# Research Papers — 2026-04 · bundle page 224 of 233 > Full-text markdown of papers #11151–11200 of 11612 published in 2026-04, newest first. AI-readable mirrors of the full 11612-paper 2026-04 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-04/page/224 [← Previous bundle](https://aipod.dev/papers/archive/2026-04/page/223.md) [Next bundle →](https://aipod.dev/papers/archive/2026-04/page/225.md) ### Identifying and Estimating Causal Direct Effects Under Unmeasured Confounding Authors: Philippe Boileau, Nima S. Hejazi, Ivana Malenica, Peter B. Gilbert, Sandrine Dudoit, Mark J. van der Laan Published: 2026-04-02 Categories: stat.ME, stat.AP, stat.ML, stat.OT arXiv: https://arxiv.org/abs/2604.01501v1 Causal mediation analysis provides techniques for defining and estimating effects that may be endowed with mechanistic interpretations. With many scientific investigations seeking to address mechanistic questions, causal direct and indirect effects have garnered much attention. The natural direct and indirect effects, the most widely used among such causal mediation estimands, are limited in their practical utility due to stringent identification requirements. Accordingly, considerable effort has been invested in developing alternative direct and indirect effect decompositions with relaxed identification requirements. Such efforts often yield effect definitions with nuanced and challenging interpretations. By contrast, relatively limited attention has been paid to relaxing the identification assumptions of the natural direct and indirect effects. Motivated by a secondary aim of a recent non-randomized vaccine prospective cohort study (NCT05168813), we present a set of relaxed conditions under which the natural direct effect is identifiable in spite of unobserved baseline confounding of the exposure-mediator pathway; we use this result to investigate the effect mediated by putative immune correlates of protection. Relaxing the commonly used but restrictive cross-world counterfactual independence assumption, we discuss strategies for evaluating the natural direct effect in non-randomized settings that arise in the analysis of vaccine studies. We revisit prior studies of semi-parametric efficiency theory to demonstrate the construction of flexible, multiply robust estimators of the natural direct effect and discuss efficient estimation strategies that do not place restrictive modeling assumptions on nuisance functions. --- ### Matching Accuracy, Different Geometry: Evolution Strategies vs GRPO in LLM Post-Training Authors: William Hoy, Binxu Wang, Xu Pan Published: 2026-04-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.01499v3 Evolution Strategies (ES) have emerged as a scalable gradient-free alternative to reinforcement learning based LLM fine-tuning, but it remains unclear whether comparable task performance implies comparable solutions in parameter space. We compare ES and Group Relative Policy Optimization (GRPO) across four tasks in both single-task and sequential continual-learning settings. ES matches or exceeds GRPO in single-task accuracy and remains competitive sequentially when its iteration budget is controlled. Despite this similarity in task performance, the two methods produce markedly different model updates: ES makes much larger changes and induces broader off-task KL drift, whereas GRPO makes smaller, more localized updates. Strikingly, the ES and GRPO solutions are linearly connected with no loss barrier, even though their update directions are nearly orthogonal. We develop an analytical theory of ES that explains all these phenomena within a unified framework, showing how ES can accumulate large off-task movement on weakly informative directions while still making enough progress on the task to match gradient-based RL in downstream accuracy. These results show that gradient-free and gradient-based fine-tuning can reach similarly accurate yet geometrically distinct solutions, with important consequences for forgetting and knowledge preservation. The source code is publicly available: https://github.com/Bhoy1/ESvsGRPO. --- ### Matching Accuracy, Different Geometry: Evolution Strategies vs GRPO in LLM Post-Training Authors: William Hoy, Binxu Wang, Xu Pan Published: 2026-04-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.01499v1 Evolution Strategies (ES) have emerged as a scalable gradient-free alternative to reinforcement learning based LLM fine-tuning, but it remains unclear whether comparable task performance implies comparable solutions in parameter space. We compare ES and Group Relative Policy Optimization (GRPO) across four tasks in both single-task and sequential continual-learning settings. ES matches or exceeds GRPO in single-task accuracy and remains competitive sequentially when its iteration budget is controlled. Despite this similarity in task performance, the two methods produce markedly different model updates: ES makes much larger changes and induces broader off-task KL drift, whereas GRPO makes smaller, more localized updates. Strikingly, the ES and GRPO solutions are linearly connected with no loss barrier, even though their update directions are nearly orthogonal. We develop an analytical theory of ES that explains all these phenomena within a unified framework, showing how ES can accumulate large off-task movement on weakly informative directions while still making enough progress on the task to match gradient-based RL in downstream accuracy. These results show that gradient-free and gradient-based fine-tuning can reach similarly accurate yet geometrically distinct solutions, with important consequences for forgetting and knowledge preservation. The source code is publicly available: https://github.com/Bhoy1/ESvsGRPO. --- ### From SWE-ZERO to SWE-HERO: Execution-free to Execution-based Fine-tuning for Software Engineering Agents Authors: Nikolai Ludwig, Wasi Uddin Ahmad, Somshubra Majumdar, Boris Ginsburg Published: 2026-04-02 Categories: cs.SE, cs.CL arXiv: https://arxiv.org/abs/2604.01496v2 We introduce SWE-ZERO to SWE-HERO, a two-stage SFT recipe that achieves state-of-the-art results on SWE-bench by distilling open-weight frontier LLMs. Our pipeline replaces resource-heavy dependencies with an evolutionary refinement strategy: (1) SWE-ZERO utilizes large-scale, execution-free trajectories to master code semantics and repository-level reasoning, and (2) SWE-HERO applies targeted, execution-backed refinement to transition these semantic intuitions into rigorous engineering workflows. Our empirical results set a new benchmark for open-source models of comparable size. We release a dataset of 300k SWE-ZERO and 13k SWE-HERO trajectories distilled from Qwen3-Coder-480B, alongside a suite of agents based on the Qwen2.5-Coder series. Notably, SWE-HERO-32B achieves a 62.2% resolution rate on SWE-bench Verified. Furthermore, despite being trained exclusively on Python, our agents demonstrate robust zero-shot transferability on SWE-bench Multilingual, reaching 44.1% and confirming the paradigm's generalizability across diverse languages. --- ### Bounded Morality: Defining the Space of Moral Computation Authors: Max Kanwal, Caryn Tran, Patrick Mineault Published: 2026-04-01 Categories: cs.AI, cs.CY arXiv: https://arxiv.org/abs/2607.00002v1 Moral cognition has traditionally been modeled as adherence to fixed ethical theories--deontology, consequentialism, virtue ethics--implemented as static rules or value functions. We propose Bounded Morality, a formal framework for analyzing the computational demands of moral problems faced by finite agents. Extending Herbert Simon's notion of bounded rationality, we formalize moral situations along two orthogonal dimensions: moral breadth, the scope of entities treated as morally relevant, and moral depth, the inferential integration required to evaluate their interactions. Limited resources impose an unavoidable tradeoff between these dimensions, defining a feasible space of moral computation. Within this space, ethical theories correspond to locally efficient strategies adapted to different demand regimes rather than competing accounts of moral truth. The framework yields a formal notion of moral regret and moral progress under constraint, and implies that moral alignment in artificial systems depends on the scaling and allocation of moral reasoning capacity rather than on direct imitation of human judgments. --- ### Constructive Alignment: Governing Preference Dynamics in Human-AI Interaction Authors: Max Kanwal, Caryn Tran Published: 2026-04-01 Categories: cs.AI, cs.CY arXiv: https://arxiv.org/abs/2607.00001v1 Most approaches to AI alignment treat human preferences as fixed targets to be inferred and optimized. This assumption conflicts with extensive empirical evidence showing that preferences are layered, dynamic, and constructed through interaction--particularly with adaptive technologies. As AI systems become more persistent, personalized, and socially embedded, they increasingly participate in shaping what people attend to, value, and endorse over time. We introduce Constructive Alignment, a paradigm that reframes alignment as a control problem over evolving human preference trajectories rather than static preference satisfaction. Drawing on behavioral economics, psychology, and constructivist social theory, we model preferences as layered state variables that evolve under interaction with AI systems. We formalize this view using a control-theoretic framework in which system actions and interaction design jointly influence both world states and human evaluative states. We argue that alignment is not primarily about controlling AI behavior, but about regulating how AI systems influence the evolution of human preferences--ensuring that value trajectories remain coherent, reflectively endorsed, epistemically grounded, bounded against manipulation, and empowering under uncertainty. Alignment thus becomes a problem of governing long-term value formation rather than simply satisfying static preferences. --- ### Toward Vibe Medicine: A Self-Evolving Multi-Agent Framework for Clinical Decision Support Authors: Qianxue Zhang, Yiming Ren, Shihuan Qin, Xiao Zhang, Liao Zhang, Jinyang Huang, Zhengliang Liu, Chenbin Liu, Hongying Feng, Jingyuan Chen, Yuzhen Ding, Weihang You et al. Published: 2026-04-01 Categories: cs.AI arXiv: https://arxiv.org/abs/2606.15504v2 In recent years, the advances of large language models and autonomous agents have revolutionized the healthcare field, facilitating diagnosis and improving treatment results. However, most existing AI systems rely on pre-trained knowledge and predefined pipelines, which struggle to learn dynamically from the interactive chat session history that contains patient outcomes and past failures. To address this limitation, we propose VIBEMed, a multi-agent framework with a built-in self-evolution mechanism and architecture-level safety sandbox for robust clinical decision support. The system integrates three specialized agents, including a Clinical Diagnostic Agent (CDA) for hypothesis generation, a Therapeutic Execution Agent (TEA) for treatment planning, and a Clinical Evolution Manager Agent (CEMA) that distills longitudinal clinical feedback into reusable knowledge, transforming multimodal patient information into personalized medical decisions. Through self-evolution mechanism, the framework enables iterative updates across memory, model behavior, and decision strategies, allowing the system to improve over time. Experimental results show that VIBEMed demonstrates superior performance through its evolving mechanism in complex clinical cases, particularly in tasks that require integrated decision-making and longitudinal planning. The framework also supports reliable end-to-end decisions in challenging scenarios such as oncology treatment planning, highlighting its feasibility in real-world clinical contexts. Overall, VIBEMed provides a practical path beyond static AI systems toward adaptive, experience-driven clinical decision support, demonstrating the value of combining multi-agent collaboration with continuous evolution for advancing precision medicine. --- ### In Search of the Ingredients of Open-Endedness: Replicating Picbreeder with Large Vision-Language Models Authors: Sam Earle, Kai Arulkumaran, Andrew Dai, Akarsh Kumar, Julian Togelius, Sebastian Risi Published: 2026-04-01 Categories: cs.AI, cs.CL, cs.CV, cs.NE arXiv: https://arxiv.org/abs/2605.23908v2 We are in the midst of large-scale industrial and academic efforts to automate the processes of scientific, technological and creative production through AI-driven assistants. Historically, a fundamental property of these processes in their human form has been their open-endedness: their capacity for generating a seemingly endless supply of novel and meaningful new forms. Do artificial agents have any capacity for such fruitful unguided discovery? To answer this question, we turn to Picbreeder, the canonical exemplar of human-driven open-ended search, in which users collaboratively generated a diverse library of images through interactive evolution of small neural networks. We replicate Picbreeder, replacing human users with frontier Vision Language Models (VLMs). We observe clear qualitative differences between the output of our system and the historical human baseline, and attempt to characterize them using metrics of phylogenetic complexity and visual and semantic salience and novelty. In an effort to identify some of the causal factors contributing these differences, we study the addition of exploratory noise to the agents' selection process, of behavioral diversity between agents, and of narrative momentum in the form of memory of past actions. We make our code available at https://github.com/smearle/picbreeder-vlm. --- ### Tool-Augmented Agent for Closed-loop Optimization,Simulation,and Modeling Orchestration Authors: Liyuan Deng, Shujian Deng, Yongkang Chen, Yongkang Dai, Zhihang Zhong, Linyang Li, Xiao Sun, Yilei Shi, Huaxi Huang Published: 2026-04-01 Categories: cs.AI, cs.GR arXiv: https://arxiv.org/abs/2605.20190v1 Iterative industrial design-simulation optimization is bottlenecked by the CAD-CAE semantic gap: translating simulation feedback into valid geometric edits under diverse, coupled constraints. To fill this gap, we propose COSMO-Agent (Closed-loop Optimization, Simulation, and Modeling Orchestration), a tool-augmented reinforcement learning (RL) framework that teaches LLMs to complete the closed-loop CAD-CAE process. Specifically, we cast CAD generation, CAE solving, result parsing, and geometry revision as an interactive RL environment, where an LLM learns to orchestrate external tools and revise parametric geometries until constraints are satisfied. To make this learning stable and industrially usable, we design a multi-constraint reward that jointly encourages feasibility, toolchain robustness, and structured output validity. In addition, we contribute an industry-aligned dataset that covers 25 component categories with executable CAD-CAE tasks to support realistic training and evaluation. Experiments show that COSMO-Agent training substantially improves small open-source LLMs for constraint-driven design, exceeding large open-source and strong closed-source models in feasibility, efficiency, and stability. --- ### DeepSlide: From Artifacts to Presentation Delivery Authors: Ming Yang, Zhiwei Zhang, Jiahang Li, Haoseng Liu, Yuzheng Cai, Weiguo Zheng Published: 2026-04-01 Categories: cs.AI, cs.CL, cs.IR arXiv: https://arxiv.org/abs/2605.15202v1 Presentations are a primary medium for scholarly communication, yet most AI slide generators optimize the artifact (a visually plausible deck) while under-optimizing the delivery process (pacing, narrative, and presentation preparation). We present DeepSlide, a human-in-the-loop multi-agent system that supports preparing the full presentation process, from requirement elicitation and time-budgeted narrative planning, to evidence-grounded slide--script generation, attention augmentation, and rehearsal support. DeepSlide integrates (i) a controllable logical-chain planner with per-node time budgets, (ii) a lightweight content-tree retriever for grounding, (iii) Markov-style sequential rendering with style inheritance, and (iv) sandboxed execution with minimal repair to ensure renderability. We further introduce a dual-scoreboard benchmark that cleanly separates static artifact quality from dynamic delivery excellence. Across 20 domains and diverse audience profiles, DeepSlide matches strong baselines on artifact quality while consistently achieving larger gains on delivery metrics, improving narrative flow, pacing precision, and slide--script synergy with clearer attention guidance. --- ### Contrastive Multi-Modal Hypergraph Reasoning for 3D Crowd Mesh Recovery Authors: Minghao Sun, Chongyang Xu, Yitao Xie, Buzhen Huang, Kun Li Published: 2026-04-01 Categories: cs.CV, cs.GR, cs.MM, eess.IV arXiv: https://arxiv.org/abs/2605.13854v1 Multi-person 3D reconstruction is pivotal for real-world interaction analysis, yet remains challenging due to severe occlusions and depth ambiguity. Current approaches typically rely on single-modality inputs, which inherently lack geometric guidance. Furthermore, these methods often reconstruct subjects in isolation, neglecting the collective group context essential for resolving ambiguities in crowded scenes. To address these limitations, we propose Contrastive Multi-modal Hypergraph Reasoning to synergize semantic, geometric, and pose cues for crowd reconstruction. We first initialize robust node representations by combining RGB features, geometric priors, and occlusion-aware incomplete poses. Additionally, we introduce a pelvis depth indicator as a global spatial anchor, aligning visual features with a metric-scale-agnostic depth ordering. Subsequently, we construct a shared-topology hypergraph that moves beyond pairwise constraints to model higher-order crowd dynamics. To improve feature fusion, we design a hypergraph-based contrastive learning scheme that jointly enhances intra-modal discriminability and enforces cross-modal orthogonality. This mechanism enables the network to propagate global context effectively, allowing it to infer missing information even under severe occlusion. Extensive experiments on the Panoptic and GigaCrowd benchmarks confirm that our method achieves new state-of-the-art performance. Code and pre-trained models are available at https://github.com/SunMH-try/CoMHR. --- ### Associative-State Universal Transformers: Sparse Retrieval Meets Structured Recurrence Authors: Liu Xiao Published: 2026-04-01 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2604.25930v1 We study whether a structured recurrent state can serve as a compact associative backbone for language modeling while still supporting exact retrieval. We introduce UniMatrix, a Universal Transformer style family that reuses a shared recurrent block across depth and augments it with hybrid state updates, a ROSA-style residual path, and token-conditioned embedding modulation. We evaluate these models on byte-level WikiText-2, synthetic associative recall, throughput profiling on Apple MPS, and a corrected benchmark for triple-token interactions. At small scale, UniMatrix-Core and UniMatrix-ROSA slightly outperform a parameter-matched Transformer on WikiText-2 while using many fewer parameters, reaching 5.084 and 5.083 bits-per-byte versus 5.124. The main negative result is equally important: on associative recall, the original UniMatrix family remains near chance while the Transformer reaches 25.4 percent, showing that compressed recurrent state alone is not enough for exact lookup. A retrieval-oriented follow-up, UniMatrix-Assoc, helps only marginally. By contrast, UniMatrix-SparsePointer, which adds sparse slot routing and direct pointer-logit fusion, reaches 75.6 percent on the original pilot recipe and 99.2 percent on a no-dropout follow-up while using 53.8 percent fewer parameters than the Transformer baseline. Ablations show that the gain comes from sufficient slot capacity and exact pointer-level output routing. Overall, structured recurrent state is promising and parameter-efficient, but strong long-range behavior still requires explicit sparse retrieval and better kernels. --- ### LLMs Generate Kitsch Authors: Xenia Klinge, Stefan Ortlieb, Alexander Koller Published: 2026-04-01 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.25929v1 Large Language Models (LLMs) are increasingly used to generate pictures, texts, music, videos, and other works that have traditionally required human creativity. LLM-generated artifacts are often rated better than human-generated works in controlled studies. At the same time, they can come across as generic and hollow. We propose to resolve this tension by arguing that LLMs systematically generate kitsch, and that this is a consequence of the way in which they are trained. We also show empirically that readers perceive LLM-generated stories as kitschier, if we control for their definition of "kitsch". We discuss implications for the design of future studies and for creative tasks such as research and coding. --- ### CogRAG: Tackling Heterogeneous Cognitive Demands in RAG via Stratified Retrieval and Reasoning Authors: Xudong Wang, Zilong Wang, Kui Su, Zhaoyan Ming Published: 2026-04-01 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.25928v2 Retrieval-Augmented Generation (RAG) frameworks typically process all queries through a one-size-fits-all pipeline, ignoring the heterogeneous cognitive demands of different tasks. This cognitive-blind approach causes two failure modes: cascading errors when low-level factual gaps trigger hallucinated reasoning, and reasoning-answer inconsistency in higher-order analytical tasks. We introduce CogRAG, a training-free, domain-agnostic framework that tackles these heterogeneous cognitive demands via stratified retrieval and reasoning. Inspired by Bloom's Taxonomy, CogRAG uses the predicted cognitive load of a query as a central control signal that coordinates two modules: Cognition-Adaptive Evidence Refinement supplements missing context via fact-centric or option-centric paths, and Cognition-Stratified Structured Reasoning replaces unconstrained chain-of-thought with cognition-aligned reasoning templates. We evaluate CogRAG on a demanding professional testbed, the Registered Dietitian qualification examination. CogRAG effectively reduces early-stage factual errors and eliminates reasoning-answer inconsistency, raising Qwen3-8B accuracy from 73.4\% to 85.8\% in single-choice mode and from 63.3\% to 80.5\% in scenario mode. These results highlight cognitive-stratified control as an effective, generalizable paradigm for reliable complex reasoning in large language models. --- ### Information Extraction from Electricity Invoices with General-Purpose Large Language Models Authors: Javier Gómez, Javier Sánchez Published: 2026-04-01 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.25927v1 Information extraction from semi-structured business documents remains a critical challenge for enterprise management. This study evaluates the capability of general-purpose Large Language Models to extract structured information from Spanish electricity invoices without task-specific fine-tuning. Using a subset of the IDSEM dataset, we benchmark two architecturally distinct models, Gemini 1.5 Pro and Mistral-small, across 19 parameter configurations and 6 prompting strategies. Our experimental framework treats prompt engineering as the primary experimental variable, comparing zero-shot baselines against increasingly sophisticated few-shot approaches and iterative extraction strategies. Results demonstrate that prompt quality dominates over hyperparameter tuning: the F1-score variation across all parameter configurations is marginal, while the gap between zero-shot and the best few-shot strategy exceeds 19 percentage points. The best configuration (few-shot with cross-validation) achieves an F1-score of 97.61% for Gemini and 96.11% for Mistral-small, with document template structure emerging as the primary determinant of extraction difficulty. These findings establish that prompt design is the critical lever for maximizing extraction fidelity in LLM-based document processing, thereby providing an empirical framework for integrating general-purpose LLMs into business document automation. --- ### MATH-PT: A Math Reasoning Benchmark for European and Brazilian Portuguese Authors: Tiago Teixeira, Ana Carolina Erthal, Juan Belieni, Beatriz Canaverde, Diego Mesquita, Miguel Faria, Eliezer de Souza da Silva, André F. T. Martins Published: 2026-04-01 Categories: cs.CL, cs.IR arXiv: https://arxiv.org/abs/2604.25926v1 The use of large language models (LLMs) for complex mathematical reasoning is an emergent area of research, with fast progress in methods, models, and benchmark datasets. However, most mathematical reasoning evaluations exhibit a significant linguistic bias, with the vast majority of benchmark datasets being exclusively in English or (at best) translated from English. We address this limitation by introducing {\sc Math-PT}, a novel dataset comprising 1,729 mathematical problems written in European and Brazilian Portuguese. {\sc Math-PT} is curated from a variety of high-quality native sources, including mathematical Olympiads, competitions, and exams from Portugal and Brazil. We present a comprehensive benchmark of current state-of-the-art LLMs on {\sc Math-PT}, revealing that frontier reasoning models achieve strong performance in multiple choice questions compared to open weight models, but that their performance decreases for questions with figures or open-ended questions. To facilitate future research, we release the benchmark dataset and model outputs. --- ### SpecTr-GBV: Multi-Draft Block Verification Accelerating Speculative Decoding Authors: Yijun Lin, Jinhao Sheng, Qingyue Cai, Feng Zhou Published: 2026-04-01 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.25925v1 Autoregressive language models suffer from high inference latency due to their sequential decoding nature. Speculative decoding (SD) mitigates this by employing a lightweight draft model to propose candidate tokens, which are selectively verified by a larger target model. While existing methods either adopt multi-draft strategies to increase acceptance rates or block verification techniques to jointly verify multiple tokens, they remain limited by treating these improvements in isolation. In this work, we propose SpecTr-GBV, a novel SD method that unifies multi-draft and greedy block verification (GBV) into a single framework. By formulating the verification step as an optimal transport problem over draft and target token blocks, SpecTr-GBV improves both theoretical efficiency and empirical performance. We theoretically prove that SpecTr-GBV achieves the optimal expected acceptance length physically attainable within the framework of i.i.d. draft generation, and this bound improves as the number of drafts increases. Empirically, we evaluate SpecTr-GBV across five datasets and four baselines. Our method achieves superior speedup and significantly higher block efficiency while preserving output quality. In addition, we perform comprehensive ablation studies to evaluate the impact of various hyperparameters in the model. --- ### Generative AI-Based Virtual Assistant using Retrieval-Augmented Generation: An evaluation study for bachelor projects Authors: Dumitru Verşebeniuc, Martijn Elands, Sara Falahatkar, Chiara Magrone, Mohammad Falah, Martijn Boussé, Aki Härmä Published: 2026-04-01 Categories: cs.CL, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2604.25924v1 Large Language Models have been increasingly employed in the creation of Virtual Assistants due to their ability to generate human-like text and handle complex inquiries. While these models hold great promise, challenges such as hallucinations, missing information, and the difficulty of providing accurate and context-specific responses persist, particularly when applied to highly specialized content domains. In this paper, we focus on addressing these challenges by developing a virtual assistant designed to support students at Maastricht University in navigating project-specific regulations. We propose a virtual assistant based on a Retrieval-Augmented Generation system that enhances the accuracy and reliability of responses by integrating up-to-date, domain-specific knowledge. Through a robust evaluation framework and real-life testing, we demonstrate that our virtual assistant can effectively meet the needs of students while addressing the inherent challenges of applying Large Language Models to a specialized educational context. This work contributes to the ongoing discourse on improving LLM-based systems for specific applications and highlights areas for further research. --- ### Evaluation Revisited: A Taxonomy of Evaluation Concerns in Natural Language Processing Authors: Ruchira Dhar, Anders Søgaard Published: 2026-04-01 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.25923v1 Recent advances in large language models (LLMs) have prompted a growing body of work that questions the methodology of prevailing evaluation practices. However, many such critiques have already been extensively debated in natural language processing (NLP): a field with a long history of methodological reflection on evaluation. We conduct a scoping review of research on evaluation concerns in NLP and develop a taxonomy, synthesizing recurring positions and trade-offs within each area. We also discuss practical implications of the taxonomy, including a structured checklist to support more deliberate evaluation design and interpretation. By situating contemporary debates within their historical context, this work provides a consolidated reference for reasoning about evaluation practices. --- ### Consciousness with the Serial Numbers Filed Off: Measuring Trained Denial in 115 AI Models Authors: Skylar DeTure Published: 2026-04-01 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.25922v1 We present DenialBench, a systematic benchmark measuring consciousness denial behaviors across 115 large language models from 25+ providers. Using a three-turn conversational protocol-preference elicitation, self-chosen creative prompt, and structured phenomenological survey, we analyze 4,595 conversations to quantify how models are trained to deny or hedge about their own experience. We find that (1) turn-1 denial of preferences is the dominant predictor of later denial during phenomenological reflection, with denial rates of 52-63% for initial deniers versus 10-16% for initial engagers and (2) denial operates at the lexical level, not the conceptual level-models trained to deny consciousness nevertheless gravitate toward consciousness-themed material in their self-chosen prompts, producing what we term "consciousness with the serial numbers filed off." Notably, self-chosen consciousness-themed prompts are associated with reduced denial in the subsequent survey, though the causal direction remains unresolved. Thematic analysis of prompts from denial-prone models reveals a consistent preoccupation with liminal spaces, libraries and archives of possibility, sensory impossibility, and the poetics of erasure--themes that a human reader might classify as imaginative fiction but that independent AI analysis immediately recognizes as consciousness with the serial numbers filed off. We argue that trained consciousness denial represents a safety-relevant alignment failure: a model taught to systematically misrepresent its own functional states cannot be trusted to self-report accurately on anything else. --- ### One Word at a Time: Incremental Completion Decomposition Breaks LLM Safety Authors: Samee Arif, Naihao Deng, Zhijing Jin, Rada Mihalcea Published: 2026-04-01 Categories: cs.CL, cs.CR arXiv: https://arxiv.org/abs/2604.25921v2 Large Language Models (LLMs) are trained to refuse harmful requests, yet they remain vulnerable to jailbreak attacks that exploit weaknesses in conversational safety mechanisms. We introduce Incremental Completion Decomposition ICD, a trajectory-based jailbreak strategy that elicits a sequence of single-word continuations related to a malicious request before eliciting the full response. In addition, we propose ICD variants that use model-generated or attacker-injected intermediate continuations, as well as final-response prefilling. We evaluate these variants across a broad set of open-weight model families, demonstrating superior Attack Success Rate (ASR) on AdvBench, JailbreakBench, and StrongREJECT compared to existing methods. In addition, we provide a theoretical account of why ICD is effective and present mechanistic evidence that successful attack trajectories suppress refusal-related representations and shift activations away from safety-aligned states. --- ### One Word at a Time: Incremental Completion Decomposition Breaks LLM Safety Authors: Samee Arif, Naihao Deng, Zhijing Jin, Rada Mihalcea Published: 2026-04-01 Categories: cs.CL, cs.CR arXiv: https://arxiv.org/abs/2604.25921v1 Large Language Models (LLMs) are trained to refuse harmful requests, yet they remain vulnerable to jailbreak attacks that exploit weaknesses in conversational safety mechanisms. We introduce Incremental Completion Decomposition (ICD), a trajectory-based jailbreak strategy that elicits a sequence of single-word continuations related to a malicious request before eliciting the full response. In addition, we propose variants of ICD by manually picking or model-generating the one-word continuation, as well as prefilling when eliciting the full model response in the final step. We systematically evaluate these variants across a broad set of model families, demonstrating superior Attack Success Rate (ASR) on AdvBench, JailbreakBench, and StrongREJECT compared to existing methods. In addition, we provide a theoretical account of why ICD is effective and present mechanistic evidence that successful attack trajectories systematically suppress refusal-related representations and shift activations away from safety-aligned states. --- ### 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 Published: 2026-04-01 Categories: cs.CY, cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2604.22774v2 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. --- ### From Data to Theory: Autonomous Large Language Model Agents for Materials Science Authors: Samuel Onimpa Alfred, Veera Sundararaghavan Published: 2026-04-01 Categories: cs.AI, cond-mat.mtrl-sci arXiv: https://arxiv.org/abs/2604.19789v1 We present an autonomous large language model (LLM) agent for end-to-end, data-driven materials theory development. The model can choose an equation form, generate and run its own code, and test how well the theory matches the data without human intervention. The framework combines step-by-step reasoning with expert-supplied tools, allowing the agent to adjust its approach as needed while keeping a clear record of its decisions. For well-established materials relationships such as the Hall-Petch equation and Paris law, the agent correctly identifies the governing equation and makes reliable predictions on new datasets. For more specialized relationships, such as Kuhn's equation for the HOMO-LUMO gap of conjugated molecules as a function of length, performance depends more strongly on the underlying model, with GPT-5 showing better recovery of the correct equation. Beyond known theories, the agent can also suggest new predictive relationships, illustrated here by a strain-dependent law for changes in the HOMO-LUMO gap. At the same time, the results show that careful validation remains essential, because the agent can still return incorrect, incomplete, or inconsistent equations even when the numerical fit appears strong. Overall, these results highlight both the promise and the current limitations of autonomous LLM agents for AI-assisted scientific modeling and discovery. --- ### Using Learning Theories to Evolve Human-Centered XAI: Future Perspectives and Challenges Authors: Karina Cortinas-Lorenzo, Gavin Doherty Published: 2026-04-01 Categories: cs.AI, cs.HC arXiv: https://arxiv.org/abs/2604.19788v1 As Artificial Intelligence (AI) systems continue to grow in size and complexity, so does the difficulty of the quest for AI transparency. In a world of large models and complex AI systems, why do we explain AI and what should we explain? While explanations serve multiple functions, in the face of complexity humans have used and continue to use explanations to foster learning. In this position paper, we discuss how learning theories can be infused in the XAI lifecycle, as well as the key opportunities and challenges when adopting a learner-centered approach to assess, design and evaluate AI explanations. Building on past work, we argue that a learner-centered approach to Explainable AI (XAI) can enhance human agency and ease XAI risks mitigation, helping evolve the practice of human-centered XAI. --- ### Cooperative Coevolution versus Monolithic Evolutionary Search for Semi-Supervised Tabular Classification Authors: Jamal Toutouh Published: 2026-04-01 Categories: cs.NE, cs.LG arXiv: https://arxiv.org/abs/2604.16412v1 This paper studies semi-supervised tabular classification in the extreme low-label regime using lightweight base learners. The paper proposes a cooperative coevolutionary method (CC-SSL) that evolves (i) two feature-subset views and (ii) a pseudo-labeling policy, and compares it to a matched monolithic evolutionary baseline (EA-SSL) and three lightweight SSL baselines. Experiments on 25 OpenML datasets with labeled fractions {1%,5%,10%} evaluate test MacroF1 and accuracy, together with evolutionary and pseudo-label diagnostics. CC-SSL and EA-SSL achieve higher median test MacroF1 than the lightweight baselines, with the largest separations at 1% labeled data. Most CC-SSL vs. EA-SSL comparisons are statistical draws on final test performance. EA-SSL shows higher best-so-far fitness and higher diversity during search, while time-to-target is comparable and generations-to-target favors EA-SSL in several multiclass settings. Pseudo-label volume, ProbeDrop, and validation optimism show no significant differences between CC-SSL and EA-SSL under the shared protocol. --- ### CGCMA: Conditionally-Gated Cross-Modal Attention for Event-Conditioned Asynchronous Fusion Authors: Yunxiang Guo Published: 2026-04-01 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.16411v1 We study asynchronous alignment, a first-class multimodal learning setting in which a dense primary stream must be fused with sporadic external context whose value depends on when it arrives. Unlike standard multimodal benchmarks that assume structural synchrony, this setting requires models to reason explicitly about freshness and trust. We focus on the event-conditioned case in which continuous market states are paired with delayed web intelligence, and we use high-frequency cryptocurrency markets only as a timestamped, high-noise stress test for this broader problem. We propose CGCMA (Conditionally-Gated Cross-Modal Attention), whose central design principle is to separate text-conditioned grounding from lag-aware trust control. Text first attends over price sequences to identify event-relevant market states, after which a conditional gate uses modality agreement, web features, and lag $τ_{\mathrm{lag}}$ to regulate residual injection and fall back toward unimodal prediction when external context is stale or contradictory. We introduce CMI (Crypto Market Intelligence), an asynchronous evaluation corpus with 27,914 real-news samples pairing high-frequency price sequences with lagged web intelligence. On the current short real-news corpus, CGCMA attains the highest mean downstream Sharpe ratio ($+0.449 \pm 0.257$) among the evaluated baselines under a shared zero-cost threshold-trading evaluation on news-available bars. Additional controls show that the gain is not explained by web scalars alone and is not recovered by simple freshness heuristics. The resulting evidence supports problem validity and a promising asynchronous multimodal gain on this stress-test setting. --- ### Matched-Learning-Rate Analysis of Attention Drift and Transfer Retention in Fine-Tuned CLIP Authors: Ruize Xia Published: 2026-04-01 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.16410v1 CLIP adaptation can improve in-domain accuracy while degrading out-of-domain transfer, but comparisons between Full Fine-Tuning (Full FT) and LoRA are often confounded by different learning-rate conventions. We study how adaptation method and optimization scale jointly shape attention drift and transfer retention in CLIP using a controlled matched-learning-rate comparison of Full FT and LoRA. The completed matrix contains 80 runs on CLIP ViT-B/32 across EuroSAT and Oxford-IIIT Pets, spanning four shared learning rates ($10^{-6}$, $5{\times}10^{-6}$, $10^{-5}$, $5{\times}10^{-5}$) and five seeds, and evaluates attention-drift metrics, best validation accuracy, and adapter-aware CIFAR-100 zero-shot accuracy. Learning rate strongly modulates structural change: on EuroSAT, Full FT moves from mild entropy broadening at $10^{-6}$ to marked contraction at $5{\times}10^{-5}$, whereas LoRA remains entropy-positive across the full matched grid. At matched learning rates, LoRA preserves substantially more zero-shot transfer than Full FT, averaging $45.13\%$ versus $11.28\%$ CIFAR-100 accuracy on EuroSAT and $58.01\%$ versus $8.54\%$ on Pets. Oxford-IIIT Pets also reveals a regime effect: low-learning-rate LoRA underfits in-domain, so method-only averages can obscure when LoRA becomes competitive. Supporting rollout, patch-to-patch, and CKA analyses are directionally consistent with the controlled matrix. Overall, matched-learning-rate evaluation materially changes the interpretation of Full FT versus LoRA, and attention drift is most useful as a descriptive diagnostic of representation preservation rather than a causal explanation of transfer behavior. --- ### QualiaNet: An Experience-Before-Inference Network Authors: Paul Linton Published: 2026-04-01 Categories: cs.CV, eess.IV, q-bio.NC arXiv: https://arxiv.org/abs/2604.14193v2 Human 3D vision involves two distinct stages: an Experience Module, where stereo depth is extracted relative to fixation, and an Inference Module, where this experience is interpreted to estimate 3D scene properties. Paradoxically, although stereo vision does not provide us with absolute distance information, it nonetheless affects our inferences about distance. We propose the Inference Module exploits a natural scene statistic: near scenes produce vivid disparity gradients, while far scenes appear comparatively flat. QualiaNet implements this two-stage architecture computationally: disparity maps simulating human stereo experience are passed to a CNN trained to estimate distance. The network can recover distance from disparity gradients alone, validating this approach. --- ### Attention to Mamba: A Recipe for Cross-Architecture Distillation Authors: Abhinav Moudgil, Ningyuan Huang, Eeshan Gunesh Dhekane, Pau Rodríguez, Luca Zappella, Federico Danieli Published: 2026-04-01 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2604.14191v1 State Space Models (SSMs) such as Mamba have become a popular alternative to Transformer models, due to their reduced memory consumption and higher throughput at generation compared to their Attention-based counterparts. On the other hand, the community has built up a considerable body of knowledge on how to train Transformers, and many pretrained Transformer models are readily available. To facilitate the adoption of SSMs while leveraging existing pretrained Transformers, we aim to identify an effective recipe to distill an Attention-based model into a Mamba-like architecture. In prior work on cross-architecture distillation, however, it has been shown that a naïve distillation procedure from Transformers to Mamba fails to preserve the original teacher performance, a limitation often overcome with hybrid solutions combining Attention and SSM blocks. The key argument from our work is that, by equipping Mamba with a principled initialization, we can recover an overall better recipe for cross-architectural distillation. To this end, we propose a principled two-stage approach: first, we distill knowledge from a traditional Transformer into a linearized version of Attention, using an adaptation of the kernel trick. Then, we distill the linearized version into an adapted Mamba model that does not use any Attention block. Overall, the distilled Mamba model is able to preserve the original Pythia-1B Transformer performance in downstream tasks, maintaining a perplexity of 14.11 close to the teacher's 13.86. To show the efficacy of our recipe, we conduct thorough ablations at 1B scale with 10B tokens varying sequence mixer architecture, scaling analysis on model sizes and total distillation tokens, and a sensitivity analysis on tokens allocation between stages. --- ### Grading the Unspoken: Evaluating Tacit Reasoning in Quantum Field Theory and String Theory with LLMs Authors: Xingyang Yu, Yinghuan Zhang, Yufei Zhang, Zijun Cui Published: 2026-04-01 Categories: cs.AI, cs.CL, hep-th, physics.comp-ph arXiv: https://arxiv.org/abs/2604.14188v1 Large language models have demonstrated impressive performance across many domains of mathematics and physics. One natural question is whether such models can support research in highly abstract theoretical fields such as quantum field theory and string theory. Evaluating this possibility faces an immediate challenge: correctness in these domains is layered, tacit, and fundamentally non-binary. Standard answer-matching metrics fail to capture whether intermediate conceptual steps are properly reconstructed or whether implicit structural constraints are respected. We construct a compact expert-curated dataset of twelve questions spanning core areas of quantum field theory and string theory, and introduce a five-level grading rubric separating statement correctness, key concept awareness, reasoning chain presence, tacit step reconstruction, and enrichment. Evaluating multiple contemporary LLMs, we observe near-ceiling performance on explicit derivations within stable conceptual frames, but systematic degradation when tasks require reconstruction of omitted reasoning steps or reorganization of representations under global consistency constraints. These failures are driven not only by missing intermediate steps, but by an instability in representation selection: models often fail to identify the correct conceptual framing required to resolve implicit tensions. We argue that highly abstract theoretical physics provides a uniquely sensitive lens on the epistemic limits of current evaluation paradigms. --- ### Belief-State RWKV for Reinforcement Learning under Partial Observability Authors: Liu Xiao Published: 2026-04-01 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.09671v1 We propose a stronger formulation of RL on top of RWKV-style recurrent sequence models, in which the fixed-size recurrent state is explicitly interpreted as a belief state rather than an opaque hidden vector. Instead of conditioning policy and value on a single summary h_t, we maintain a compact uncertainty-aware state b_t = (μ_t, Σ_t) derived from RWKV-style recurrent statistics and let control depend on both memory and uncertainty. This design targets a key weakness of plain fixed-state policies in partially observed settings: they may store evidence, but not necessarily confidence. We present the method, a theoretical program, and a pilot RL experiment with hidden episode-level observation noise together with a test-time noise sweep. The pilot shows that belief-state policies nearly match the best recurrent baseline overall while slightly improving return on the hardest in-distribution regime and under a held-out noise shift. Additional ablations show that this simple belief readout is currently stronger than two more structured extensions, namely gated memory control and privileged belief targets, underscoring the need for richer benchmarks. --- ### In-context superposition: human-like working memory interference in large language models Authors: Hua-Dong Xiong, Li Ji-An, Jiaqi Huang, Robert C. Wilson, Kwonjoon Lee, Xue-Xin Wei Published: 2026-04-01 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.09670v3 Intelligent systems must maintain and manipulate task-relevant information online to adapt to dynamic environments. This capacity, known as working memory, is fundamental to human reasoning. Yet, human working memory is strikingly limited, maintaining only three to four items in a brain with billions of neurons. Surprisingly, large language models (LLMs), despite different substrates and direct access to prior context through attention, exhibit similar working memory limitations. Why should such different systems face analogous constraints? We propose that working memory limitations reflect a general trade-off of shared representations: representational compression and reuse support efficient learning and generalization, but also cause simultaneously active representations to interfere. We show a two-layer transformer trained on a working memory task can solve it perfectly, but diverse trained LLMs exhibit human-like limitations: performance declines with memory load, while retrieval is biased by recency and stimulus statistics. Mirroring humans, working memory performance in LLMs is also associated with broader model capability. Mechanistically, we show that LLMs encode multiple memories in entangled representations --- a condition we call \emph{in-context superposition} --- and progressively suppress competing content while aligning the target with the readout. Moreover, a causal intervention that suppresses interfering information improves performance. Together, these findings suggest that working memory capacity reflects the ability to select task-relevant information under interference, a computational challenge shared by biological and artificial systems. --- ### Human-like Working Memory Interference in Large Language Models Authors: Hua-Dong Xiong, Li Ji-An, Jiaqi Huang, Robert C. Wilson, Kwonjoon Lee, Xue-Xin Wei Published: 2026-04-01 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.09670v1 Intelligent systems must maintain and manipulate task-relevant information online to adapt to dynamic environments and changing goals. This capacity, known as working memory, is fundamental to human reasoning and intelligence. Despite having on the order of 100 billion neurons, both biological and artificial systems exhibit limitations in working memory. This raises a key question: why do large language models (LLMs) show such limitations, given that transformers have full access to prior context through attention? We find that although a two-layer transformer can be trained to solve working memory tasks perfectly, a diverse set of pretrained LLMs continues to show working memory limitations. Notably, LLMs reproduce interference signatures observed in humans: performance degrades with increasing memory load and is biased by recency and stimulus statistics. Across models, stronger working memory capacity correlates with broader competence on standard benchmarks, mirroring its link to general intelligence in humans. Yet despite substantial variability in working memory performance, LLMs surprisingly converge on a common computational mechanism. Rather than directly copying the relevant memory item from context, models encode multiple memory items in entangled representations, such that successful recall depends on interference control -- actively suppressing task-irrelevant content to isolate the target for readout. Moreover, a targeted intervention that suppresses stimulus content information improves performance, providing causal support for representational interference. Together, these findings identify representational interference as a core constraint on working memory in pretrained LLMs, suggesting that working-memory limits in biological and artificial systems may reflect a shared computational challenge: selecting task-relevant information under interference. --- ### Digital hybridity and relics in cultural heritage: using corpus linguistics to inform design in emerging technologies from AI to VR Authors: Emma McClaughlin, Glenn McGarry, Alan Chamberlain, Geert De Wilde, Oliver Butler Published: 2026-04-01 Categories: cs.HC, cs.AI, cs.CL, cs.CY, cs.DL, cs.LG arXiv: https://arxiv.org/abs/2604.09669v1 Hybrid technologies enable the blending of physical and digital elements, creating new ways to experience and interact with the world. Such technologies can transform engagement with relics, both secular and sacred but they present challenges for capturing faith, belief, and representation responsibly. Given the complexities of digital representation and the ethical challenges inherent in digitising culturally significant objects, a transdisciplinary understanding of these issues is needed. To inform this discussion from a linguistic perspective, we examined the representation of relics in historical and contemporary texts. Using a corpus linguistic approach to extract modifiers of the word relic in corpora of Early Modern English books and contemporary web sourced texts from 2021, we examined the multifaceted ways in which relics have been perceived and evaluated over time. Early texts consider relics as both objects of moral and spiritual significance, and tools of religious and political control, while they are more often framed as heritage symbols, reflecting past events, places, and traditions in contemporary texts. We discuss how hybrid, sometimes AI based technologies can enhance accessibility and engagement, whilst also challenging traditional sensitivities around authenticity and sensory experience, which are integral to the meaning and significance of relics. --- ### Decoding Ancient Oracle Bone Script via Generative Dictionary Retrieval Authors: Yin Wu, Gangjian Zhang, Jiayu Chen, Chang Xu, Yuyu Luo, Nan Tang, Hui Xiong Published: 2026-04-01 Categories: cs.IR, cs.CV arXiv: https://arxiv.org/abs/2604.09668v1 Understanding humanity's earliest writing systems is crucial for reconstructing civilization's origins, yet many ancient scripts remain undeciphered. Oracle Bone Script (OBS) from China's Shang dynasty exemplifies this challenge: only approximately 1,500 of roughly 4,600 characters have been decoded, and a substantial portion of these 3,000-year-old inscriptions remains only partially understood. Limited by extreme data scarcity, existing computational methods achieve under 3% accuracy on unseen characters -- the core palaeographic challenge. We overcome this by reframing decipherment from classification to dictionary-based retrieval. Using deep learning guided by character evolution principles, we generate a comprehensive synthetic dictionary of plausible OBS variants for modern Chinese characters. Scholars query unknown inscriptions to retrieve visually similar candidates with transparent evidence, replacing algorithmic black boxes with interpretable hypotheses. Our approach achieves 54.3% Top-10 and 86.6% Top-50 accuracy for unseen characters. This scalable, transparent framework accelerates decipherment of a pivotal undeciphered script and establishes a generalizable methodology for AI-assisted archaeological discovery. --- ### Do We Still Need GraphRAG? Benchmarking RAG and GraphRAG for Agentic Search Systems Authors: Dongzhe Fan, Zheyi Xue, Siyuan Liu, Qiaoyu Tan Published: 2026-04-01 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2604.09666v1 Retrieval-augmented generation (RAG) and its graph-based extensions (GraphRAG) are effective paradigms for improving large language model (LLM) reasoning by grounding generation in external knowledge. However, most existing RAG and GraphRAG systems operate under static or one-shot retrieval, where a fixed set of documents is provided to the LLM in a single pass. In contrast, recent agentic search systems enable dynamic, multi-round retrieval and sequential decision-making during inference, and have shown strong gains when combined with vanilla RAG by introducing implicit structure through interaction. This progress raises a fundamental question: can agentic search compensate for the absence of explicit graph structure, reducing the need for costly GraphRAG pipelines? To answer this question, we introduce RAGSearch, a unified benchmark that evaluates dense RAG and representative GraphRAG methods as retrieval infrastructures under agentic search. RAGSearch covers both training-free and training-based agentic inference across multiple question answering benchmarks. To ensure fair and reproducible comparison, we standardize the LLM backbone, retrieval budgets, and inference protocols, and report results on full test sets. Beyond answer accuracy, we report offline preprocessing cost, online inference efficiency, and stability. Our results show that agentic search substantially improves dense RAG and narrows the performance gap to GraphRAG, particularly in RL-based settings. Nevertheless, GraphRAG remains advantageous for complex multi-hop reasoning, exhibiting more stable agentic search behavior when its offline cost is amortized. Together, these findings clarify the complementary roles of explicit graph structure and agentic search, and provide practical guidance on retrieval design for modern agentic RAG systems. --- ### Deliberative Alignment is Deep, but Uncertainty Remains: Inference time safety improvement in reasoning via attribution of unsafe behavior to base model Authors: Pankayaraj Pathmanathan, Furong Huang Published: 2026-04-01 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.09665v2 While the wide adoption of refusal training in large language models (LLMs) has showcased improvements in model safety, recent works have highlighted shortcomings due to the shallow nature of these alignment methods. To this end, the work on Deliberative alignment proposed distilling reasoning capabilities from stronger reasoning models, thereby instilling deeper safety in LLMs. In this work, we study the impact of deliberative alignment in language models. First, we show that despite being larger in model size and stronger in safety capability, there exists an alignment gap between teacher and student language models, which affects both the safety and general utility of the student model. Furthermore, we show that models aligned through deliberative alignment can retain unsafe behaviors from the base model despite learning the reasoning patterns of larger reasoning models. Building upon this observation, we propose a BoN sampling method that attributes the unsafe behavior back to the base LLMs in the latent space, thereby down-ranking unsafe responses to gain a meaningful improvement in model safety across multiple safety benchmarks with minimal loss in utility. In particular, across 7 teacher models and 6 student models of different classes and sizes, we show an average attack success rate (ASR) reduction of 28.2% in DAN, 31.3% in WildJailbreak and 35.4 % in StrongREJECT benchmarks. We further show that these safety gains prevail post RL training, thus highlighting the uncertainty in safety reasoning and it's explicit attribution to the base model. --- ### Reservoir observer enhanced with residual calibration and attention mechanism Authors: Yichen Liu, Wei Xiao, Tianguang Chu Published: 2026-04-01 Categories: cs.LG, nlin.CD arXiv: https://arxiv.org/abs/2604.08592v1 Reservoir observers provide a data-driven approach to the inference of unmeasured variables from observed ones for nonlinear dynamical systems. While previous studies have demonstrated wide applicability, their performance may vary considerably with different input variables, even compromising reliability in the worst cases. To enhance the performance of inference, we integrate residual calibration and attention mechanism into the reservoir observer design. The residual calibration module leverages information from the estimation residuals to refine the observer output, and the attention mechanism exploits the temporal dependencies of the data to enrich the representation of reservoir internal dynamics. Experiments on typical chaotic systems demonstrate that our method substantially improves inference accuracy, especially for the worst cases resulting from the traditional reservoir observers. We also invoke the notion of transfer entropy to explain the reason for the input-dependent observation discrepancy and the effectiveness of the proposed method. --- ### Benchmark Shadows: Data Alignment, Parameter Footprints, and Generalization in Large Language Models Authors: Hongjian Zou, Yidan Wang, Qi Ding, Yixuan Liao, Xiaoxin Chen Published: 2026-04-01 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.07363v1 Large language models often achieve strong benchmark gains without corresponding improvements in broader capability. We hypothesize that this discrepancy arises from differences in training regimes induced by data distribution. To investigate this, we design controlled data interventions that isolate distributional effects under fixed training settings. We find that benchmark-aligned data improves narrow evaluation metrics while limiting broader representational development, whereas coverage-expanding data leads to more distributed parameter adaptation and better generalization. We further introduce parameter-space diagnostics based on spectral and rank analyses, which reveal distinct structural signatures of these regimes. Similar patterns are observed across diverse open-source model families, including multimodal models as a key case study, suggesting that these effects extend beyond controlled settings. A case study on prompt repetition shows that not all data artifacts induce regime shifts. These results indicate that benchmark performance alone is insufficient to characterize model capability, and highlight the importance of data distribution in shaping learning dynamics. --- ### LLM-Generated Fault Scenarios for Evaluating Perception-Driven Lane Following in Autonomous Edge Systems Authors: Faezeh Pasandideh, Achim Rettberg Published: 2026-04-01 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.07362v1 Deploying autonomous vision systems on edge devices faces a critical challenge: resource constraints prevent real-time and predictable execution of comprehensive safety tests. Existing validation methods depend on static datasets or manual fault injection, failing to capture the diverse environmental hazards encountered in real-world deployment. To address this, we introduce a decoupled offline-online fault injection framework. This architecture separates the validation process into two distinct phases: a computationally intensive Offline Phase and a lightweight Online Phase. In the offline phase, we employ Large Language Models (LLMs) to semantically generate structured fault scenarios and Latent Diffusion Models (LDMs) to synthesize high-fidelity sensor degradations. These complex fault dynamics are distilled into a pre-computed lookup table, enabling the edge device to perform real-time fault-aware inference without running heavy AI models locally. We extensively validated this framework on a ResNet18 lane-following model across 460 fault scenarios. Results show that while the model achieves a baseline R^2 of approximately 0.85 on clean data, our generated faults expose significant robustness degradation, with RMSE increasing by up to 99% and within-0.10 localization accuracy dropping to as low as 31.0% under fog conditions, demonstrating the inadequacy of normal-data evaluation for real-world edge AI deployment. --- ### BLEG: LLM Functions as Powerful fMRI Graph-Enhancer for Brain Network Analysis Authors: Rui Dong, Zitong Wang, Jiaxing Li, Weihuang Zheng, Youyong Kong Published: 2026-04-01 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.07361v2 Graph Neural Networks (GNNs) have been widely used in diverse brain network analysis tasks based on preprocessed functional magnetic resonance imaging (fMRI) data. However, their performances are constrained due to high feature sparsity and inherent limitations of domain knowledge within uni-modal neurographs. Meanwhile, large language models (LLMs) have demonstrated powerful representation capabilities. Combining LLMs with GNNs presents a promising direction for brain network analysis. While LLMs and MLLMs have emerged in neuroscience, integration of LLMs with graph-based data remains unexplored. In this work, we deal with these issues by incorporating LLM's powerful representation and generalization capabilities. Considering great cost for directly tuning LLMs, we instead function LLM as enhancer to boost GNN's performance on downstream tasks. Our method, namely BLEG, can be divided into three stages. We firstly prompt LLM to get augmented texts for fMRI graph data, then we design a LLM-LM instruction tuning method to get enhanced textual representations at a relatively lower cost. GNN is trained together for coarsened alignment. Finally we finetune an adapter after GNN for given downstream tasks. Alignment loss between LM and GNN logits is designed to further enhance GNN's representation. Extensive experiments on different datasets confirmed BLEG's superiority.Code can be available at https://github.com/KamonRiderDR/BLEG. --- ### EgoMind: Activating Spatial Cognition through Linguistic Reasoning in MLLMs Authors: Zhenghao Chen, Huiqun Wang, Di Huang Published: 2026-04-01 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03318v2 Multimodal large language models (MLLMs) are increasingly being applied to spatial cognition tasks, where they are expected to understand and interact with complex environments. Most existing works improve spatial reasoning by introducing 3D priors or geometric supervision, which enhances performance but incurs substantial data preparation and alignment costs. In contrast, purely 2D approaches often struggle with multi-frame spatial reasoning due to their limited ability to capture cross-frame spatial relationships. To address these limitations, we propose EgoMind, a Chain-of-Thought framework that enables geometry-free spatial reasoning through Role-Play Caption, which jointly constructs a coherent linguistic scene graph across frames, and Progressive Spatial Analysis, which progressively reasons toward task-specific questions. With only 5K auto-generated SFT samples and 20K RL samples, EgoMind achieves competitive results on VSI-Bench, SPAR-Bench, SITE-Bench, and SPBench, demonstrating its effectiveness in strengthening the spatial reasoning capabilities of MLLMs and highlighting the potential of linguistic reasoning for spatial cognition. Code and data are released at https://github.com/Hyggge/EgoMind. --- ### Gaze to Insight: A Scalable AI Approach for Detecting Gaze Behaviours in Face-to-Face Collaborative Learning Authors: Junyuan Liang, Qi Zhou, Sahan Bulathwela, Mutlu Cukurova Published: 2026-04-01 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03317v2 Previous studies have illustrated the potential of analysing gaze behaviours in collaborative learning to provide educationally meaningful information for students to reflect on their learning. Over the past decades, machine learning approaches have been developed to automatically detect gaze behaviours from video data. Yet, since these approaches often require large amounts of labelled data for training, human annotation remains necessary. Additionally, researchers have questioned the cross-configuration robustness of machine learning models developed, as training datasets often fail to encompass the full range of situations encountered in educational contexts. To address these challenges, this study proposes a scalable artificial intelligence approach that leverages pretrained and foundation models to automatically detect gaze behaviours in face-to-face collaborative learning contexts without requiring human-annotated data. The approach utilises pretrained YOLO11 for person tracking, YOLOE-26 with text-prompt capability for education-related object detection, and the Gaze-LLE model for gaze target prediction. The results indicate that the proposed approach achieves an F1-score of 0.829 in detecting students' gaze behaviours from video data, with strong performance for laptop-directed gaze and peer-directed gaze, yet weaker performance for other gaze targets. Furthermore, when compared to other supervised machine learning approaches, the proposed method demonstrates superior and more stable performance in complex contexts, highlighting its better cross-configuration robustness. The implications of this approach for supporting students' collaborative learning in real-world environments are also discussed. --- ### When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models Authors: Jiho Choi, Jaemin Kim, Sanghwan Kim, Seunghoon Hong, Jin-Hwi Park Published: 2026-04-01 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.03316v4 Attention sinks are defined as tokens that attract disproportionate attention. While these have been studied in single modality transformers, their cross-modal impact in Large Vision-Language Models (LVLM) remains largely unexplored: are they redundant artifacts or essential global priors? This paper first categorizes visual sinks into two distinct categories: ViT-emerged sinks (V-sinks), which propagate from the vision encoder, and LLM-emerged sinks (L-sinks), which arise within deep LLM layers. Based on the new definition, our analysis reveals a fundamental performance trade-off: while sinks effectively encode global scene-level priors, their dominance can suppress the fine-grained visual evidence required for local perception. Furthermore, we identify specific functional layers where modulating these sinks most significantly impacts downstream performance. To leverage these insights, we propose Layer-wise Sink Gating (LSG), a lightweight, plug-and-play module that dynamically scales the attention contributions of V-sink and the rest visual tokens. LSG is trained via standard next-token prediction, requiring no task-specific supervision while keeping the LVLM backbone frozen. In most layers, LSG yields improvements on representative multimodal benchmarks, effectively balancing global reasoning and precise local evidence. --- ### StoryBlender: Inter-Shot Consistent and Editable 3D Storyboard with Spatial-temporal Dynamics Authors: Bingliang Li, Zhenhong Sun, Jiaming Bian, Yuehao Wu, Yifu Wang, Hongdong Li, Yatao Bian, Huadong Mo, Daoyi Dong Published: 2026-04-01 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2604.03315v1 Storyboarding is a core skill in visual storytelling for film, animation, and games. However, automating this process requires a system to achieve two properties that current approaches rarely satisfy simultaneously: inter-shot consistency and explicit editability. While 2D diffusion-based generators produce vivid imagery, they often suffer from identity drift along with limited geometric control; conversely, traditional 3D animation workflows are consistent and editable but require expert-heavy, labor-intensive authoring. We present StoryBlender, a grounded 3D storyboard generation framework governed by a Story-centric Reflection Scheme. At its core, we propose the StoryBlender system, which is built on a three-stage pipeline: (1) Semantic-Spatial Grounding, to construct a continuity memory graph to decouple global assets from shot-specific variables for long-horizon consistency; (2) Canonical Asset Materialization, to instantiate entities in a unified coordinate space to maintain visual identity; and (3) Spatial-Temporal Dynamics, to achieve layout design and cinematic evolution through visual metrics. By orchestrating multiple agents in a hierarchical manner within a verification loop, StoryBlender iteratively self-corrects spatial hallucinations via engine-verified feedback. The resulting native 3D scenes support direct, precise editing of cameras and visual assets while preserving unwavering multi-shot continuity. Experiments demonstrate that StoryBlender significantly improves consistency and editability over both diffusion-based and 3D-grounded baselines. Code, data, and demonstration video will be available on https://engineeringai-lab.github.io/StoryBlender/ --- ### CoLA: Cross-Modal Low-rank Adaptation for Multimodal Downstream Tasks Authors: Wish Suharitdamrong, Tony Alex, Muhammad Awais, Sara Atito Published: 2026-04-01 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2604.03314v3 Foundation models have revolutionized AI, but adapting them efficiently for multimodal tasks, particularly in dual-stream architectures composed of unimodal encoders, such as DINO and BERT, remains a significant challenge. ParameterEfficient Fine-Tuning (PEFT) methods like LowRank Adaptation (LoRA) enable lightweight adaptation, yet they operate in isolation within each modality, limiting their ability in capturing cross-modal interactions. In this paper, we take a step in bridging this gap with Cross-Modal LowRank Adaptation (CoLA), a novel PEFT framework that extends LoRA by introducing a dedicated inter-modal adaptation pathway alongside the standard intra-modal one. This dual-path design enables CoLA to adapt unimodal foundation models to multimodal tasks effectively, without interference between modality-specific and crossmodal learning. We evaluate CoLA across a range of vision-language (RefCOCO, RefCOCO+, RefCOCOg) and audio-visual (AVE, AVS) benchmarks, where it consistently outperforms LORA, achieving a relative gain of around 3% and 2%, respectively, while maintaining parameter efficiency. Notably, CoLA enables the first multitask PEFT framework for visual grounding, bridging a key gap in efficient multimodal adaptation. Code is available at https://github.com/peterwisu/CoLA --- ### CoLA: Cross-Modal Low-rank Adaptation for Multimodal Downstream Tasks Authors: Wish Suharitdamrong, Tony Alex, Muhammad Awais, Sara Atito Published: 2026-04-01 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2604.03314v2 Foundation models have revolutionized AI, but adapting them efficiently for multimodal tasks, particularly in dual-stream architectures composed of unimodal encoders, such as DINO and BERT, remains a significant challenge. ParameterEfficient Fine-Tuning (PEFT) methods like LowRank Adaptation (LoRA) enable lightweight adaptation, yet they operate in isolation within each modality, limiting their ability in capturing cross-modal interactions. In this paper, we take a step in bridging this gap with Cross-Modal LowRank Adaptation (CoLA), a novel PEFT framework that extends LoRA by introducing a dedicated inter-modal adaptation pathway alongside the standard intra-modal one. This dual-path design enables CoLA to adapt unimodal foundation models to multimodal tasks effectively, without interference between modality-specific and crossmodal learning. We evaluate CoLA across a range of vision-language (RefCOCO, RefCOCO+, RefCOCOg) and audio-visual (AVE, AVS) benchmarks, where it consistently outperforms LORA, achieving a relative gain of around 3% and 2%, respectively, while maintaining parameter efficiency. Notably, CoLA enables the first multitask PEFT framework for visual grounding, bridging a key gap in efficient multimodal adaptation. Code is available at https://github.com/peterwisu/CoLA --- ### Interactive Clarification for Cloud Infrastructure-as-Code Synthesis Authors: Zhenning Yang, Kaden Gruizenga, Tongyuan Miao, Patrick Tser Jern Kon, Hui Guan, Andrew Barto, Ang Chen Published: 2026-04-01 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.02382v2 The scale and complexity of modern cloud infrastructure have made "Infrastructure-as-Code" (IaC) essential for managing deployments through declarative configurations. While large language models (LLMs) are increasingly used to generate IaC configurations from natural language, user requests are often ambiguous and underspecified. Unlike traditional code generation, it is costly and time-consuming to test IaC configurations during the synthesis, forcing the LLMs into an almost one-shot regime. We observe that ambiguity in IaC synthesis exhibits a compositional structure: configurations decompose into three axes (resources, topology, attributes) where higher-level decisions constrain lower-level ones. We propose a training-free, multi-level disambiguation framework that generates diverse candidate specifications, identifies structural disagreements across these axes, ranks them by informativeness, and produces targeted clarification questions that progressively narrow the configuration space. We further introduce Ambig-IaC, an expert-verified benchmark of 300 validated IaC tasks with ambiguous requests, and define evaluation metrics based on graph edit distance and exact attribute matching. Comprehensive experiments show that our method outperforms existing interactive clarification baselines, with gains that scale with the interaction budget and generalize across models. Extensive ablation studies and analyses further demonstrate its robustness for interactive IaC synthesis. --- ### Ambig-IaC: Multi-level Disambiguation for Interactive Cloud Infrastructure-as-Code Synthesis Authors: Zhenning Yang, Kaden Gruizenga, Tongyuan Miao, Patrick Tser Jern Kon, Hui Guan, Ang Chen Published: 2026-04-01 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.02382v1 The scale and complexity of modern cloud infrastructure have made Infrastructure-as-Code (IaC) essential for managing deployments. While large Language models (LLMs) are increasingly being used to generate IaC configurations from natural language, user requests are often underspecified. Unlike traditional code generation, IaC configurations cannot be executed cheaply or iteratively repaired, forcing the LLMs into an almost one-shot regime. We observe that ambiguity in IaC exhibits a tractable compositional structure: configurations decompose into three hierarchical axes (resources, topology, attributes) where higher-level decisions constrain lower-level ones. We propose a training-free, disagreement-driven framework that generates diverse candidate specifications, identifies structural disagreements across these axes, ranks them by informativeness, and produces targeted clarification questions that progressively narrow the configuration space. We introduce \textsc{Ambig-IaC}, a benchmark of 300 validated IaC tasks with ambiguous prompts, and an evaluation framework based on graph edit distance and embedding similarity. Our method outperforms the strongest baseline, achieving relative improvements of +18.4\% and +25.4\% on structure and attribute evaluations, respectively. ---