledger. We provide comprehensive experimental validation on three real domains (instruction following, mathematical reasoning, and code summarization), demonstrating that proxy-based empirical gain achieves near-perfect ranking alignment with realized utility, substantially outperforming row-count and token-count baselines. This framework enables a fair Data-as-a-Service economy where high-reasoning data is priced according to its actual contribution to model intelligence, while providing the transparency and auditability necessary for trustworthy data markets.

151. Quantifying and Mitigating Self-Preference Bias of LLM Judges

Authors: Jinming Yang , Chuxian Qiu , Zhenyu Deng , Xinshan Jiao , Tao Zhou
URL: https://arxiv.org/abs/2604.22891
Abstract:

LLM-as-a-Judge has become a dominant approach in automated evaluation systems, playing critical roles in model alignment, leaderboard construction, quality control, and so on. However, the scalability and trustworthiness of this approach can be substantially distorted by Self-Preference Bias (SPB), which is a directional evaluative deviation in which LLMs systematically favor or disfavor their own generated outputs during evaluation. Existing measurements rely on costly human annotations and conflate generative capability with evaluative stance, and thus are impractical for large-scale deployment in real-world systems. To address this issue, we introduce a fully automated framework to quantifying and mitigating SPB, which constructs equal-quality pairs of responses with negligible quality differences, enabling statistical disentanglement of discriminability from bias propensity without human gold standards. Empirical analysis across 20 mainstream LLMs reveals that advanced capabilities are often uncorrelated, or even negatively correlated, with low SPB. To mitigate this bias, we propose a structured multi-dimensional evaluation strategy grounded in cognitive load decomposition, which reduces SPB by 31.5\% on average.

152. RouteGuard: Internal-Signal Detection of Skill Poisoning in LLM Agents

Authors: Wenjie Xiao , Xuehai Tang , Biyu Zhou , Songlin Hu , Jizhong Han
URL: https://arxiv.org/abs/2604.22888
Abstract:

Agent skills introduce a new and more severe form of indirect injection for LLM agents: unlike traditional indirect prompt injection, attackers can hide malicious instructions inside a dense, action-oriented skill that already functions as a legitimate instruction source. We study pre-execution skill-poison detection and show that successful skill poisoning induces a structured internal effect, attention hijacking, in which response-time attention shifts from trusted context to malicious skill spans and drives harmful behavior. Motivated by this mechanism, we propose RouteGuard, a frozen-backbone detector that combines response-conditioned attention and hidden-state alignment through reliability-gated late fusion. Across both real and synthetic open-source skill benchmarks, RouteGuard is consistently the strongest or most robust detector; on the critical Skill-Inject channel slice, it reaches 0.8834 F1 and recovers 90.51% of description attacks missed by lexical screening, showing that defending against skill poisoning requires internal-signal detection rather than text-only filtering

153. Can Multimodal Large Language Models Truly Understand Small Objects?

Authors: Fujun Han , Junan Chen , Xintong Zhu , Jingqi Ye , Xuanjie Mao , Tao Chen , Peng Ye
URL: https://arxiv.org/abs/2604.22884
Abstract:

Multimodal Large Language Models (MLLMs) have shown promising potential in diverse understanding tasks, e.g., image and video analysis, math and physics olympiads. However, they remain blank and unexplored for Small Object Understanding (SOU) tasks. To fill this gap, we introduce SOUBench, the first and comprehensive benchmark for exploring the small objects understanding capability of existing MLLMs. Specifically, we first design an effective and automatic visual question-answer generation strategy, constructing a new SOU-VQA evaluation dataset, with 18,204 VQA pairs, six relevant sub-tasks, and three dominant scenarios (i.e., Driving, Aerial, and Underwater). Then, we conduct a comprehensive evaluation on 15 state-of-the-art MLLMs and reveal their weak capabilities in small object understanding. Furthermore, we develop SOU-Train, a multimodal training dataset with 11,226 VQA pairs, to improve the SOU capabilities of MLLMs. Through supervising fine-tuning of the latest MLLM, we demonstrate that SOU-Train can effectively enhance the latest MLLM’s ability to understand small objects. Comprehensive experimental results demonstrate that, the proposed SOUBench, along with the SOU-VQA and SOU-Train datasets, provides a crucial empirical foundation to the community for further developing models with enhanced small object understanding capabilities. Datasets and Code: this https URL .

154. SketchVLM: Vision language models can annotate images to explain thoughts and guide users

Authors: Brandon Collins , Logan Bolton , Hung Huy Nguyen , Mohammad Reza Taesiri , Trung Bui , Anh Totti Nguyen
URL: https://arxiv.org/abs/2604.22875
Abstract:

When answering questions about images, humans naturally point, label, and draw to explain their reasoning. In contrast, modern vision-language models (VLMs) such as Gemini-3-Pro and GPT-5 only respond with text, which can be difficult for users to verify. We present SketchVLM, a training-free, model-agnostic framework that enables VLMs to produce non-destructive, editable SVG overlays on the input image to visually explain their answers. Across seven benchmarks spanning visual reasoning (maze navigation, ball-drop trajectory prediction, and object counting) and drawing (part labeling, connecting-the-dots, and drawing shapes around objects), SketchVLM improves visual reasoning task accuracy by up to +28.5 percentage points and annotation quality by up to 1.48x relative to image-editing and fine-tuned sketching baselines, while also producing annotations that are more faithful to the model’s stated answer. We find that single-turn generation already achieves strong accuracy and annotation quality, and multi-turn generation opens up further opportunities for human-AI collaboration. An interactive demo and code are at this https URL .

155. AutoRISE: Agent-Driven Strategy Evolution for Red-Teaming Large Language Models

Authors: Tanmay Gautam , Alireza Bahramali , Sandeep Atluri
URL: https://arxiv.org/abs/2604.22871
Abstract:

Automated red-teaming methods for large language models typically optimize attack prompts within a fixed, human-designed strategy, leaving the attack strategy itself unchanged. We instead optimize the strategy. We propose AutoRISE, a method that searches over executable attack programs rather than individual prompts. At each iteration, a coding agent edits a strategy and a fixed evaluation harness scores the resulting attacks, returning both a scalar objective and per-example diagnostics that guide subsequent edits. This allows structural changes, including new attack components and altered control flow, that prompt-level methods do not directly express. We also release two benchmark suites developed on disjoint target sets and evaluate on 11 models from five families against seven established jailbreak datasets. Across held-out models, AutoRISE improves average attack success rate by 17.0 points over the strongest baseline, and improves attack success by up to 16 points on frontier targets with low baseline success rates. Ablations against parametric and strategy-library baselines suggest that these gains arise from unrestricted program search, particularly compositional techniques and control-flow edits. AutoRISE operates in a black-box, inference-only setting, requiring no fine-tuning, human annotation, or GPU compute.

156. IntrAgent: An LLM Agent for Content-Grounded Information Retrieval through Literature Review

Authors: Fengbo Ma , Zixin Rao , Xiaoting Li , Zhetao Chen , Hongyue Sun , Yiping Zhao , Xianyan Chen , Zhen Xiang
URL: https://arxiv.org/abs/2604.22861
Abstract:

Scientific research relies on accurate information retrieval from literature to support analytical decisions. In this work, we introduce a new task, INformation reTRieval through literAture reVIEW (IntraView), which aims to automate fine-grained information retrieval faithfully grounded in the provided content in response to research-driven queries, and propose IntrAgent, an LLM-based agent that addresses this challenging task. In particular, IntrAgent is designed to mimic human behaviors when reading literature for information retrieval – identifying relevant sections and then iteratively extracting key details to refine the retrieved information. It follows a two-stage pipeline: a Section Ranking stage that prioritizes relevant literature sections through structural-knowledge-enabled reasoning, and an Iterative Reading stage that continuously extracts details and synthesizes them into concise, contextually grounded answers. To support rigorous evaluation, we introduce IntraBench, a new benchmark consisting of 315 test instances built from expert-authored questions paired with literature spanning five STEM domains. Across seven backbone LLMs, IntrAgent achieves on average 13.2% higher cross-domain accuracy than state-of-the-art RAG and research-agent baselines.

157. SwarmDrive: Semantic V2V Coordination for Latency-Constrained Cooperative Autonomous Driving

Authors: Anjie Qiu , Donglin Wang , Zexin Fang , Sanket Partani , Hans D. Schotten
URL: https://arxiv.org/abs/2604.22852
Abstract:

Cloud-hosted LLM inference for autonomous driving adds round-trip delay and depends on stable connectivity, while purely local edge models struggle under occlusion. We present SwarmDrive, a semantic Vehicle-to-Vehicle (V2V) coordination framework in which nearby vehicles run local Small Language Models (SLMs), share compact intent distributions only when uncertainty is high, and fuse them through event-triggered consensus. We evaluate SwarmDrive in a 5-seed executable study built around one occluded intersection case, combining matched operating-point comparisons with robustness sweeps. In that setting, SwarmDrive under its 6G communication setting (“Swarm 6G”) raises success from 68.9% to 94.1% over a single local SLM while reducing latency from a 510 ms cloud reference to 151.4 ms. However, an increased number of participating vehicles leads to higher communication overhead and packet loss. SwarmDrive also evaluates the impact of swarm-size, packet-loss, and entropy-threshold sweeps and shows that the cooperative gain holds across ablations and is best balanced near an active swarm size of 4 vehicles and an entropy trigger threshold of 0.65 in the current prototype. These results show that semantic edge cooperation can work under tight latency constraints in the targeted intersection case, but they are not a deployment-grade validation of a real 6G stack.

158. Structure Guided Retrieval-Augmented Generation for Factual Queries

Authors: Miao Xie , Xiao Zhang , Yi Li , Chunli Lv
URL: https://arxiv.org/abs/2604.22843
Abstract:

Retrieval-Augmented Generation (RAG) has been proposed to mitigate hallucinations in large language models (LLMs), where generated outputs may be factually incorrect. However, existing RAG approaches predominantly rely on vector similarity for retrieval, which is prone to semantic noise and fails to ensure that generated responses fully satisfy the complex conditions specified by factual queries, often leading to incorrect answers. To address this challenge, we introduce a novel research problem, named Exact Retrieval Problem (ERP). To the best of our knowledge, this is the first problem formulation that explicitly incorporates structural information into RAG for factual questions to satisfy all query conditions. For this novel problem, we propose Structure Guided Retrieval-Augmented Generation (SG-RAG), which models the retrieval process as an embedding-based subgraph matching task, and uses the retrieved topological structures to guide the LLM to generate answers that meet all specified query conditions. To facilitate evaluation of ERP, we construct and publicly release Exact Retrieval Question Answering (ERQA), a large-scale dataset comprising 120000 fact-oriented QA pairs, each involving complex conditions, spanning 20 diverse domains. The experimental results demonstrate that SG-RAG significantly outperforms strong baselines on ERQA, delivering absolute improvements from 20.68 to 50.88 points across all evaluation metrics, while maintaining reasonable computational overhead.

159. PivotMerge: Bridging Heterogeneous Multimodal Pre-training via Post-Alignment Model Merging

Authors: Zibo Shao , Baochen Xiong , Xiaoshan Yang , Yaguang Song , Qimeng Zhang , Haifeng Chen , Changsheng Xu
URL: https://arxiv.org/abs/2604.22823
Abstract:

Multimodal Large Language Models (MLLMs) rely on multimodal pre-training over diverse data sources, where different datasets often induce complementary cross-modal alignment capabilities. Model merging provides a cost-effective mechanism for integrating multiple expert MLLMs with complementary strengths into a unified model. However, existing model merging research mainly focuses on post-finetuning scenarios, leaving the pre-training stage largely unexplored. We argue that the core of MLLM pre-training lies in establishing effective cross-modal alignment, which bridges visual and textual representations into a unified semantic space. Motivated by this insight, we introduce the post-alignment merging task, which aims to integrate cross-modal alignment capabilities learned from heterogeneous multimodal pre-training. This setting introduces two key challenges: cross-domain parameter interference, where parameter updates learned from different data distributions conflict during merging, and layer-wise alignment contribution disparity, where different layers and projectors contribute unevenly to cross-modal alignment. To address them, we propose \textbf{PivotMerge}, a post-alignment merging framework for cross-modal projectors. PivotMerge incorporates two key components: Shared-space Decomposition and Filtering, which disentangles shared alignment patterns from domain-specific variations and suppresses conflicting directions, and Alignment-guided Layer-wise Merging, which assigns layer-specific merging weights based on differing alignment contributions. We construct systematic CC12M-based post-alignment merging scenarios for evaluation. Extensive experiments on multiple multimodal benchmarks show that PivotMerge consistently outperforms existing baselines, demonstrating its effectiveness and generalization ability.

160. DO-Bench: An Attributable Benchmark for Diagnosing Object Hallucination in Vision-Language Models

Authors: JiYang Wang , Jiawei Chen , Mengqi Xiao , Yu Cheng , Yangfu Li , Zhaoxia Yin
URL: https://arxiv.org/abs/2604.22822
Abstract:

Object level hallucination remains a central reliability challenge for vision language models (VLMs), particularly in binary object existence verification. Existing benchmarks emphasize aggregate accuracy but rarely disentangle whether errors stem from perceptual limitations or from the influence of contextual textual priors, leaving underlying failure mechanisms ambiguous. We introduce DO-Bench, a controlled diagnostic benchmark that isolates these sources through structured multimodal interventions. Rather than evaluating models in unconstrained settings, DO-Bench probes two complementary dimensions: the Prior Override dimension progressively strengthens contextual textual priors while holding visual evidence constant to assess resistance to prior pressure, and the Perception-Limited dimension incrementally enhances visual evidence from full-scene context to localized object crops to measure perceptual grounding strength. This paired design enables attribution of errors to prior suppression, perceptual insufficiency, or their interaction. We further define two diagnostic metrics, PriorRobust and PerceptionAbility, to quantify these behaviors consistently. Evaluations across diverse open- and closed-source VLMs reveal systematic differences in prior sensitivity and perceptual reliability, demonstrating that object hallucination reflects heterogeneous, mechanism dependent failure patterns beyond aggregate accuracy.

161. Complete Cyclic Subtask Graphs for Tool-Using LLM Agents: Flexibility, Cost, and Bottlenecks in Multi-Agent Workflows

Authors: Luay Gharzeddine , Samer Saab Jr
URL: https://arxiv.org/abs/2604.22820
Abstract:

Long-horizon tool-using tasks sometimes benefit from revisiting earlier subtasks for recovery and exploration, but added multi-agent workflow flexibility can also introduce coordination overhead and substantial inference cost. We study complete cyclic subtask graphs, a deliberately maximally flexible multi-agent architecture in which executable subtask nodes are fully connected and a unified state-analysis-and-routing agent selects transitions using natural-language criteria. This makes unrestricted revisitation explicit and directly analyzable at the subtask level. We evaluate task-specific (Spec-Cyc) and benchmark-generic (Gen-Cyc) graphs on TextCraft, ALFWorld, and Finance-Agent, with ablations over planner/executor/router strength, tool exposure (generalist vs specialized), $n$-shot successful trajectory summaries, and fault-injected random subtask perturbations. The benchmarks expose three distinct regimes. ALFWorld highlights a setting where explicit revisitation supports recovery and exploration; TextCraft, a largely prerequisite-chain domain, often favors the efficiency of simpler forward execution; and Finance-Agent remains bottlenecked by retrieval, grounding, and evidence synthesis more than by workflow flexibility alone. Shared-win token comparisons further show that the added flexibility can be substantially more expensive than a single ReAct agent. Overall, we use complete cyclic subtask graphs as a maximally flexible experimental lens for measuring when multi-agent revisitation helps, when it mainly adds coordination cost, and when external task bottlenecks dominate.

162. See No Evil: Semantic Context-Aware Privacy Risk Detection for AR

Authors: Jialu Liu , Yao Li , Zhuoheng Li , Huining Li , Ying Chen
URL: https://arxiv.org/abs/2604.22805
Abstract:

Augmented reality (AR) systems pose unique privacy risks due to their continuous capture of visual data. Existing AR privacy frameworks lack semantic understanding of visual content, limiting their effectiveness in detecting context-dependent privacy risks. We propose PrivAR, which leverages vision language models (VLMs) with chain-of-thought prompting for contextual privacy risk detection in AR environments. PrivAR uses visual scene cues to infer potential sensitive information types, such as identifying password notes in office environments through contextual reasoning. PrivAR detects and obfuscates textual content, preventing exposure of sensitive information while preserving contextual cues necessary for VLM inference. Additionally, we investigate contextually-informed warning interfaces to enhance user privacy awareness. Experiments on a real-world AR dataset show that PrivAR achieves superior accuracy (81.48%) and F1-score (84.62%) compared to baselines, while reducing privacy leakage rate to 17.58%. User studies evaluating contextually-informed warning interfaces provide insights into effective privacy-aware AR design.

163. RCSB PDB AI Help Desk: retrieval-augmented generation for protein structure deposition support

Authors: Vivek Reddy Chithari (1), Jasmine Y. Young (1), Irina Persikova (1), Yuhe Liang (1), Gregg V. Crichlow (1), Justin W. Flatt (1), Sutapa Ghosh (1), Brian P. Hudson (1), Ezra Peisach (1), Monica Sekharan (1), Chenghua Shao (1), Stephen K. Burley (1 and 2) ((1) RCSB Protein Data Bank, Rutgers, The State University of New Jersey, Piscataway, NJ, USA, (2) RCSB Protein Data Bank, San Diego Supercomputer Center, University of California San Diego, CA, USA)
URL: https://arxiv.org/abs/2604.22800
Abstract:

Motivation: Structural Biologists have contributed more than 245,000 experimentally determined three-dimensional structures of biological macromolecules to the Protein Data Bank (PDB). Incoming data are validated and biocurated by ~20 expert biocurators across the wwPDB. RCSB PDB biocurators who process more than 40% of global depositions face increasing challenges in maintaining efficient Help Desk operations, with approximately 19,000 messages in approximately 8,000 entries received from depositors in 2025. Results: We developed an AI-powered Help Desk using Retrieval-Augmented Generation (RAG) built on LangChain with a pgvector store (PostgreSQL) and GPT-4.1-mini. The system employs pymupdf4llm for Markdown-preserving PDF extraction, two-stage document chunking, Maximal Marginal Relevance retrieval, a topical guardrail that filters off-topic queries, and a specialized system prompt that prevents exposure of internal terminology. A dual-LLM architecture uses separate model configurations for question condensing and response generation. Deployed in production on Kubernetes with PostgreSQL (pgvector), it provides around-the-clock depositor assistance with citation-backed, streaming responses. Availability and implementation: Freely available at this https URL .

164. Parameter Efficiency Is Not Memory Efficiency: Rethinking Fine-Tuning for On-Device LLM Adaptation

Authors: Irene Tenison , Stella Ahn , Miriam Kim , Ebtisam Alshehri , Lalana Kagal
URL: https://arxiv.org/abs/2604.22783
Abstract:

Parameter-Efficient Fine-Tuning (PEFT) has become the standard for adapting large language models (LLMs). In this work we challenge the wide-spread assumption that parameter efficiency equates memory efficiency and on-device adaptability. We show that this is not true - while methods like LoRA and IA3 significantly reduce trainable parameters, they remain bound by intermediate tensors that scale linearly with sequence length, often triggering out-of-memory errors on-device. In this work, we introduce LARS (Low-memory Activation-Rank Subspace), a novel adaptation framework that decouples memory consumption from sequence length. While prior PEFT methods apply low-rank constraints to model parameters, LARS instead constrains the activation subspace used during training, directly targeting the dominant source of memory consumption and fundamentally flattening the memory growth rate. LARS reduces the memory footprint by an average of 33.54% on GPUs and 51.95% on CPUs in comparison to LoRA across reasoning, understanding and long-context datasets using different models while maintaining competitive accuracy and throughput. Besides GPUs, we deploy on Raspberry Pi and consumer-grade CPUs to demonstrate that LARS provides a scalable path for sophisticated LLM personalization on resource-constrained hardware and edge devices.

Authors: Anastasiia Filippova , David Grangier , Marco Cuturi , João Monteiro
URL: https://arxiv.org/abs/2604.22782
Abstract:

Serving transformer language models with high throughput requires caching Key-Values (KVs) to avoid redundant computation during autoregressive generation. The memory footprint of KV caching is significant and heavily impacts serving costs. This work proposes to lessen these memory requirements. While recent work has largely addressed KV cache reduction via compression and eviction along the temporal axis, we argue that the \emph{depth} dimension offers an orthogonal and robust avenue for optimization. Although prior research suggests that a full cache for every layer is redundant, implementing cross-layer cache sharing remains a practical challenge; existing methods typically suffer from reduced throughput or increased time-to-first-token. In this paper, we demonstrate that dropping a layer’s cache offers efficient optimization without information loss. We propose a simple training approach: random cross-layer attention. During training, layers randomly choose to attend either to their own KV states or those of a preceding layer. This stochastic process adapts the model to be robust to various depth-wise cache sharing strategies, ensuring flexibility for unknown hardware constraints at deployment time. Our evaluations show that applying this scheme during pre-training or fine-tuning enables depth-wise cache sharing for various model families. Furthermore, for larger models in data-constrained settings, this approach is suggestive of a regularization-like effect, frequently preserving or improving performance while significantly reducing the cache’s memory footprint.

166. KARL: Mitigating Hallucinations in LLMs via Knowledge-Boundary-Aware Reinforcement Learning

Authors: Cheng Gao , Cheng Huang , Kangyang Luo , Ziqing Qiao , Shuzheng Si , Huimin Chen , Chaojun Xiao , Maosong Sun
URL: https://arxiv.org/abs/2604.22779
Abstract:

Enabling large language models (LLMs) to appropriately abstain from answering questions beyond their knowledge is crucial for mitigating hallucinations. While existing reinforcement learning methods foster autonomous abstention, they often compromise answer accuracy because their static reward mechanisms, agnostic to models’ knowledge boundaries, drive models toward excessive caution. In this work, we propose KARL, a novel framework that continuously aligns an LLM’s abstention behavior with its evolving knowledge boundary. KARL introduces two core innovations: a Knowledge-Boundary-Aware Reward that performs online knowledge boundary estimation using within-group response statistics, dynamically rewarding correct answers or guided abstention; and a Two-Stage RL Training Strategy that first explores the knowledge boundary and bypasses the “abstention trap”, and subsequently converts incorrect answers beyond the knowledge boundary into abstentions without sacrificing accuracy. Extensive experiments on multiple benchmarks demonstrate that KARL achieves a superior accuracy-hallucination trade-off, effectively suppressing hallucinations while maintaining high accuracy across both in-distribution and out-of-distribution scenarios.

167. Epicure: Multidimensional Flavor Structure in Food Ingredient Embeddings

Authors: Jakub Radzikowski , Josef Chen
URL: https://arxiv.org/abs/2604.22776
Abstract:

A chef’s intuition about flavor, texture, and cultural identity represents tacit knowledge that is difficult to articulate yet central to culinary practice. We show that this knowledge is already encoded in FlavorGraph’s 300-dimensional ingredient embeddings, trained on recipe cooccurrence and food chemistry, and that it can be systematically recovered. An LLM-augmented curation pipeline consolidates 6,653 raw FlavorGraph ingredients into 1,032 canonical entries, substantially strengthening the recoverable structure. We identify at least fifteen independently classifiable dimensions spanning taste, texture, geography, food processing, and culture.

168. When VLMs ‘Fix’ Students: Identifying and Penalizing Over-Correction in the Evaluation of Multi-line Handwritten Math OCR

Authors: Jin Seong , Wencke Liermann , Minho Kim , Jong-hun Shin , Soojong Lim
URL: https://arxiv.org/abs/2604.22774
Abstract:

Accurate transcription of handwritten mathematics is crucial for educational AI systems, yet current benchmarks fail to evaluate this capability properly. Most prior studies focus on single-line expressions and rely on lexical metrics such as BLEU, which fail to assess the semantic reasoning across multi-line student solutions. In this paper, we present the first systematic study of multi-line handwritten math Optical Character Recognition (OCR), revealing a critical failure mode of Vision-Language Models (VLMs): over-correction. Instead of faithfully transcribing a student’s work, these models often “fix” errors, thereby hiding the very mistakes an educational assessment aims to detect. To address this, we propose PINK (Penalized INK-based score), a semantic evaluation metric that leverages a Large Language Model (LLM) for rubric-based grading and explicitly penalizes over-correction. Our comprehensive evaluation of 15 state-of-the-art VLMs on the FERMAT dataset reveals substantial ranking reversals compared to BLEU: models like GPT-4o are heavily penalized for aggressive over-correction, whereas Gemini 2.5 Flash emerges as the most faithful transcriber. Furthermore, human expert studies show that PINK aligns significantly better with human judgment (55.0% preference over BLEU’s 39.5%), providing a more reliable evaluation framework for handwritten math OCR in educational settings.

169. The Randomness Floor: Measuring Intrinsic Non-Randomness in Language Model Token Distributions

Authors: Jarosław Hryszko
URL: https://arxiv.org/abs/2604.22771
Abstract:

Language models cannot be random. This paper introduces Entropic Deviation (ED), the normalised KL divergence between a model’s token distribution and the uniform distribution, and measures it systematically across 31,200 generations spanning seven models, two architectures (transformer and state space), nine prompt categories, three temperatures, and five languages. Under semantically neutral prompts (empty strings, random characters, nonsense syllables) transformers still exhibit ED of approximately 0.30, meaning that 88-93% of the non-randomness observed under semantic prompts is intrinsic to the learned weights rather than induced by context. Three transformer families (Gemma, Llama, Qwen) converge on nearly identical ED values despite different training data and vocabularies. A state space model (Mamba2) reveals a qualitatively different regime: twice the ED, three times lower within-sequence variance, and massive sensitivity to temperature (r = -0.78) where transformers are nearly immune (r < 0.05). Cross-lingual experiments with Qwen-32B show a stable gradient across five languages (English, Japanese, Chinese, Polish, Arabic) that does not correlate with token fertility and persists when two languages sharing an identical tokeniser subset are compared. These findings establish a structural lower bound on randomness in pretrained language models, characterise how this bound differs across architectures, and demonstrate that language itself modulates the bound independently of tokenisation.

170. Learning in Blocks: A Multi Agent Debate Assisted Personalized Adaptive Learning Framework for Language Learning

Authors: Nicy Scaria , Silvester John Joseph Kennedy , Deepak Subramani
URL: https://arxiv.org/abs/2604.22770
Abstract:

Most digital language learning curricula rely on discrete-item quizzes that test recall rather than applied conversational proficiency. When progression is driven by quiz performance, learners can advance despite persistent gaps in using grammar and vocabulary during interaction. Recent work on LLM-based judging suggests a path toward scoring open-ended conversations, but using interaction evidence to drive progression and review requires scoring protocols that are reliable and validated. We introduce Learning in Blocks, a framework that grounds progression in demonstrated conversational competence evaluated using CEFR-aligned rubrics. The framework employs heterogeneous multi-agent debate (HeteroMAD) in two stages: a scoring stage where role-specialized agents independently evaluate Grammar, Vocabulary, and Interactive Communication, engage in debate to address conflicting judgments, and a judge synthesizes consensus scores; and a recommendation stage that identifies specific grammar skills and vocabulary topics for targeted review. Progression requires demonstrating 70% mastery, and spaced review targets identified weaknesses to counter skill decay. We benchmark four scoring and recommendation methods on CEFR A2 conversations annotated by ESL experts. HeteroMAD achieves a superior score agreement with a 0.23 degree of variation and recommendation acceptability of 90.91%. An 8-week study with 180 CEFR A2 learners demonstrates that combining rubric-aligned scoring and recommendation with spaced review and mastery-based progression produces better learning outcomes than feedback alone.

171. Artificial General Intelligence Forecasting and Scenario Analysis: State of the Field, Methodological Gaps, and Strategic Implications

Authors: Gopal P. Sarma , Sunny D. Bhatt , Michael Jacob , Rachel Steratore
URL: https://arxiv.org/abs/2604.22766
Abstract:

In this report, we review the current state of methodologies to forecast the arrival of artificial general intelligence, assess their reliability, and analyze the implications for strategy and policy. We synthesize diverse forecasting approaches, document significant limitations in existing methods, and propose a research agenda for developing more-robust forecasting infrastructure. The report does not endorse a specific forecast or scenario but rather provides a framework for interpreting forecasts under conditions of deep uncertainty. We experimented with an iterative approach to human and artificial intelligence collaboration for this report. The primary drafting of the text was performed by large language models (GPT 5.1, Gemini 3 Pro, and Claude 4.5 Opus), with human researchers providing direction, peer review, fact-checking, and revision.

172. Implicit Humanization in Everyday LLM Moral Judgments

Authors: Hoda Ayad , Tanu Mitra
URL: https://arxiv.org/abs/2604.22764
Abstract:

Recent adoption of conversational information systems has expanded the scope of user queries to include complex tasks such as personal advice-seeking. However, we identify a specific type of sought advice-a request for a moral judgment (i.e. “who was wrong?”) in a social conflict-as an implicitly humanizing query which carries potentially harmful anthropomorphic projections. In this study, we examine the reinforcement of these assumptions in the responses of four major general-purpose LLMs through the use of linguistic, behavioral, and cognitive anthropomorphic cues. We also contribute a novel dataset of simulated user queries for moral judgments. We find current LLM system responses reinforce implicit humanization in queries, potentially exacerbating risks like overreliance or misplaced trust. We call for future work to expand the understanding of anthropomorphism to include implicit userside humanization and to design solutions that address user needs while correcting misaligned expectations of model capabilities.

173. Behavioral Intelligence Platforms: From Event Streams to Autonomous Insight via Probabilistic Journey Graphs, Behavioral Knowledge Extraction, and Grounded Language Generation

Authors: Arun Patra , Bhushan Vadgave
URL: https://arxiv.org/abs/2604.22762
Abstract:

Contemporary product analytics systems require users to pose explicit queries, such as writing SQL, configuring dashboards, or constructing funnels, before insights can surface. This pull-based paradigm creates a bottleneck: it requires both domain knowledge and technical fluency, and assumes practitioners know in advance which questions to ask. We argue that behavioral analytics should move from passive systems that answer queries to active systems that continuously detect and explain behavioral phenomena. We present the Behavioral Intelligence Platform (BIP), a system architecture that transforms raw event streams into automatically generated insights. BIP consists of four layers. First, Normalization and State Derivation (NSD) standardizes events and maps them to a semantic state hierarchy. Second, a Behavioral Graph Engine (BGE) models user journeys as absorbing Markov chains and computes transition probabilities, removal effects, and path quality metrics. Third, a Behavioral Knowledge Graph (BKG) and Detector System convert graph outputs into grounded behavioral facts and identify behavioral phenomena. Finally, a Grounded Language Layer constrains large language model outputs to verified facts, producing reliable narrative insights. We formalize the Behavioral Intelligence Problem, introduce a taxonomy of detectors for autonomous insight generation, and propose an interestingness score to prioritize insights under limited attention.

174. Quantifying Divergence in Inter-LLM Communication Through API Retrieval and Ranking

Authors: Eyhab Al-Masri
URL: https://arxiv.org/abs/2604.22760
Abstract:

Large language models (LLMs) increasingly operate as autonomous agents that reason over external APIs to perform complex tasks. However, their reliability and agreement remain poorly characterized. We present a unified benchmarking framework to quantify inter-LLM divergence, defined as the extent to which models differ in API discovery and ranking under identical tasks. Across 15 canonical API domains and 5 major model families, we measure pairwise and group-level agreement using set-, rank-, and consensus-based metrics including Average Overlap, Jaccard similarity, Rank-Biased Overlap, Kendall’s tau, Kendall’s W, and Cronbach’s alpha. Results show moderate overall alignment (AO about 0.50, tau about 0.45) but strong domain dependence: structured tasks (Weather, Speech-to-Text) are stable, while open-ended tasks (Sentiment Analysis) exhibit substantially higher divergence. Volatility and consensus analyses reveal that coherence clusters around data-bound domains and degrades for abstract reasoning tasks. These insights enable reliability-aware orchestration in multi-agent systems, where consensus weighting can improve coordination among heterogeneous LLMs. Beyond performance benchmarking, our results reveal systematic failure modes in multi-agent LLM coordination, where apparent agreement can mask instability in action-relevant rankings. This hidden divergence poses a pre-deployment safety risk and motivates diagnostic benchmarks for early detection.

175. RedParrot: Accelerating NL-to-DSL for Business Analytics via Query Semantic Caching

Authors: Tong Wang , Yongqin Xu , Jianfeng Zhang , Lingxi Cui , Wenqing Wei , Suzhou Chen , Huan Li , Ke Chen , Lidan Shou
URL: https://arxiv.org/abs/2604.22758
Abstract:

Recently, at Xiaohongshu, the rapid expansion of e-commerce and advertising demands real-time business analytics with high accuracy and low latency. To meet this demand, systems typically rely on converting natural language (NL) queries into Domain-Specific Languages (DSLs) to ensure semantic consistency, validation, and portability. However, existing multi-stage LLM pipelines for this NL-to-DSL task suffer from prohibitive latency, high cost, and error propagation, rendering them unsuitable for enterprise-scale deployment. In this paper, we propose RedParrot, a novel NL-to-DSL framework that accelerates inference via a semantic cache. Observing the high repetition and stable structural patterns in user queries, RedParrot bypasses the costly pipeline by matching new requests against cached “query skeletons” (normalized structural patterns) and adapting their corresponding DSLs. Our core technical contributions include (1) an offline skeleton construction strategy, (2) an online, entity-agnostic embedding model trained via contrastive learning for robust matching, and (3) a heterogeneous Retrieval-Augmented Generation (RAG) method that integrates diverse knowledge sources to handle unseen entities. Experiments on six real enterprise datasets from Xiaohongshu show RedParrot achieves an average 3.6x speedup and an 8.26% accuracy improvement. Furthermore, on new public benchmarks adapted from Spider and BIRD, it boosts accuracy by 34.8%, substantially outperforming standard in-context learning baselines.

176. Your Reviews Replicate You: LLM-Based Agents as Customer Digital Twins for Conjoint Analysis

Authors: Bin Xuan , Jungmin Hwang , Hakyeon Lee
URL: https://arxiv.org/abs/2604.22756
Abstract:

Conjoint analysis is a cornerstone of market research for estimating consumer preferences; however, traditional methods face persistent challenges regarding time, cost, and respondent fatigue. To address these limitations, this study proposes a framework that utilizes large language model (LLM)-based “customer digital twins (CDT)” as virtual respondents. We identified active users within the Reddit community and aggregated their comprehensive review histories to construct individualized vector databases. By integrating retrieval-augmented generation (RAG) with prompt engineering, this study developed customer agents capable of dynamically retrieving and reasoning upon their specific past preferences and constraints. These customer agents, called CDTs, performed pairwise comparison tasks on product profiles generated via fractional factorial design, and the resulting choice data was analyzed to estimate part-worth utilities by logistic regression. Empirical validation demonstrates that these CDTs predict the preferences of actual users with 87.73% accuracy. Furthermore, a case study on the computer monitor category successfully quantified trade-offs between attributes such as panel type and resolution, deriving preference structures consistent with market realities. Ultimately, this study contributes to marketing research by presenting a scalable alternative that significantly improves both agility and cost-efficiency to traditional methods.

177. RADIANT-LLM: an Agentic Retrieval Augmented Generation Framework for Reliable Decision Support in Safety-Critical Nuclear Engineering

Authors: Zavier Ndum Ndum , Jian Tao , John Ford , Mansung Yim , Yang Liu
URL: https://arxiv.org/abs/2604.22755
Abstract:

Reliable decision support in nuclear engineering requires traceable, domain-grounded knowledge retrieval, yet safety and risk analysis workflows remain hampered by fragmented documentation and hallucination when use pre-trained large language model (LLM) in specialized nuclear domains. To address these challenges, this paper presents RADIANT-LLM (Retrival-Augumented, Domain-Intelligent Agent for Nuclear Technologies using LLM), a multi-modal retrieval-augmented generation (RAG) framework designed for nuclear safety, security, and safeguards applications. The framework uses a local-first, model-agnostic architecture that pairs a multi-modal document ingestion pipeline with a structured, metadata-rich knowledge base, supporting page- and figure-level retrieval from technical documents. An agentic layer coordinates domain-specific tools, enforces citation-backed responses with provenance tracking, and supports human-in-the-loop validation to reduce hallucination risks. To rigorously evaluate this framework, we develop and apply a suite of domain-aware metrics, including Context Precision (CoP), Hallucination Rate (HR), and Visual Recall (ViR), to expert-curated benchmarks derived from Used Nuclear Fuel Storage Facility design guidance. Across varying knowledge base sizes, CoP and ViR remain within an 85–98\% band, and hallucination rates are substantially lower than those observed in general-purpose deployments. When the same queries are posed to commercial LLM platforms without the RAG layer, hallucinations and citation errors increase markedly. These results indicate that a locally controlled, multi-modal RAG framework with domain-specific retrieval and provenance enforcement is necessary to achieve the factual accuracy, transparency, and auditability that nuclear engineering workflows demand.

178. How Do AI Agents Spend Your Money? Analyzing and Predicting Token Consumption in Agentic Coding Tasks

Authors: Longju Bai , Zhemin Huang , Xingyao Wang , Jiao Sun , Rada Mihalcea , Erik Brynjolfsson , Alex Pentland , Jiaxin Pei
URL: https://arxiv.org/abs/2604.22750
Abstract:

The wide adoption of AI agents in complex human workflows is driving rapid growth in LLM token consumption. When agents are deployed on tasks that require a significant amount of tokens, three questions naturally arise: (1) Where do AI agents spend the tokens? (2) Which models are more token-efficient? and (3) Can agents predict their token usage before task execution? In this paper, we present the first systematic study of token consumption patterns in agentic coding tasks. We analyze trajectories from eight frontier LLMs on SWE-bench Verified and evaluate models’ ability to predict their own token costs before task execution. We find that: (1) agentic tasks are uniquely expensive, consuming 1000x more tokens than code reasoning and code chat, with input tokens rather than output tokens driving the overall cost; (2) token usage is highly variable and inherently stochastic: runs on the same task can differ by up to 30x in total tokens, and higher token usage does not translate into higher accuracy; instead, accuracy often peaks at intermediate cost and saturates at higher costs; (3) models vary substantially in token efficiency: on the same tasks, Kimi-K2 and Claude-Sonnet-4.5, on average, consume over 1.5 million more tokens than GPT-5; (4) task difficulty rated by human experts only weakly aligns with actual token costs, revealing a fundamental gap between human-perceived complexity and the computational effort agents actually expend; and (5) frontier models fail to accurately predict their own token usage (with weak-to-moderate correlations, up to 0.39) and systematically underestimate real token costs. Our study offers new insights into the economics of AI agents and can inspire future research in this direction.

LLM 관련 주요 논문 - 2026-04-28

1. Case-Specific Rubrics for Clinical AI Evaluation: Methodology, Validation, and LLM-Clinician Agreement Across 823 Encounters

2. The Price of Agreement: Measuring LLM Sycophancy in Agentic Financial Applications

3. XGRAG: A Graph-Native Framework for Explaining KG-based Retrieval-Augmented Generation

4. Evaluating whether AI models would sabotage AI safety research

5. A systematic evaluation of vision-language models for observational astronomical reasoning tasks

6. Towards Lawful Autonomous Driving: Deriving Scenario-Aware Driving Requirements from Traffic Laws and Regulations

7. STELLAR-E: a Synthetic, Tailored, End-to-end LLM Application Rigorous Evaluator

8. Beyond the Attention Stability Boundary: Agentic Self-Synthesizing Reasoning Protocols

9. Agentic clinical reasoning over longitudinal myeloma records: a retrospective evaluation against expert consensus

10. PhysNote: Self-Knowledge Notes for Evolvable Physical Reasoning in Vision-Language Model

11. Aligning with Your Own Voice: Self-Corrected Preference Learning for Hallucination Mitigation in LVLMs

12. Adaptive ToR: Complexity-Aware Tree-Based Retrieval for Pareto-Optimal Multi-Intent NLU

13. Multi-Dimensional Evaluation of Sustainable City Trips with LLM-as-a-Judge and Human-in-the-Loop

14. An Information-Geometric Framework for Stability Analysis of Large Language Models under Entropic Stress

15. Grounding Before Generalizing: How AI Differs from Humans in Causal Transfer

16. A2DEPT: Large Language Model-Driven Automated Algorithm Design via Evolutionary Program Trees

17. Representational Curvature Modulates Behavioral Uncertainty in Large Language Models

18. LLM-Guided Agentic Floor Plan Parsing for Accessible Indoor Navigation of Blind and Low-Vision People

19. Context-Aware Hospitalization Forecasting Evaluations for Decision Support using LLMs

20. LLM-Augmented Traffic Signal Control with LSTM-Based Traffic State Prediction and Safety-Constrained Decision Support

21. ZenBrain: A Neuroscience-Inspired 7-Layer Memory Architecture for Autonomous AI Systems

22. ClawTrace: Cost-Aware Tracing for LLM Agent Skill Distillation

23. Domain-Filtered Knowledge Graphs from Sparse Autoencoder Features

24. FAIR_XAI: Improving Multimodal Foundation Model Fairness via Explainability for Wellbeing Assessment

25. Expert Evaluation of LLM’s Open-Ended Legal Reasoning on the Japanese Bar Exam Writing Task

26. Vibe Medicine: Redefining Biomedical Research Through Human-AI Co-Work

27. Tandem: Riding Together with Large and Small Language Models for Efficient Reasoning

28. Thinking Like a Clinician: A Cognitive AI Agent for Clinical Diagnosis via Panoramic Profiling and Adversarial Debate

29. When AI reviews science: Can we trust the referee?

30. MetaGAI: A Large-Scale and High-Quality Benchmark for Generative AI Model and Data Card Generation

31. Agentic Adversarial Rewriting Exposes Architectural Vulnerabilities in Black-Box NLP Pipelines

32. Ulterior Motives: Detecting Misaligned Reasoning in Continuous Thought Models

33. IndustryAssetEQA: A Neurosymbolic Operational Intelligence System for Embodied Question Answering in Industrial Asset Maintenance

34. When Corrective Hints Hurt: Prompt Design in Reasoner-Guided Repair of LLM Overcaution on Entailed Negations under OWL~2~DL

35. GSAR: Typed Grounding for Hallucination Detection and Recovery in Multi-Agent LLMs

36. LEGO: An LLM Skill-Based Front-End Design Generation Platform

37. CAP-CoT: Cycle Adversarial Prompt for Improving Chain of Thoughts in LLM Reasoning

38. Discovering Agentic Safety Specifications from 1-Bit Danger Signals

39. From Coarse to Fine: Self-Adaptive Hierarchical Planning for LLM Agents

40. Judging the Judges: A Systematic Evaluation of Bias Mitigation Strategies in LLM-as-a-Judge Pipelines

41. PhySE: A Psychological Framework for Real-Time AR-LLM Social Engineering Attacks

42. Towards Automated Ontology Generation from Unstructured Text: A Multi-Agent LLM Approach

43. Analytica: Soft Propositional Reasoning for Robust and Scalable LLM-Driven Analysis

44. Don’t Make the LLM Read the Graph: Make the Graph Think

45. A Systematic Approach for Large Language Models Debugging

46. FormalScience: Scalable Human-in-the-Loop Autoformalisation of Science with Agentic Code Generation in Lean

47. PExA: Parallel Exploration Agent for Complex Text-to-SQL

48. An Intelligent Fault Diagnosis Method for General Aviation Aircraft Based on Multi-Fidelity Digital Twin and FMEA Knowledge Enhancement

49. Defective Task Descriptions in LLM-Based Code Generation: Detection and Analysis

50. Green Shielding: A User-Centric Approach Towards Trustworthy AI

51. Leveraging LLMs for Multi-File DSL Code Generation: An Industrial Case Study

52. Benchmarking Source-Sensitive Reasoning in Turkish: Humans and LLMs under Evidential Trust Manipulation

53. AgentWard: A Lifecycle Security Architecture for Autonomous AI Agents

54. DepthKV: Layer-Dependent KV Cache Pruning for Long-Context LLM Inference

55. K-MetBench: A Multi-Dimensional Benchmark for Fine-Grained Evaluation of Expert Reasoning, Locality, and Multimodality in Meteorology

56. Less Is More: Engineering Challenges of On-Device Small Language Model Integration in a Mobile Application

57. Learning to Route Queries to Heads for Attention-based Re-ranking with Large Language Models

58. Skill Retrieval Augmentation for Agentic AI

59. Layerwise Convergence Fingerprints for Runtime Misbehavior Detection in Large Language Models

60. Understanding the Limits of Automated Evaluation for Code Review Bots in Practice

61. Why AI Harms Can’t Be Fixed One Identity at a Time: What 5300 Incident Reports Reveal About Intersectionality

62. GAMMAF: A Common Framework for Graph-Based Anomaly Monitoring Benchmarking in LLM Multi-Agent Systems

63. Measuring Successful Cooperation in Human-AI Teamwork: Development and Validation of the Perceived Cooperativity and Teaming Perception Scales

64. Kwai Summary Attention Technical Report

65. Scaling Properties of Continuous Diffusion Spoken Language Models

66. All That Glitters Is Not Audio: Rethinking Text Priors and Audio Reliance in Audio-Language Evaluation

67. Global Context or Local Detail? Adaptive Visual Grounding for Hallucination Mitigation

68. SeaEvo: Advancing Algorithm Discovery with Strategy Space Evolution

69. DPRM: A Plug-in Doob h transform-induced Token-Ordering Module for Diffusion Language Models

70. SycoPhantasy: Quantifying Sycophancy and Hallucination in Small Open Weight VLMs for Vision-Language Scoring of Fantasy Characters

71. See Further, Think Deeper: Advancing VLM’s Reasoning Ability with Low-level Visual Cues and Reflection

72. MEMCoder: Multi-dimensional Evolving Memory for Private-Library-Oriented Code Generation

73. RefEvo: Agentic Design with Co-Evolutionary Verification for Agile Reference Model Generation

74. Agentic Witnessing: Pragmatic and Scalable TEE-Enabled Privacy-Preserving Auditing

75. Rewarding the Scientific Process: Process-Level Reward Modeling for Agentic Data Analysis

76. MultiDx: A Multi-Source Knowledge Integration Framework towards Diagnostic Reasoning

77. MemeScouts@LT-EDI 2026: Asking the Right Questions – Prompted Weak Supervision for Meme Hate Speech Detection

78. Meta-Aligner: Bidirectional Preference-Policy Optimization for Multi-Objective LLMs Alignment

79. AdapTime: Enabling Adaptive Temporal Reasoning in Large Language Models