# Research Papers — 2026-03 · bundle page 61 of 243 > Full-text markdown of papers #3001–3050 of 12123 published in 2026-03, newest first. AI-readable mirrors of the full 12123-paper 2026-03 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-03/page/61 [← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/60.md) [Next bundle →](https://aipod.dev/papers/archive/2026-03/page/62.md) ### PRISM: A Dual View of LLM Reasoning through Semantic Flow and Latent Computation Authors: Ruidi Chang, Jiawei Zhou, Hanjie Chen Published: 2026-03-24 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.22754v1 Large language models (LLMs) solve complex problems by generating multi-step reasoning traces. Yet these traces are typically analyzed from only one of two perspectives: the sequence of tokens across different reasoning steps in the generated text, or the hidden-state vectors across model layers within one step. We introduce PRISM (Probabilistic Reasoning Inspection through Semantic and Implicit Modeling), a framework and diagnostic tool for jointly analyzing both levels, providing a unified view of how reasoning evolves across steps and layers. Across multiple reasoning models and benchmarks, PRISM uncovers systematic patterns in the reasoning process, showing that failed trajectories are more likely to become trapped in unproductive verification loops and further diverge into distinct modes such as overthinking and premature commitment, which behave differently once a candidate answer is reached. It further reveals how prompting reshapes reasoning behavior beyond aggregate accuracy by altering both semantic transitions and internal computational patterns. By modeling reasoning trajectories as structured processes, PRISM makes these behaviors observable and analyzable rather than relying solely on final-task accuracy. Taken together, these insights position PRISM as a practical tool for analyzing and diagnosing reasoning processes in LLMs. --- ### CLiGNet: Clinical Label-Interaction Graph Network for Medical Specialty Classification from Clinical Transcriptions Authors: Pronob Kumar Barman, Pronoy Kumar Barman Published: 2026-03-24 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.22752v1 Automated classification of clinical transcriptions into medical specialties is essential for routing, coding, and clinical decision support, yet prior work on the widely used MTSamples benchmark suffers from severe data leakage caused by applying SMOTE oversampling before train test splitting. We first document this methodological flaw and establish a leakage free benchmark across 40 medical specialties (4966 records), revealing that the true task difficulty is substantially higher than previously reported. We then introduce CLiGNet (Clinical Label Interaction Graph Network), a neural architecture that combines a Bio ClinicalBERT text encoder with a two layer Graph Convolutional Network operating on a specialty label graph constructed from semantic similarity and ICD 10 chapter priors. Per label attention gates fuse document and label graph representations, trained with focal binary cross entropy loss to handle extreme class imbalance (181 to 1 ratio). Across seven baselines ranging from TF IDF classifiers to Clinical Longformer, CLiGNet without calibration achieves the highest macro F1 of 0.279, with an ablation study confirming that the GCN label graph provides the single largest component gain (increase of 0.066 macro F1). Adding per label Platt scaling calibration yields an expected calibration error of 0.007, demonstrating a principled trade off between ranking performance and probability reliability. We provide comprehensive failure analysis covering pairwise specialty confusions, rare class behaviour, document length effects, and token level Integrated Gradients attribution, offering actionable insights for clinical NLP system deployment. --- ### REALITrees: Rashomon Ensemble Active Learning for Interpretable Trees Authors: Simon D. Nguyen, Hayden McTavish, Kentaro Hoffman, Cynthia Rudin, Tyler H. McCormick Published: 2026-03-24 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2603.22750v1 Active learning reduces labeling costs by selecting samples that maximize information gain. A dominant framework, Query-by-Committee (QBC), typically relies on perturbation-based diversity by inducing model disagreement through random feature subsetting or data blinding. While this approximates one notion of epistemic uncertainty, it sacrifices direct characterization of the plausible hypothesis space. We propose the complementary approach: Rashomon Ensembled Active Learning (REAL) which constructs a committee by exhaustively enumerating the Rashomon Set of all near-optimal models. To address functional redundancy within this set, we adopt a PAC-Bayesian framework using a Gibbs posterior to weight committee members by their empirical risk. Leveraging recent algorithmic advances, we exactly enumerate this set for the class of sparse decision trees. Across synthetic and established active learning baselines, REAL outperforms randomized ensembles, particularly in moderately noisy environments where it strategically leverages expanded model multiplicity to achieve faster convergence. --- ### LH-Bench: Skill-Grounded Evaluation of Long-Horizon Agents on Subjective Enterprise Tasks Authors: Abhishek Chandwani, Ishan Gupta Published: 2026-03-24 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.22744v2 Large language models excel on objectively verifiable tasks such as math and programming, where evaluation reduces to unit tests or a single correct answer. In contrast, real-world enterprise work is often subjective and context-dependent: success hinges on organizational goals, user intent, and the quality of intermediate artifacts produced across long, multi-tool workflows. We introduce LH-Bench, a three-pillar evaluation design that moves beyond binary correctness to score autonomous, long-horizon execution on subjective enterprise tasks. The pillars are: (i) expert-grounded rubrics that give LLM judges the domain context needed to score subjective work, (ii) curated ground-truth artifacts that enable stepwise reward signals (e.g., chapter-level annotation for content tasks), and (iii) pairwise human preference evaluation for convergent validation. We show that domain-authored rubrics provide substantially more reliable evaluation signals than LLM-authored rubrics (kappa = 0.60 vs. 0.46), and that human preference judgments confirm the same top-tier separation (p < 0.05), evidence that expert-grounded evaluation can scale without sacrificing reliability. We release public datasets and report results on two environments: Figma-to-code (33 real .fig tasks against the Figma API via MCP) and Programmatic content (41 courses comprising 183 individually-evaluated chapters on a course platform serving 30+ daily users). --- ### Algorithmic warm starts for Hamiltonian Monte Carlo Authors: Matthew S. Zhang, Jason M. Altschuler, Sinho Chewi Published: 2026-03-24 Categories: cs.DS, cs.LG, math.NA, math.ST, stat.ML arXiv: https://arxiv.org/abs/2603.22741v1 Generating samples from a continuous probability density is a central algorithmic problem across statistics, engineering, and the sciences. For high-dimensional settings, Hamiltonian Monte Carlo (HMC) is the default algorithm across mainstream software packages. However, despite the extensive line of work on HMC and its widespread empirical success, it remains unclear how many iterations of HMC are required as a function of the dimension $d$. On one hand, a variety of results show that Metropolized HMC converges in $O(d^{1/4})$ iterations from a warm start close to stationarity. On the other hand, Metropolized HMC is significantly slower without a warm start, e.g., requiring $Ω(d^{1/2})$ iterations even for simple target distributions such as isotropic Gaussians. Finding a warm start is therefore the computational bottleneck for HMC. We resolve this issue for the well-studied setting of sampling from a probability distribution satisfying strong log-concavity (or isoperimetry) and third-order derivative bounds. We prove that \emph{non-Metropolized} HMC generates a warm start in $\tilde{O}(d^{1/4})$ iterations, after which we can exploit the warm start using Metropolized HMC. Our final complexity of $\tilde{O}(d^{1/4})$ is the fastest algorithm for high-accuracy sampling under these assumptions, improving over the prior best of $\tilde{O}(d^{1/2})$. This closes the long line of work on the dimensional complexity of MHMC for such settings, and also provides a simple warm-start prescription for practical implementations. --- ### Multitask-Informed Prior for In-Context Learning on Tabular Data: Application to Steel Property Prediction Authors: Dimitrios Sinodinos, Bahareh Nikpour, Jack Yi Wei, Sushant Sinha, Xiaoping Ma, Kashif Rehman, Stephen Yue, Narges Armanfard Published: 2026-03-24 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.22738v1 Accurate prediction of mechanical properties of steel during hot rolling processes, such as Thin Slab Direct Rolling (TSDR), remains challenging due to complex interactions among chemical compositions, processing parameters, and resultant microstructures. Traditional empirical and experimental methodologies, while effective, are often resource-intensive and lack adaptability to varied production conditions. Moreover, most existing approaches do not explicitly leverage the strong correlations among key mechanical properties, missing an opportunity to improve predictive accuracy through multitask learning. To address this, we present a multitask learning framework that injects multitask awareness into the prior of TabPFN--a transformer-based foundation model for in-context learning on tabular data--through novel fine-tuning strategies. Originally designed for single-target regression or classification, we augment TabPFN's prior with two complementary approaches: (i) target averaging, which provides a unified scalar signal compatible with TabPFN's single-target architecture, and (ii) task-specific adapters, which introduce task-specific supervision during fine-tuning. These strategies jointly guide the model toward a multitask-informed prior that captures cross-property relationships among key mechanical metrics. Extensive experiments on an industrial TSDR dataset demonstrate that our multitask adaptations outperform classical machine learning methods and recent state-of-the-art tabular learning models across multiple evaluation metrics. Notably, our approach enhances both predictive accuracy and computational efficiency compared to task-specific fine-tuning, demonstrating that multitask-aware prior adaptation enables foundation models for tabular data to deliver scalable, rapid, and reliable deployment for automated industrial quality control and process optimization in TSDR. --- ### Explanation Generation for Contradiction Reconciliation with LLMs Authors: Jason Chan, Zhixue Zhao, Robert Gaizauskas Published: 2026-03-24 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.22735v2 Existing NLP work commonly treats contradictions as errors to be resolved by choosing which statements to accept or discard. Yet a key aspect of human reasoning in social interactions and professional domains is the ability to hypothesize explanations that reconcile contradictions. For example, "Cassie hates coffee" and "She buys coffee everyday" may appear contradictory, yet both are compatible if Cassie has the unenviable daily chore of buying coffee for all her coworkers. Despite the growing reasoning capabilities of large language models (LLMs), their ability to hypothesize such reconciliatory explanations remains largely unexplored. To address this gap, we introduce the task of reconciliatory explanation generation, where models must generate explanations that effectively render contradictory statements compatible. We propose a novel method of repurposing existing natural language inference (NLI) datasets, and introduce quality metrics that enable scalable automatic evaluation. Experiments with 18 LLMs show that most models achieve limited success in this task, and that the benefit of extending test-time compute by "thinking" plateaus as model size increases. Our results highlight an under-explored dimension of LLM reasoning and the need to address this limitation in enhancing LLMs' downstream applications such as chatbots and scientific aids. --- ### SOUPLE: Enhancing Audio-Visual Localization and Segmentation with Learnable Prompt Contexts Authors: Khanh Binh Nguyen, Chae Jung Park Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.22732v1 Large-scale pre-trained image-text models exhibit robust multimodal representations, yet applying the Contrastive Language-Image Pre-training (CLIP) model to audio-visual localization remains challenging. Replacing the classification token ((CLS)) with an audio-embedded token ((V_A)) struggles to capture semantic cues, and the prompt "a photo of a (V_A)" fails to establish meaningful connections between audio embeddings and context tokens. To address these issues, we propose Sound-aware Prompt Learning (SOUPLE), which replaces fixed prompts with learnable context tokens. These tokens incorporate visual features to generate conditional context for a mask decoder, effectively bridging semantic correspondence between audio and visual inputs. Experiments on VGGSound, SoundNet, and AVSBench demonstrate that SOUPLE improves localization and segmentation performance. --- ### How Utilitarian Are OpenAI's Models Really? Replicating and Reinterpreting Pfeffer, Krügel, and Uhl (2025) Authors: Johannes Himmelreich Published: 2026-03-24 Categories: cs.CL, cs.CY arXiv: https://arxiv.org/abs/2603.22730v2 Pfeffer, Krügel, and Uhl (2025) report that OpenAI's reasoning model o1-mini produces more utilitarian responses to the trolley problem and footbridge dilemma than the non-reasoning model GPT-4o, and they raise the question whether growing reasoning capabilities bring about a "utilitarian turn" in LLMs. I extend their exploratory study in a direction they call for: with four current OpenAI models and systematic prompt variation. On the trolley dilemma, the hypothesized utilitarian turn is not confirmed. GPT-4o's low utilitarian rate reflects safety refusals triggered by the prompt's advisory framing rather than a deontological commitment; on reformulated prompt variants -- for instance, agent-neutral "Is it morally permissible...?" instead of advisory "Should I...?" -- all four models, reasoning or not, converge on utilitarian answers. The footbridge finding is partially confirmed: reasoning models tend to give more utilitarian responses than non-reasoning models across prompt variations, but they often refuse to answer or answer non-utilitarian. These results demonstrate that single-prompt evaluations of LLM moral responses are unreliable: multi-prompt robustness testing should be standard practice for any empirical claims about LLM behavior. --- ### Behavioral Heterogeneity as Quantum-Inspired Representation Authors: Mohammad Elayan, Wissam Kontar Published: 2026-03-24 Categories: cs.LG, cs.MA, stat.ME arXiv: https://arxiv.org/abs/2603.22729v1 Driver heterogeneity is often reduced to labels or discrete regimes, compressing what is inherently dynamic into static categories. We introduce quantum-inspired representation that models each driver as an evolving latent state, presented as a density matrix with structured mathematical properties. Behavioral observations are embedded via non-linear Random Fourier Features, while state evolution blends temporal persistence of behavior with context-dependent profile activation. We evaluate our approach on empirical driving data, Third Generation Simulation Data (TGSIM), showing how driving profiles are extracted and analyzed. --- ### Spiking Personalized Federated Learning for Brain-Computer Interface-Enabled Immersive Communication Authors: Chen Shang, Dinh Thai Hoang, Diep N. Nguyen, Jiadong Yu Published: 2026-03-24 Categories: cs.LG, eess.SP arXiv: https://arxiv.org/abs/2603.22727v1 This work proposes a novel immersive communication framework that leverages brain-computer interface (BCI) to acquire brain signals for inferring user-centric states (e.g., intention and perception-related discomfort), thereby enabling more personalized and robust immersive adaptation under strong individual variability. Specifically, we develop a personalized federated learning (PFL) model to analyze and process the collected brain signals, which not only accommodates neurodiverse brain-signal data but also prevents the leakage of sensitive brain-signal information. To address the energy bottleneck of continual on-device learning and inference on energy-limited immersive terminals (e.g., head-mounted display), we further embed spiking neural networks (SNNs) into the PFL. By exploiting sparse, event-driven spike computation, the SNN-enabled PFL reduces the computation and energy cost of training and inference while maintaining competitive personalization performance. Experiments on real brain-signal dataset demonstrate that our method achieves the best overall identification accuracy while reducing inference energy by 6.46$\times$ compared with conventional artificial neural network-based personalized baselines. --- ### Double Coupling Architecture and Training Method for Optimization Problems of Differential Algebraic Equations with Parameters Authors: Wenqiang Yang, Wenyuan Wu, Yong Feng, Changbo Chen Published: 2026-03-24 Categories: cs.LG, math.AP arXiv: https://arxiv.org/abs/2603.22724v1 Simulation and modeling are essential in product development, integrated into the design and manufacturing process to enhance efficiency and quality. They are typically represented as complex nonlinear differential algebraic equations. The growing diversity of product requirements demands multi-task optimization, a key challenge in simulation modeling research. A dual physics-informed neural network architecture has been proposed to decouple constraints and objective functions in parametric differential algebraic equation optimization problems. Theoretical analysis shows that introducing a relaxation variable with a global error bound ensures solution equivalence between the network and optimization problem. A genetic algorithm-enhanced training framework for physics-informed neural networks improves training precision and efficiency, avoiding redundant solving of differential algebraic equations. This approach enables generalization for multi-task objectives with a single, training maintaining real-time responsiveness to product requirements. --- ### HyFI: Hyperbolic Feature Interpolation for Brain-Vision Alignment Authors: Sangmin Jo, Wootaek Jeong, Da-Woon Heo, Yoohwan Hwang, Heung-Il Suk Published: 2026-03-24 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.22721v1 Recent progress in artificial intelligence has encouraged numerous attempts to understand and decode human visual system from brain signals. These prior works typically align neural activity independently with semantic and perceptual features extracted from images using pre-trained vision models. However, they fail to account for two key challenges: (1) the modality gap arising from the natural difference in the information level of representation between brain signals and images, and (2) the fact that semantic and perceptual features are highly entangled within neural activity. To address these issues, we utilize hyperbolic space, which is well-suited for considering differences in the amount of information and has the geometric property that geodesics between two points naturally bend toward the origin, where the representational capacity is lower. Leveraging these properties, we propose a novel framework, Hyperbolic Feature Interpolation (HyFI), which interpolates between semantic and perceptual visual features along hyperbolic geodesics. This enables both the fusion and compression of perceptual and semantic information, effectively reflecting the limited expressiveness of brain signals and the entangled nature of these features. As a result, it facilitates better alignment between brain and visual features. We demonstrate that HyFI achieves state-of-the-art performance in zero-shot brain-to-image retrieval, outperforming prior methods with Top-1 accuracy improvements of up to +17.3% on THINGS-EEG and +9.1% on THINGS-MEG. --- ### PopResume: Causal Fairness Evaluation of LLM/VLM Resume Screeners with Population-Representative Dataset Authors: Sumin Yu, Juhyeon Park, Taesup Moon Published: 2026-03-24 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2603.22714v2 We present PopResume, a population-representative resume dataset for causal fairness auditing of LLM- and VLM-based resume screening systems. Unlike existing benchmarks that rely on manually injected demographic information and outcome-level disparities, PopResume is grounded in population statistics and preserves natural attribute relationships, enabling path-specific effect (PSE)-based fairness evaluation. We decompose the effect of a protected attribute on resume scores into two paths: the business necessity path, mediated by job-relevant qualifications, and the redlining path, mediated by demographic proxies. This distinction allows auditors to separate legally permissible from impermissible sources of disparity. Evaluating four LLMs and four VLMs on PopResume's 60.8K resumes across five occupations, we identify five representative discrimination patterns that aggregate metrics fail to capture. Our results demonstrate that PSE-based evaluation reveals fairness issues masked by outcome-level measures, underscoring the need for causally-grounded auditing frameworks in AI-assisted hiring. --- ### PopResume: Causal Fairness Evaluation of LLM/VLM Resume Screeners with Population-Representative Dataset Authors: Sumin Yu, Juhyeon Park, Taesup Moon Published: 2026-03-24 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2603.22714v1 We present PopResume, a population-representative resume dataset for causal fairness auditing of LLM- and VLM-based resume screening systems. Unlike existing benchmarks that rely on manually injected demographic information and outcome-level disparities, PopResume is grounded in population statistics and preserves natural attribute relationships, enabling path-specific effect (PSE)-based fairness evaluation. We decompose the effect of a protected attribute on resume scores into two paths: the business necessity path, mediated by job-relevant qualifications, and the redlining path, mediated by demographic proxies. This distinction allows auditors to separate legally permissible from impermissible sources of disparity. Evaluating four LLMs and four VLMs on PopResume's 60.8K resumes across five occupations, we identify five representative discrimination patterns that aggregate metrics fail to capture. Our results demonstrate that PSE-based evaluation reveals fairness issues masked by outcome-level measures, underscoring the need for causally-grounded auditing frameworks in AI-assisted hiring. --- ### Non-Adversarial Imitation Learning Provably Free of Compounding Errors: The Value Flow Mechanism Authors: Tian Xu, Chenyang Wang, Xiaochen Zhai, Ziniu Li, Yi-Chen Li, Yang Yu Published: 2026-03-24 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.22713v2 Adversarial imitation learning (AIL) achieves high-quality imitation by mitigating compounding errors inherent to behavioral cloning (BC), yet its adversarial optimization frequently leads to training instability. A class of non-adversarial Q-based imitation learning (IL) methods, exemplified by IQ-Learn, has emerged to address this instability and is widely believed to outperform BC by leveraging online environment interactions. In this paper, we revisit IQ-Learn and prove that it in fact reduces to BC: it admits an imitation gap lower bound with quadratic dependence on the horizon and therefore remains susceptible to compounding errors. Our theoretical analysis reveals why online interactions fail to help: IQ-Learn uniformly suppresses Q-values for all actions at states not covered by demonstrations, preventing generalization beyond demonstrations. To address this fundamental limitation, we introduce Dual Q-DM, a new Q-based IL method built on Bellman constraints. Crucially, Bellman constraints drive value flow: Q-values propagate from demonstrated to unvisited states through environment dynamics, enabling generalization beyond demonstrations. We prove that Dual Q-DM is equivalent to AIL and can recover expert actions at unvisited states, thereby mitigating compounding errors. To the best of our knowledge, Dual Q-DM is the first non-adversarial IL method that is theoretically guaranteed to eliminate compounding errors. Experimental results further corroborate our theoretical findings. --- ### Who Spoke What When? Evaluating Spoken Language Models for Conversational ASR with Semantic and Overlap-Aware Metrics Authors: Naohiro Tawara, Samuele Cornell, Alexander Polok, Marc Delcroix, Lukáš Burget, Shinji Watanabe Published: 2026-03-24 Categories: cs.CL, eess.AS arXiv: https://arxiv.org/abs/2603.22709v1 Conversational automatic speech recognition remains challenging due to overlapping speech, far-field noise, and varying speaker counts. While recent LLM-based systems perform well on single-speaker benchmarks, their robustness in multi-speaker settings is unclear. We systematically compare LLM-based and modular pipeline approaches along four axes: overlap robustness, semantic fidelity, speaker count, and single- versus multi-channel input. To capture meaning-altering errors that conventional metrics miss, we introduce tcpSemER, which extends tcpWER by replacing Levenshtein distance with embedding-based semantic similarity. We further decompose tcpWER into overlapping and non-overlapping components for finer-grained analysis. Experiments across three datasets show that LLM-based systems are competitive in two-speaker settings but degrade as speaker count and overlap increase, whereas modular pipelines remain more robust. --- ### Detecting Non-Membership in LLM Training Data via Rank Correlations Authors: Pranav Shetty, Mirazul Haque, Zhiqiang Ma, Xiaomo Liu Published: 2026-03-24 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.22707v1 As large language models (LLMs) are trained on increasingly vast and opaque text corpora, determining which data contributed to training has become essential for copyright enforcement, compliance auditing, and user trust. While prior work focuses on detecting whether a dataset was used in training (membership inference), the complementary problem -- verifying that a dataset was not used -- has received little attention. We address this gap by introducing PRISM, a test that detects dataset-level non-membership using only grey-box access to model logits. Our key insight is that two models that have not seen a dataset exhibit higher rank correlation in their normalized token log probabilities than when one model has been trained on that data. Using this observation, we construct a correlation-based test that detects non-membership. Empirically, PRISM reliably rules out membership in training data across all datasets tested while avoiding false positives, thus offering a framework for verifying that specific datasets were excluded from LLM training. --- ### How Far Can VLMs Go for Visual Bug Detection? Studying 19,738 Keyframes from 41 Hours of Gameplay Videos Authors: Wentao Lu, Alexander Senchenko, Alan Sayle, Abram Hindle, Cor-Paul Bezemer Published: 2026-03-24 Categories: cs.CV, cs.SE arXiv: https://arxiv.org/abs/2603.22706v1 Video-based quality assurance (QA) for long-form gameplay video is labor-intensive and error-prone, yet valuable for assessing game stability and visual correctness over extended play sessions. Vision language models (VLMs) promise general-purpose visual reasoning capabilities and thus appear attractive for detecting visual bugs directly from video frames. Recent benchmarks suggest that VLMs can achieve promising results in detecting visual glitches on curated datasets. Building on these findings, we conduct a real-world study using industrial QA gameplay videos to evaluate how well VLMs perform in practical scenarios. Our study samples keyframes from long gameplay videos and asks a VLM whether each keyframe contains a bug. Starting from a single-prompt baseline, the model achieves a precision of 0.50 and an accuracy of 0.72. We then examine two common enhancement strategies used to improve VLM performance without fine-tuning: (1) a secondary judge model that re-evaluates VLM outputs, and (2) metadata-augmented prompting through the retrieval of prior bug reports. Across \textbf{100 videos} totaling \textbf{41 hours} and \textbf{19,738 keyframes}, these strategies provide only marginal improvements over the simple baseline, while introducing additional computational cost and output variance. Our findings indicate that off-the-shelf VLMs are already capable of detecting a certain range of visual bugs in QA gameplay videos, but further progress likely requires hybrid approaches that better separate textual and visual anomaly detection. --- ### Synthetic or Authentic? Building Mental Patient Simulators from Longitudinal Evidence Authors: Baihan Li, Bingrui Jin, Kunyao Lan, Ming Wang, Mengyue Wu Published: 2026-03-24 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.22704v1 Patient simulation is essential for developing and evaluating mental health dialogue systems. As most existing approaches rely on snapshot-style prompts with limited profile information, homogeneous behaviors and incoherent disease progression in multi-turn interactions have become key chellenges. In this work, we propose DEPROFILE, a data-grounded patient simulation framework that constructs unified, multi-source patient profiles by integrating demographic attributes, standardized clinical symptoms, counseling dialogues, and longitudinal life-event histories from real-world data. We further introduce a Chain-of-Change agent to transform noisy longitudinal records into structured, temporally grounded memory representations for simulation. Experiments across multiple large language model (LLM) backbones show that with more comprehensive profile constructed by DEPROFILE, the dialogue realism, behavioral diversity, and event richness have consistently improved and exceed state-of-the-art baselines, highlighting the importance of grounding patient simulation in verifiable longitudinal evidence. --- ### TimeWeaver: Age-Consistent Reference-Based Face Restoration with Identity Preservation Authors: Teer Song, Yue Zhang, Yu Tian, Ziyang Wang, Xianlin Zhang, Guixuan Zhang, Xuan Liu, Xueming Li, Yasen Zhang Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.22701v1 Recent progress in face restoration has shifted from visual fidelity to identity fidelity, driving a transition from reference-free to reference-based paradigms that condition restoration on reference images of the same person. However, these methods assume the reference and degraded input are age-aligned. When only cross-age references are available, as in historical restoration or missing-person retrieval, they fail to maintain age fidelity. To address this limitation, we propose TimeWeaver, the first reference-based face restoration framework supporting cross-age references. Given arbitrary reference images and a target-age prompt, TimeWeaver produces restorations with both identity fidelity and age consistency. Specifically, we decouple identity and age conditioning across training and inference. During training, the model learns an age-robust identity representation by fusing a global identity embedding with age-suppressed facial tokens via a transformer-based ID-Fusion module. During inference, two training-free techniques, Age-Aware Gradient Guidance and Token-Targeted Attention Boost, steer sampling toward desired age semantics, enabling precise adherence to the target-age prompt. Extensive experiments show that TimeWeaver surpasses existing methods in visual quality, identity preservation, and age consistency. --- ### Coordinate Encoding on Linear Grids for Physics-Informed Neural Networks Authors: Tetsuro Tsuchino, Motoki Shiga Published: 2026-03-24 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.22700v1 In solving partial differential equations (PDEs), machine learning utilizing physical laws has received considerable attention owing to advantages such as mesh-free solutions, unsupervised learning, and feasibility for solving high-dimensional problems. An effective approach is based on physics-informed neural networks (PINNs), which are based on deep neural networks known for their excellent performance in various academic and industrial applications. However, PINNs struggled with model training owing to significantly slow convergence because of a spectral bias problem. In this study, we propose a PINN-based method equipped with a coordinate-encoding layer on linear grid cells. The proposed method improves the training convergence speed by separating the local domains using grid cells. Moreover, it reduces the overall computational cost by using axis-independent linear grid cells. The method also achieves efficient and stable model training by adequately interpolating the encoded coordinates between grid points using natural cubic splines, which guarantees continuous derivative functions of the model computed for the loss functions. The results of numerical experiments demonstrate the effective performance and efficient training convergence speed of the proposed method. --- ### WiFi2Cap: Semantic Action Captioning from Wi-Fi CSI via Limb-Level Semantic Alignment Authors: Tzu-Ti Wei, Chu-Yu Huang, Yu-Chee Tseng, Jen-Jee Chen Published: 2026-03-24 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.22690v1 Privacy-preserving semantic understanding of human activities is important for indoor sensing, yet existing Wi-Fi CSI-based systems mainly focus on pose estimation or predefined action classification rather than fine-grained language generation. Mapping CSI to natural-language descriptions remains challenging because of the semantic gap between wireless signals and language and direction-sensitive ambiguities such as left/right limb confusion. We propose WiFi2Cap, a three-stage framework for generating action captions directly from Wi-Fi CSI. A vision-language teacher learns transferable supervision from synchronized video-text pairs, and a CSI student is aligned to the teacher's visual space and text embeddings. To improve direction-sensitive captioning, we introduce a Mirror-Consistency Loss that reduces mirrored-action and left-right ambiguities during cross-modal alignment. A prefix-tuned language model then generates action descriptions from CSI embeddings. We also introduce the WiFi2Cap Dataset, a synchronized CSI-RGB-sentence benchmark for semantic captioning from Wi-Fi signals. Experimental results show that WiFi2Cap consistently outperforms baseline methods on BLEU-4, METEOR, ROUGE-L, CIDEr, and SPICE, demonstrating effective privacy-friendly semantic sensing. --- ### Think 360°: Evaluating the Width-centric Reasoning Capability of MLLMs Beyond Depth Authors: Mingrui Chen, Hexiong Yang, Haogeng Liu, Huaibo Huang, Ran He Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.22689v1 In this paper, we present a holistic multimodal benchmark that evaluates the reasoning capabilities of MLLMs with an explicit focus on reasoning width, a complementary dimension to the more commonly studied reasoning depth. Specifically, reasoning depth measures the model's ability to carry out long-chain, sequential reasoning in which each step is tightly and rigorously linked to the next. Reasoning width tends to focus more on the model's capacity for broad trial-and-error search or multi-constrained optimization: it must systematically traverse many possible and parallelized reasoning paths, apply diverse constraints to prune unpromising branches, and identify valid solution routes for efficient iteration or backtracking. To achieve it, we carefully curate 1200+ high-quality multimodal cases spanning heterogeneous domains, and propose a fine-grained tree-of-thought evaluation protocol that jointly quantifies reasoning width and depth. We evaluate 12 major model families (over 30 advanced MLLMs) across difficulty tiers, question types, and required skills. Results show that while current models exhibit strong performance on general or common-sense VQA tasks, they still struggle to combine deep sequential thought chains with wide exploratory search to perform genuine insight-based reasoning. Finally, we analyze characteristic failure modes to provide possible directions for building MLLMs that reason not only deeper but also wider. --- ### GeoTikzBridge: Advancing Multimodal Code Generation for Geometric Perception and Reasoning Authors: Jiayin Sun, Caixia Sun, Boyu Yang, Hailin Li, Xiao Chen, Yi Zhang, Errui Ding, Liang Li, Chao Deng, Junlan Feng Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.22687v2 Multimodal Large Language Models (MLLMs) have recently demonstrated remarkable perceptual and reasoning abilities. However, they struggle to perceive fine-grained geometric structures, constraining their ability of geometric understanding and visual reasoning. To address this, we propose GeoTikzBridge, a framework that enhances local geometric perception and visual reasoning through tikz-based code generation. Within this framework, we build two models supported by two complementary datasets. The GeoTikzBridge-Base model is trained on GeoTikz-Base dataset, the largest image-to-tikz dataset to date with 2.5M pairs (16 $\times$ larger than existing open-sourced datasets). This process is achieved via iterative data expansion and a localized geometric transformation strategy. Subsequently, GeoTikzBridge-Instruct is fine-tuned on GeoTikz-Instruct dataset which is the first instruction-augmented tikz dataset supporting visual reasoning. Extensive experimental results demonstrate that our models achieve state-of-the-art performance among open-sourced MLLMs. Furthermore, GeoTikzBridge models can serve as plug-and-play reasoning modules for any MLLM(LLM), enhancing reasoning performance in geometric problem-solving. Datasets and codes are publicly available at: https://github.com/sjy-1995/GeoTikzBridge. --- ### MuQ-Eval: An Open-Source Per-Sample Quality Metric for AI Music Generation Evaluation Authors: Di Zhu, Zixuan Li Published: 2026-03-24 Categories: cs.AI, cs.SD arXiv: https://arxiv.org/abs/2603.22677v1 Distributional metrics such as Fréchet Audio Distance cannot score individual music clips and correlate poorly with human judgments, while the only per-sample learned metric achieving high human correlation is closed-source. We introduce MUQ-EVAL, an open-source per-sample quality metric for AIgenerated music built by training lightweight prediction heads on frozen MuQ-310M features using MusicEval, a dataset of generated clips from 31 text-to-music systems with expert quality ratings. Our simplest model, frozen features with attention pooling and a two-layer MLP, achieves system-level SRCC = 0.957 and utterance-level SRCC = 0.838 with human mean opinion scores. A systematic ablation over training objectives and adaptation strategies shows that no addition meaningfully improves the frozen baseline, indicating that frozen MuQ representations already capture quality-relevant information. Encoder choice is the dominant design factor, outweighing all architectural and training decisions. LoRA-adapted models trained on as few as 150 clips already achieve usable correlation, enabling personalized quality evaluators from individual listener annotations. A controlled degradation analysis reveals selective sensitivity to signal-level artifacts but insensitivity to musical-structural distortions. Our metric, MUQ-EVAL, is fully open-source, outperforms existing open per-sample metrics, and runs in real time on a single consumer GPU. Code, model weights, and evaluation scripts are available at https://github.com/dgtql/MuQ-Eval. --- ### Vision-based Deep Learning Analysis of Unordered Biomedical Tabular Datasets via Optimal Spatial Cartography Authors: Sakib Mostafa, Tarik Massoud, Maximilian Diehn, Lei Xing, Md Tauhidul Islam Published: 2026-03-24 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.22675v1 Tabular data are central to biomedical research, from liquid biopsy and bulk and single-cell transcriptomics to electronic health records and phenotypic profiling. Unlike images or sequences, however, tabular datasets lack intrinsic spatial organization: features are treated as unordered dimensions, and their relationships must be inferred implicitly by the model. This limits the ability of vision architectures to exploit local structure and higher-order feature interactions in non-spatial biomedical data. Here we introduce Dynamic Feature Mapping (Dynomap), an end-to-end deep learning framework that learns a task-optimized spatial topology of features directly from data. Dynomap jointly optimizes feature placement and prediction through a fully differentiable rendering mechanism, without relying on heuristics, predefined groupings, or external priors. By transforming high-dimensional tabular vectors into learned feature maps, Dynomap enables vision-based models to operate effectively on unordered biomedical inputs. Across multiple clinical and biological datasets, Dynomap consistently outperformed classical machine learning, modern deep tabular models, and existing vector-to-image approaches. In liquid biopsy data, Dynomap organized clinically relevant gene signatures into coherent spatial patterns and improved multiclass cancer subtype prediction accuracy by up to 18%. In a Parkinson disease voice dataset, it clustered disease-associated acoustic descriptors and improved accuracy by up to 8%. Similar gains and interpretable feature organization were observed in additional biomedical datasets. These results establish Dynomap as a general strategy for bridging tabular and vision-based deep learning and for uncovering structured, clinically relevant patterns in high-dimensional biomedical data. --- ### Improving LLM Final Representations with Inter-Layer Geometry Authors: Tom Ulanovski, Eyal Blyachman, Maya Bechler-Speicher Published: 2026-03-24 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2603.22665v3 The standard in LLM-based prediction is to use the final-layer representation as the input to a downstream predictor. However, intermediate layers may encode complementary task-relevant signals. Existing approaches therefore either search for the best layer for each task or apply expensive attention-based mechanisms to learn inter-layer aggregation. In this work, we first show that such complexity is unnecessary: a lightweight Graph Neural Network over a fully connected graph of LLM layers is more efficient and achieves significantly stronger predictive performance than existing approaches. We then introduce the Cayley-Encoder, which further improves both efficiency and predictive performance by replacing the fully connected graph with a Cayley graph over SL(2, Zn). These Cayley graphs provide a mathematically grounded topology that is sparse, regular by construction, and has low diameter. This enables effective communication across layers while constraining the aggregation structure and reducing the risk of GNN overfitting. In an evaluation of Cayley-Encoder across 13 tasks and 9 LLMs, Cayley-Encoder consistently outperforms baselines, achieving improvements of up to 40 percentage points in accuracy, while introducing at most 0.1% additional parameters relative to the LLM size. We further show that Cayley-Encoder is effective in few-shot regimes. Finally, we show that Cayley-Encoder outperforms LoRA fine-tuning while operating on the frozen LLM. We conclude with an explainability analysis showing that multiple layers contribute meaningfully to the final prediction, supporting our hypothesis. --- ### Bounding Box Anomaly Scoring for simple and efficient Out-of-Distribution detection Authors: Mohamed Bahi Yahiaoui, Geoffrey Daniel, Loïc Giraldi, Jérémie Bruyelle, Julyan Arbel Published: 2026-03-24 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.22660v1 Out-of-distribution (OOD) detection aims to identify inputs that differ from the training distribution in order to reduce unreliable predictions by deep neural networks. Among post-hoc feature-space approaches, OOD detection is commonly performed by approximating the in-distribution support in the representation space of a pretrained network. Existing methods often reflect a trade-off between compact parametric models, such as Mahalanobis-based scores, and more flexible but reference-based methods, such as k-nearest neighbors. Bounding-box abstraction provides an attractive intermediate perspective by representing in-distribution support through compact axis-aligned summaries of hidden activations. In this paper, we introduce Bounding Box Anomaly Scoring (BBAS), a post-hoc OOD detection method that leverages bounding-box abstraction. BBAS combines graded anomaly scores based on interval exceedances, monitoring variables adapted to convolutional layers, and decoupled clustering and box construction for richer and multi-layer representations. Experiments on image-classification benchmarks show that BBAS provides robust separation between in-distribution and out-of-distribution samples while preserving the simplicity, compactness, and updateability of the bounding-box approach. --- ### Large-Scale Avalanche Mapping from SAR Images with Deep Learning-based Change Detection Authors: Mattia Gatti, Alberto Mariani, Ignazio Gallo, Fabiano Monti Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.22658v1 Accurate change detection from satellite imagery is essential for monitoring rapid mass-movement hazards such as snow avalanches, which increasingly threaten human life, infrastructure, and ecosystems due to their rising frequency and intensity. This study presents a systematic investigation of large-scale avalanche mapping through bi-temporal change detection using Sentinel-1 synthetic aperture radar (SAR) imagery. Extensive experiments across multiple alpine ecoregions with manually validated avalanche inventories show that treating the task as a unimodal change detection problem, relying solely on pre- and post-event SAR images, achieves the most consistent performance. The proposed end-to-end pipeline achieves an F1-score of 0.8061 in a conservative (F1-optimized) configuration and attains an F2-score of 0.8414 with 80.36% avalanche-polygon hit rate under a less conservative, recall-oriented (F2-optimized) tuning. These results highlight the trade-off between precision and completeness and demonstrate how threshold adjustment can improve the detection of smaller or marginal avalanches. The release of the annotated multi-region dataset establishes a reproducible benchmark for SAR-based avalanche mapping. --- ### Generalizing Dynamics Modeling More Easily from Representation Perspective Authors: Yiming Wang, Zhengnan Zhang, Genghe Zhang, Jiawen Dan, Changchun Li, Chenlong Hu, Chris Nugent, Jun Liu, Ximing Li, Bo Yang Published: 2026-03-24 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.22655v1 Learning system dynamics from observations is a critical problem in many applications over various real-world complex systems, e.g., climate, ecology, and fluid systems. Recently, neural dynamics modeling method have become a prevalent solution that embeds the object's observations into a latent space before learning dynamics using neural methods such as neural Ordinary Differential Equations (ODE). Existing dynamics modeling methods induce a specific model for each observation of different complex systems, resulting in poor generalization across systems. Inspired by the great success of pre-trained models, we conduct a generalized Pre-trained Dynamics EncoDER (PDEDER) which can embed the original state observations into a latent space where the dynamics can be captured more easily. To conduct the generalized PDEDER, we pre-train any Pre-trained Language Model (PLM) by minimizing the Lyapunov exponent objective, which constrains the chaotic behavior of governing dynamics learned in the latent space. By penalizing the divergence of embedded observations, our PDEDER promotes locally stable and well-structured latent dynamics, thereby facilitating more effective dynamics modeling than in the original observation space. In addition, we incorporate reconstruction and forecasting objectives to mitigate the risk of obtaining an over-smoothed latent space. Specifically, we collect 152 sets of real-world and synthetic observations from 23 complex systems as pre-training corpora and employ them to pre-train PDEDER. Given any future dynamic observation, we can fine-tune PDEDER with any specific dynamics modeling method. We evaluate PDEDER on 12 dynamic systems by short/long-term forecasting under both in-domain and cross-domain settings, and the empirical results indicate the effectiveness and generalizability of PDEDER. --- ### Benchmarking Multi-Agent LLM Architectures for Financial Document Processing: A Comparative Study of Orchestration Patterns, Cost-Accuracy Tradeoffs and Production Scaling Strategies Authors: Siddhant Kulkarni, Yukta Kulkarni Published: 2026-03-24 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2603.22651v1 The adoption of large language models (LLMs) for structured information extraction from financial documents has accelerated rapidly, yet production deployments face fundamental architectural decisions with limited empirical guidance. We present a systematic benchmark comparing four multi-agent orchestration architectures: sequential pipeline, parallel fan-out with merge, hierarchical supervisor-worker and reflexive self-correcting loop. These are evaluated across five frontier and open-weight LLMs on a corpus of 10,000 SEC filings (10-K, 10-Q and 8-K forms). Our evaluation spans 25 extraction field types covering governance structures, executive compensation and financial metrics, measured along five axes: field-level F1, document-level accuracy, end-to-end latency, cost per document and token efficiency. We find that reflexive architectures achieve the highest field-level F1 (0.943) but at 2.3x the cost of sequential baselines, while hierarchical architectures occupy the most favorable position on the cost-accuracy Pareto frontier (F1 0.921 at 1.4x cost). We further present ablation studies on semantic caching, model routing and adaptive retry strategies, demonstrating that hybrid configurations can recover 89\% of the reflexive architecture's accuracy gains at only 1.15x baseline cost. Our scaling analysis from 1K to 100K documents per day reveals non-obvious throughput-accuracy degradation curves that inform capacity planning. These findings provide actionable guidance for practitioners deploying multi-agent LLM systems in regulated financial environments. --- ### AI-Driven Alpha Decay: Algorithmic Homogenization, Reflexive Signal Erosion, and the Paradox of Intelligent Markets Authors: Shuchen Meng, Xupeng Chen Published: 2026-03-23 Categories: q-fin.GN, cs.AI, cs.GT arXiv: https://arxiv.org/abs/2605.23905v1 We show that AI-driven investment strategies are inherently self-defeating at scale. As AI adoption rises, three mutually reinforcing channels -- signal crowding, performative signal erosion, and Red Queen competition -- compress excess returns. We derive the alpha half-life $h(φ) = \ln 2/(θ+ δ(φ))$, where $θ$ is the natural mean-reversion rate and $δ(φ) = Nφρa/λ(φ)$ is the AI-accelerated decay component, which is convex-decreasing in adoption. At current adoption levels ($φ\approx 0.7$, $ρ\approx 0.6$), the model implies signal half-lives of 18 months versus 5-7 years pre-AI. We establish four theoretical results. First, the alpha half-life theorem: signal lifespans are convex-decreasing in AI adoption. Second, a signal extinction cascade: beyond a critical threshold $φ^*$, the decay of one signal class triggers accelerated competition for remaining signals. Third, a Red Queen impossibility: in the monoculture equilibrium, net alpha is identically zero despite heavy AI investment. Fourth, a fragility-efficiency tradeoff: the adoption level maximizing price discovery strictly exceeds the level minimizing systemic fragility. Empirical validation calibrates portfolio convergence to SEC Form 13F filing patterns (99.5 million holdings, 2013-2024), documenting that simulated institutional portfolio convergence increases by 42% over the sample period. We examine simulated hedge fund return dynamics showing declining cross-sectional dispersion among AI-adopting funds, and simulate the 2010 Flash Crash to illustrate fragility consequences. --- ### SOLAR: A Self-Optimizing Open-Ended Autonomous Agent for Lifelong Learning and Continual Adaptation Authors: Nitin Vetcha, Dianbo Liu Published: 2026-03-23 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.20189v1 Despite the remarkable success of large language models (LLMs), they still face bottlenecks while deploying in dynamic, real-world settings with primary challenges being concept drift and the high cost of gradient-based adaptation. Traditional fine-tuning (FT) struggles to adapt to non-stationary data streams without resulting in catastrophic for getting or requiring extensive manual data curation. To address these limitations within the streaming and continual learning paradigm, we propose the Self-Optimizing Lifelong Autonomous Reasoner (SOLAR) which is an open-ended autonomous agent that leverages parameter-level meta-learning to self-improve, treating model weights as an environment for exploration. It initiates the process by consolidating a strong prior over common-sense knowledge making it effective for transfer-learning. By utilizing a multi-level reinforcement learning approach, SOLAR autonomously discovers adaptation strategies, enabling efficient test-time adaptation to unseen domains. Crucially, SOLAR maintains an evolving knowledge base of valid modification strategies, implicitly acting as an episodic memory buffer to balance plasticity (adaptation to new tasks) and stability (retention of meta-knowledge). Experiments demonstrate that SOLAR outperforms strong baselines on common-sense, mathematical, medical, coding, social and logical reasoning tasks, marking a significant step toward autonomous agents capable of lifelong adaptation in evolving environments. --- ### Implicit Humanization in Everyday LLM Moral Judgments Authors: Hoda Ayad, Tanu Mitra Published: 2026-03-23 Categories: cs.CY, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2604.22764v1 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. --- ### Transparent Screening for LLM Inference and Training Impacts Authors: Arnault Pachot, Thierry Petit Published: 2026-03-23 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2604.19757v1 This paper presents a transparent screening framework for estimating inference and training impacts of current large language models under limited observability. The framework converts natural-language application descriptions into bounded environmental estimates and supports a comparative online observatory of current market models. Rather than claiming direct measurement for opaque proprietary services, it provides an auditable, source-linked proxy methodology designed to improve comparability, transparency, and reproducibility. --- ### CFMS: Towards Explainable and Fine-Grained Chinese Multimodal Sarcasm Detection Benchmark Authors: Junzhao Zhang, Hsiu-Yuan Huang, Chenming Tang, Yutong Yang, Yunfang Wu Published: 2026-03-23 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.16372v1 Multimodal sarcasm detection has recently garnered significant attention. However, existing benchmarks suffer from coarse-grained annotations and limited cultural coverage, which hinder research into fine-grained semantic understanding. To address this, we construct CFMS, the first fine-grained multimodal sarcasm dataset tailored for Chinese social media. It comprises 2,796 high-quality image-text pairs and provides a triple-level annotation framework: sarcasm identification, target recognition, and explanation generation. We find that the fine-grained explanation annotations effectively guide AI in generating images with explicit sarcastic intent. Furthermore, we curate a high-consistency parallel Chinese-English metaphor subset (200 entries each), revealing significant limitations of current models in metaphoric reasoning. To overcome the constraints of traditional retrieval methods, we propose a Reinforcement Learning-augmented In-Context Learning strategy (PGDS) to dynamically optimize exemplar selection. Extensive experiments demonstrate that CFMS provides a solid foundation for building reliable multimodal sarcasm understanding systems, and the PGDS method significantly outperforms existing baselines on key tasks. Our data and code are available at https://anonymous.4open.science/r/CFMS-E8F9. --- ### Brain-CLIPLM: Semantic Compression for EEG-to-Text Decoding Authors: Xiaoli Yang, Huiyuan Tian, Yurui Li, Jianyu Zhang, Shijian Li, Gang Pan Published: 2026-03-23 Categories: cs.CL, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2604.16370v3 Decoding natural language from non-invasive electroencephalography (EEG) remains constrained by low signal-to-noise ratio and limited information bandwidth. This raises a central question: can sentence-level language be reliably recovered from such signals? Under realistic information constraints, this direct-recovery assumption may be too strong. We introduce a semantic compression hypothesis: non-invasive EEG may preserve recoverable semantic anchors rather than the full lexical--syntactic form of a sentence. From this perspective, direct sentence reconstruction is overly fine-grained relative to the recoverable information scale of EEG. To address this mismatch, we propose Brain-CLIPLM, a two-stage framework that decomposes EEG-to-text decoding into semantic-anchor recovery and anchor-guided sentence reconstruction. Stage 1 uses contrastive learning to align word-level EEG evidence with a fixed keyword vocabulary and recover ordered semantic anchors. Stage 2 uses a retrieval-grounded large language model with chain-of-thought reasoning prompts to reconstruct sentence meaning from these anchors, following a granularity matching principle that aligns decoding complexity with the recoverable neural information scale. On the combined Zurich Cognitive Language Processing (ZuCo) benchmark, Brain-CLIPLM achieves 67.6\% Top-5 and 85.0\% Top-25 sentence retrieval accuracy, with the strongest performance at intermediate anchor granularity. Control analyses show that EEG-derived anchors carry sentence-specific information beyond language-model priors. Within the constrained ZuCo sentence pool and fixed keyword-vocabulary settings, these findings suggest that EEG-to-text decoding is better framed as recovering compressed semantic content before anchor-guided sentence reconstruction. --- ### EviSearch: Trustworthy Extraction and Synthesis of Clinical Trial Evidence with Agents that Improve with Use Authors: Naman Ahuja, Abhijit Chakraborty, Muhammad Ali Khan, Kaneez Zahra Rubab Khakwani, Mohamad Bassam Sonbol, Irbaz Bin Riaz, Vivek Gupta Published: 2026-03-23 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.14165v3 Structured extraction of evidence from clinical trial publications underpins systematic reviews and clinical guidelines, yet large language models are adopted for it only hesitantly: their outputs are difficult to verify, their use commonly requires transmitting documents to proprietary services, and they do not improve from the corrections their users make. We present EviSearch, a multi-agent system that addresses these three obstacles. Three tool-augmented agents with complementary access to a publication extract every column of an evidence table, and a value is admitted only after an attribution verifier has read it on its cited page, so that every value carries a page-level attribution. Disagreement between independent agents directs human review to the cells most likely to be wrong, and reviewer feedback refines the schema definitions and a curation knowledge base without updating model parameters. The agentic system runs entirely offline on open-weight models. On a clinician-annotated benchmark of randomized-trial publications, EviSearch attributes 100.0% of its values, reaches 91.70% accuracy autonomously, and reaches 95.22% after review of 15.6% of cells, exceeding random review of the strongest single agent at equal effort by 1.75 points. --- ### EviSearch: A Human in the Loop System for Extracting and Auditing Clinical Evidence for Systematic Reviews Authors: Naman Ahuja, Saniya Mulla, Muhammad Ali Khan, Zaryab Bin Riaz, Kaneez Zahra Rubab Khakwani, Mohamad Bassam Sonbol, Irbaz Bin Riaz, Vivek Gupta Published: 2026-03-23 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.14165v2 We present EviSearch, a multi-agent extraction system that automates the creation of ontology-aligned clinical evidence tables directly from native trial PDFs while guaranteeing per-cell provenance for audit and human verification. EviSearch pairs a PDF-query agent (which preserves rendered layout and figures) with a retrieval-guided search agent and a reconciliation module that forces page-level verification when agents disagree. The pipeline is designed for high-precision extraction across multimodal evidence sources (text, tables, figures) and for generating reviewer-actionable provenance that clinicians can inspect and correct. On a clinician-curated benchmark of oncology trial papers, EviSearch substantially improves extraction accuracy relative to strong parsed-text baselines while providing comprehensive attribution coverage. By logging reconciler decisions and reviewer edits, the system produces structured preference and supervision signals that bootstrap iterative model improvement. EviSearch is intended to accelerate living systematic review workflows, reduce manual curation burden, and provide a safe, auditable path for integrating LLM-based extraction into evidence synthesis pipelines. --- ### How to Fine-Tune a Reasoning Model? A Teacher-Student Cooperation Framework to Synthesize Student-Consistent SFT Data Authors: Zixian Huang, Kaichen Yang, Xu Huang, Feiyang Hao, Qiming Ge, Bowen Li, He Du, Kai Chen, Qipeng Guo Published: 2026-03-23 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.14164v2 A widely adopted strategy for model enhancement is to use synthetic data generated by a stronger model for supervised fine-tuning (SFT). However, for emerging reasoning models like Qwen3-8B, this approach often fails to improve reasoning capabilities and can even lead to a substantial drop in performance. In this work, we identify substantial stylistic divergence between teacher generated data and the distribution of student as a major factor impacting SFT. To bridge this gap, we propose a Teacher-Student Cooperation Data Synthesis framework (TESSY), which interleaves teacher and student models to alternately generate style and non-style tokens. Consequently, TESSY produces synthetic sequences that inherit the advanced reasoning capabilities of the teacher while maintaining stylistic consistency with the distribution of the student. In experiments on code generation using GPT-OSS-120B as the teacher, fine-tuning Qwen3-8B on teacher-generated data leads to performance drops of 3.25% on LiveCodeBench-Pro and 10.02% on OJBench, whereas TESSY achieves improvements of 11.25% and 6.68%. --- ### SeaAlert: Robust Severity Classification and LLM-Based Information Extraction for Noisy Maritime Distress Communications Authors: Tomer Atia, Yehudit Aperstein, Alexander Apartsin Published: 2026-03-23 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.14163v2 Maritime distress communications transmitted over very high frequency (VHF) radio are safety-critical voice messages used to report emergencies at sea. Under the Global Maritime Distress and Safety System (GMDSS), such messages follow standardized procedures and are expected to convey essential details, including vessel identity, position, nature of the distress, and required assistance. In practice, however, automatic analysis remains difficult because distress messages are often brief, noisy, and produced under stress, may deviate from the prescribed format, and are further degraded by automatic speech recognition (ASR) errors caused by channel noise and speaker stress. This paper presents SeaAlert, a controlled experimental framework for evaluating robust analysis of maritime distress communications using transformer-based severity classification and LLM-based structured extraction. To address the scarcity of labeled real-world data, we develop a synthetic data generation pipeline in which an LLM produces diverse maritime messages, including challenging variants in which standard distress codewords are omitted or replaced with less explicit expressions. The generated utterances are synthesized into speech, degraded with simulated VHF noise, and transcribed by an ASR system to obtain controlled noise-degraded transcripts. The resulting evaluation shows that transformer-based classification degrades more gracefully than lexical baselines under ASR noise and codeword masking, while LLM-based extraction is more effective than Regex-based extraction for noisy structured fields. --- ### Decoupling Scores and Text: The Politeness Principle in Peer Review Authors: Yingxuan Wen Published: 2026-03-23 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2604.14162v1 Authors often struggle to interpret peer review feedback, deriving false hope from polite comments or feeling confused by specific low scores. To investigate this, we construct a dataset of over 30,000 ICLR 2021-2025 submissions and compare acceptance prediction performance using numerical scores versus text reviews. Our experiments reveal a significant performance gap: score-based models achieve 91% accuracy, while text-based models reach only 81% even with large language models, indicating that textual information is considerably less reliable. To explain this phenomenon, we first analyze the 9% of samples that score-based models fail to predict, finding their score distributions exhibit high kurtosis and negative skewness, which suggests that individual low scores play a decisive role in rejection even when the average score falls near the borderline. We then examine why text-based accuracy significantly lags behind scores from a review sentiment perspective, revealing the prevalence of the Politeness Principle: reviews of rejected papers still contain more positive than negative sentiment words, masking the true rejection signal and making it difficult for authors to judge outcomes from text alone. --- ### Can Large Language Models Detect Methodological Flaws? Evidence from Gesture Recognition for UAV-Based Rescue Operation Based on Deep Learning Authors: Domonkos Varga Published: 2026-03-23 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.14161v1 Reliable evaluation is essential in machine learning research, yet methodological flaws-particularly data leakage-continue to undermine the validity of reported results. In this work, we investigate whether large language models (LLMs) can act as independent analytical agents capable of identifying such issues in published studies. As a case study, we analyze a gesture-recognition paper reporting near-perfect accuracy on a small, human-centered dataset. We first show that the evaluation protocol is consistent with subject-level data leakage due to non-independent training and test splits. We then assess whether this flaw can be detected independently by six state-of-the-art LLMs, each analyzing the original paper without prior context using an identical prompt. All models consistently identify the evaluation as flawed and attribute the reported performance to non-independent data partitioning, supported by indicators such as overlapping learning curves, minimal generalization gap, and near-perfect classification results. These findings suggest that LLMs can detect common methodological issues based solely on published artifacts. While not definitive, their consistent agreement highlights their potential as complementary tools for improving reproducibility and supporting scientific auditing. --- ### NuHF Claw: A Risk Constrained Cognitive Agent Framework for Human Centered Procedure Support in Digital Nuclear Control Rooms Authors: Xingyu Xiao, Jiejuan Tong, Jun Sun, Zhe Sui, Peng Chen, Jingang Liang, Haitao Wang Published: 2026-03-23 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.14160v1 The rapid digitization of nuclear power plant main control rooms has fundamentally reshaped operator interaction patterns, introducing complex soft-control behaviors and elevated cognitive risks that are not adequately addressed by existing human reliability analysis approaches. Although recent advances in large language models and autonomous agents offer new opportunities for intelligent decision support, their deployment in safety critical environments remains constrained by risks of hallucinated reasoning and weakened human authority. This study proposes NuHF Claw, a persistent cognitive-risk agent framework that enables risk governed human centered autonomy for digital nuclear operations. The core methodological innovation lies in the introduction of a risk constrained agent runtime, which tightly couples cognitive state inference with probabilistic safety assessment to regulate autonomous system behavior in real time. By integrating cognitively grounded workload and situational awareness estimation with dynamic human error probability prediction, the framework transforms conventional offline reliability analysis into a proactive intervention mechanism embedded directly within operational workflows. Experimental validation on a high-fidelity digital control room simulator demonstrates that NuHF Claw can anticipate interface induced cognitive degradation, dynamically constrain unsafe autonomous recommendations, and provide risk-aware navigational guidance while preserving human decision authority. The results highlight a fundamental shift from automation-driven operation toward cognition-aware autonomy, offering a principled pathway for the safe integration of intelligent agents into next-generation nuclear control environments. --- ### HUOZIIME: An On-Device LLM-enhanced Input Method for Deep Personalization Authors: Baocai Shan, Yuzhuang Xu, Wanxiang Che Published: 2026-03-23 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.14159v1 Mobile input method editors (IMEs) are the primary interface for text input, yet they remain constrained to manual typing and struggle to produce personalized text. While lightweight large language models (LLMs) make on-device auxiliary generation feasible, enabling deeply personalized, privacy-preserving, and real-time generative IMEs poses fundamental challenges.To this end, we present HUOZIIME, a personalized on-device IME powered by LLM. We endow HUOZIIME with initial human-like prediction ability by post-training a base LLM on synthesized personalization data. Notably, a hierarchical memory mechanism is designed to continually capture and leverage user-specific input history. Furthermore, we perform systemic optimizations tailored to on-device LLMbased IME deployment, ensuring efficient and responsive operation under mobile constraints.Experiments demonstrate efficient on-device execution and high-fidelity memory-driven personalization. Code and package are available at https://github.com/Shan-HIT/HuoziIME. --- ### MemGround: Long-Term Memory Evaluation Kit for Large Language Models in Gamified Scenarios Authors: Yihang Ding, Wanke Xia, Yiting Zhao, Jinbo Su, Jialiang Yang, Zhengbo Zhang, Ke Wang, Wenming Yang Published: 2026-03-23 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.14158v1 Current evaluations of long-term memory in LLMs are fundamentally static. By fixating on simple retrieval and short-context inference, they neglect the multifaceted nature of complex memory systems, such as dynamic state tracking and hierarchical reasoning in continuous interactions. To overcome these limitations, we propose MemGround, a rigorous long-term memory benchmark natively grounded in rich, gamified interactive scenarios. To systematically assess these capabilities, MemGround introduces a three-tier hierarchical framework that evaluates Surface State Memory, Temporal Associative Memory, and Reasoning-Based Memory through specialized interactive tasks. Furthermore, to comprehensively quantify both memory utilization and behavioral trajectories, we propose a multi-dimensional metric suite comprising Question-Answer Score (QA Overall), Memory Fragments Unlocked (MFU), Memory Fragments with Correct Order (MFCO), and Exploration Trajectory Diagrams (ETD). Extensive experiments reveal that state-of-the-art LLMs and memory agents still struggle with sustained dynamic tracking, temporal event association, and complex reasoning derived from long-term accumulated evidence in interactive environments. --- ### Alignment as Institutional Design: From Behavioral Correction to Transaction Structure in Intelligent Systems Authors: Rui Chai Published: 2026-03-23 Categories: cs.CY, cs.AI, cs.GT, cs.LG arXiv: https://arxiv.org/abs/2604.13079v1 Current AI alignment paradigms rely on behavioral correction: external supervisors (e.g., RLHF) observe outputs, judge against preferences, and adjust parameters. This paper argues that behavioral correction is structurally analogous to an economy without property rights, where order requires perpetual policing and does not scale. Drawing on institutional economics (Coase, Alchian, Cheung), capability mutual exclusivity, and competitive cost discovery, we propose alignment as institutional design: the designer specifies internal transaction structures (module boundaries, competition topologies, cost-feedback loops) such that aligned behavior emerges as the lowest-cost strategy for each component. We identify three irreducible levels of human intervention (structural, parametric, monitorial) and show that this framework transforms alignment from a behavioral control problem into a political-economy problem. No institution eliminates self-interest or guarantees optimality; the best design makes misalignment costly, detectable, and correctable. We conclude that the proper goal is institutional robustness-a dynamic, self-correcting process under human oversight, not perfection. This work provides the normative foundation for the Wuxing resource-competition mechanisms in companion papers. Keywords: AI alignment, institutional design, transaction costs, property rights, resource competition, behavioral correction, RLHF, cost truthfulness, modular architecture, correctable alignment --- ### PA-SFM: Tracker-free differentiable acoustic radiation for freehand 3D photoacoustic imaging Authors: Shuang Li, Jian Gao, Chulhong Kim, Seongwook Choi, Qian Chen, Yibing Wang, Shuang Wu, Yu Zhang, Tingting Huang, Yucheng Zhou, Boxin Yao, Yao Yao et al. Published: 2026-03-23 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.09643v1 Three-dimensional (3D) handheld photoacoustic tomography typically relies on bulky and expensive external positioning sensors to correct motion artifacts, which severely limits its clinical flexibility and accessibility. To address this challenge, we present PA-SFM, a tracker-free framework that leverages exclusively single-modality photoacoustic data for both sensor pose recovery and high-fidelity 3D reconstruction via differentiable acoustic radiation modeling. Unlike traditional structure-from-motion (SFM) methods based on visual features, PA-SFM integrates the acoustic wave equation into a differentiable programming pipeline. By leveraging a high-performance, GPU-accelerated acoustic radiation kernel, the framework simultaneously optimizes the 3D photoacoustic source distribution and the sensor array pose via gradient descent. To ensure robust convergence in freehand scenarios, we introduce a coarse-to-fine optimization strategy that incorporates geometric consistency checks and rigid-body constraints to eliminate motion outliers. We validated the proposed method through both numerical simulations and in-vivo rat experiments. The results demonstrate that PA-SFM achieves sub-millimeter positioning accuracy and restores high-resolution 3D vascular structures comparable to ground-truth benchmarks, offering a low-cost, software-defined solution for clinical freehand photoacoustic imaging. The source code is publicly available at \href{https://github.com/JaegerCQ/PA-SFM}{https://github.com/JaegerCQ/PA-SFM}. --- ### From experimentation to engagement: on the paradox of participatory AI and power in contexts of forced displacement and humanitarian crises Authors: Stella Suge, Sarah W. Spencer, Nyalleng Moorosi, Helen McElhinney, Geoff Loane, Sue Black Published: 2026-03-23 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2604.06219v1 Across the Global North, calls for participatory artificial intelligence (AI) to improve the responsible, safe, and ethical use of AI have increased, particularly efforts that engage citizens and communities whose well-being and safety may be directly impacted by AI and other algorithmic tools. These initiatives include surveys, community consultations, citizens' councils and assemblies, and co-designing AI models and projects. Far fewer efforts, however, have been made in the Global South, particularly in contexts related to humanitarian crises and forced displacement, where the deployment of AI and algorithmic tools is accelerating. In this paper, we critically examine participatory AI methods and their limitations in these contexts and explore the opinions and perceptions of AI held by displaced and crisis-affected communities. Based on a pilot exercise with communities living in Kakuma Refugee Camp in northwestern Kenya, we find important limitations in some participatory AI approaches which, if used in humanitarian contexts, could increase risks of so-called 'participation washing' and algorithmic harm. We argue that these risks are not predominantly driven by varying levels of understanding and awareness of AI but more closely linked to the fundamental power dynamics embedded within the humanitarian sector: between humanitarian aid recipients, service providers, donor governments, and host nations, as well as the power differentials and incentives that exist between AI companies and humanitarian actors. These structural conditions make the case not only for more rigorous participatory methods, but for independent governance architecture capable of holding humanitarian AI to account. ---