# Research Papers — 2026-04 · bundle page 38 of 233 > Full-text markdown of papers #1851–1900 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/38 [← Previous bundle](https://aipod.dev/papers/archive/2026-04/page/37.md) [Next bundle →](https://aipod.dev/papers/archive/2026-04/page/39.md) ### Identity-Decoupled Anonymization for Visual Evidence in Multi-modal Retrieval-Augmented Generation Authors: Zehua Cheng, Wei Dai, Jiahao Sun Published: 2026-04-26 Categories: cs.CV, cs.IR arXiv: https://arxiv.org/abs/2604.23584v1 Multi-modal retrieval-augmented generation (MRAG) systems retrieve visual evidence from large image corpora to ground the responses of large multi-modal models, yet the retrieved images frequently contain human faces whose identities constitute sensitive personal information. Existing anonymization techniques that destroy the non-identity visual cues that downstream reasoning depends on or fail to provide principled privacy guarantees. We propose Identity-Decoupled MRAG, a framework that interposes a generative anonymization module between retrieval and generation. Our approach consists of three components: (i)a disentangled variational encoder that factorizes each face into an identity code and a spatially-structured attribute code, regularized by a mutual-information penalty and a gradient-based independence term; (ii)a manifold-aware rejection sampler that replaces the identity code with a synthetic one guaranteed to be both distinct from the original and realistic; and (iii)a conditional latent diffusion generator that synthesizes the anonymized face from the replacement identity and the preserved attributes, distilled into a latent consistency model for low-latency deployment. Privacy is enforced through a multi-oracle ensemble of face recognition models with a hinge-based loss that halts optimization once identity similarity drops below the impostor-regime threshold. --- ### AgentEval: DAG-Structured Step-Level Evaluation for Agentic Workflows with Error Propagation Tracking Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu Published: 2026-04-26 Categories: cs.SE, cs.CL arXiv: https://arxiv.org/abs/2604.23581v1 Agentic systems that chain reasoning, tool use, and synthesis into multi-step workflows are entering production, yet prevailing evaluation practices like end-to-end outcome checks and ad-hoc trace inspection systematically mask the intermediate failures that dominate real-world error budgets. We present AgentEval, a framework that formalizes agent executions as evaluation directed acyclic graphs (DAGs), where each node carries typed quality metrics assessed by a calibrated LLM judge (GPT-4o), classified through a hierarchical failure taxonomy (3 levels, 21 subcategories), and linked to upstream dependencies for automated root cause attribution. An ablation study isolates the impact of DAG-based dependency modeling: it alone contributes +22 percentage points to failure detection recall and +34 pp to root cause accuracy over flat step-level evaluation with identical judges and rubrics. Across three production workflows (450 test cases, two agent model families, predominantly sequential architectures with a 12% non-DAG trace rate), AgentEval achieves 2.17x higher failure detection recall than end-to-end evaluation (0.89 vs. 0.41), Cohen's kappa = 0.84 agreement with human experts, and 72% root cause accuracy against an 81% human ceiling. Cross-system evaluation on tau-bench and SWE-bench traces confirms transferability (failure detection recall >= 0.78) without taxonomy or rubric modification. A 4-month pilot with 18 engineers detected 23 pre-release regressions through CI/CD-integrated regression testing, reducing median root-cause identification time from 4.2 hours to 22 minutes and driving measurable failure rate reductions in two workflows. --- ### PhysCodeBench: Benchmarking Physics-Aware Symbolic Simulation of 3D Scenes via Self-Corrective Multi-Agent Refinement Authors: Tianyidan Xie, Peiyu Wang, Hu Jiaxin, Yuyi Qian, Yuxuan Wang, Shenyi Wang, Rui Ma, Yanlun Peng, Lanjun Wang, Ying Tai, Jian Yang, Zili Yi Published: 2026-04-26 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2604.23580v2 Translating natural-language descriptions of physical phenomena into executable simulation code requires both programming expertise and physical reasoning. Current large language models (LLMs) lack this combination: they frequently produce code that runs but simulates the wrong physics. We introduce PhysCodeBench, the first benchmark for this task, with 1,200 expert-validated examples spanning four physical domains. Its evaluation suite, PhysCodeEval, goes beyond executability and visual fidelity to measure physical correctness directly from the engine state via conservation-law residuals and expert-written assertions, and supports cross-engine evaluation to disentangle physics reasoning from API fluency. As a reference method, we propose the Self-Corrective Multi-Agent Refinement Framework (SMRF), which decouples physics-aware error correction from code generation through specialized agents. This design is motivated by our finding that targeted correction, rather than generic iterative refinement, is the key driver of physical accuracy. SMRF nearly triples the physical-assertion pass rate of the best proprietary baseline (70.6\% vs.\ 23.8\%) and retains its advantage under cross-engine transfer. --- ### PhysCodeBench: Benchmarking Physics-Aware Symbolic Simulation of 3D Scenes via Self-Corrective Multi-Agent Refinement Authors: Tianyidan Xie, Peiyu Wang, Yuyi Qian, Yuxuan Wang, Rui Ma, Ying Tai, Song Wu, Qian Wang, Lanjun Wang, Zili Yi Published: 2026-04-26 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2604.23580v1 Physics-aware symbolic simulation of 3D scenes is critical for robotics, embodied AI, and scientific computing, requiring models to understand natural language descriptions of physical phenomena and translate them into executable simulation environments. While large language models (LLMs) excel at general code generation, they struggle with the semantic gap between physical descriptions and simulation implementation. We introduce PhysCodeBench, the first comprehensive benchmark for evaluating physics-aware symbolic simulation, comprising 700 manually-crafted diverse samples across mechanics, fluid dynamics, and soft-body physics with expert annotations. Our evaluation framework measures both code executability and physical accuracy through automated and visual assessment. Building on this, we propose a Self-Corrective Multi-Agent Refinement Framework (SMRF) with three specialized agents (simulation generator, error corrector, and simulation refiner) that collaborate iteratively with domain-specific validation to produce physically accurate simulations. SMRF achieves 67.7 points overall performance compared to 36.3 points for the best baseline among evaluated SOTA models, representing a 31.4-point improvement. Our analysis demonstrates that error correction is critical for accurate physics-aware symbolic simulation and that specialized multi-agent approaches significantly outperform single-agent methods across the tested physical domains. --- ### LLMs Reading the Rhythms of Daily Life: Aligned Understanding for Behavior Prediction and Generation Authors: Fanjin Meng, Jingtao Ding, Nian Li, Yizhou Sun, Yong Li Published: 2026-04-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.23578v1 Human daily behavior unfolds as complex sequences shaped by intentions, preferences, and context. Effectively modeling these behaviors is crucial for intelligent systems such as personal assistants and recommendation engines. While recent advances in deep learning and behavior pre-training have improved behavior prediction, key challenges remain--particularly in handling long-tail behaviors, enhancing interpretability, and supporting multiple tasks within a unified framework. Large language models (LLMs) offer a promising direction due to their semantic richness, strong interpretability, and generative capabilities. However, the structural and modal differences between behavioral data and natural language limit the direct applicability of LLMs. To address this gap, we propose Behavior Understanding Alignment (BUA), a novel framework that integrates LLMs into human behavior modeling through a structured curriculum learning process. BUA employs sequence embeddings from pretrained behavior models as alignment anchors and guides the LLM through a three-stage curriculum, while a multi-round dialogue setting introduces prediction and generation capabilities. Experiments on two real-world datasets demonstrate that BUA significantly outperforms existing methods in both tasks, highlighting its effectiveness and flexibility in applying LLMs to complex human behavior modeling. --- ### RouteNLP: Closed-Loop LLM Routing with Conformal Cascading and Distillation Co-Optimization Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu Published: 2026-04-26 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2604.23577v1 Serving diverse NLP workloads with large language models is costly: at one enterprise partner, inference costs exceeded $200K/month despite over 70% of queries being routine tasks well within the capability of smaller models. We present RouteNLP, a closed-loop framework that routes queries across a tiered model portfolio to minimize cost while satisfying per-task quality constraints. The framework integrates three components: a difficulty-aware router with shared task-conditioned representations trained on preference data and quality signals; confidence-calibrated cascading that uses conformal prediction for distribution-free threshold initialization; and a distillation-routing co-optimization loop that clusters escalation failures, applies targeted knowledge distillation to cheaper models, and automatically retrains the router, yielding over twice the cost improvement of untargeted distillation. In an 8-week pilot deployment processing ~5K queries/day at an enterprise customer-service division, RouteNLP reduced inference costs by 58% while maintaining 91% response acceptance and reducing p99 latency from 1,847 ms to 387 ms. On a six-task benchmark spanning finance, customer service, and legal domains, the framework achieves 40-85% cost reduction while retaining 96-100% quality on structured tasks and 96-98% on generation tasks, with human evaluation confirming that 74.5% of routed generation outputs match or exceed frontier-model quality. --- ### CAPSULE: Control-Theoretic Action Perturbations for Safe Uncertainty-Aware Reinforcement Learning Authors: Rahul Narava, Siddharth Verma, Ojas Jain, Shashi Shekhar Jha, Mayank Shekhar Jha Published: 2026-04-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.23576v1 Ensuring safe exploration in high-dimensional systems with unknown dynamics remains a significant challenge. Existing safe reinforcement learning methods often provide safety guarantees only in expectation, which can still lead to safety violations. Control-theoretic approaches, in contrast, offer hard constraint-based safety guarantees but typically assume access to known system dynamics or require accurate estimation of control-affine models. In this paper, we propose a safe reinforcement learning framework that learns a probabilistic control-affine dynamics model in an offline setting. The learned model is leveraged to explicitly construct control barrier functions (CBFs) that incorporate model uncertainty to provide conservative safety constraints. These CBF constraints are enforced through an online constraint-based action correction mechanism, enabling safe exploration without overly restricting task performance. Empirical evaluations on nonlinear, complex continuous-control benchmarks demonstrate that our approach achieves returns comparable to those of existing baselines while significantly reducing safety violations. --- ### The Collapse of Heterogeneity in Silicon Philosophers Authors: Yuanming Shi, Andreas Haupt Published: 2026-04-26 Categories: cs.CY, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2604.23575v2 Silicon samples are increasingly used as a low-cost substitute for human panels and have been shown to reproduce aggregate human opinion with high fidelity. We show that, in the alignment-relevant domain of philosophy, silicon samples systematically collapse heterogeneity. Using data from $N = {277}$ professional philosophers drawn from PhilPeople profiles, we evaluate seven proprietary and open-source large language models on their ability to replicate individual philosophical positions and to preserve cross-question correlation structures across philosophical domains. We find that language models substantially over-correlate philosophical judgments, producing artificial consensus across domains. This collapse is associated in part with specialist effects, whereby models implicitly assume that domain specialists hold highly similar philosophical views. We assess the robustness of these findings by studying the impact of DPO fine-tuning and by validating results against the full PhilPapers 2020 Survey ($N = {1785}$). We conclude by discussing implications for alignment, evaluation, and the use of silicon samples as substitutes for human judgment. The code of this project can be found at https://github.com/stanford-del/silicon-philosophers. --- ### PhysLayer: Language-Guided Layered Animation with Depth-Aware Physics Authors: Tianyidan Xie, Zhentao Huang, Mingjie Wang, Xin Huang, Jun Zhou, Minglun Gong, Zili Yi Published: 2026-04-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.23574v1 Existing image-to-video generation methods often produce physically implausible motions and lack precise control over object dynamics. While prior approaches have incorporated physics simulators, they remain confined to 2D planar motions and fail to capture depth-aware spatial interactions. We introduce PhysLayer, a novel framework enabling language-guided, depth-aware layered animation of static images. PhysLayer consists of three key components: First, a language-guided scene understanding module that utilizes vision foundation models to decompose scenes into depth-based layers by analyzing object composition, material properties, and physical parameters. Second, a depth-aware layered physics simulation that extends 2D rigid-body dynamics with depth motion and perspective-consistent scaling, enabling more realistic object interactions without requiring full 3D reconstruction. Third, a physics-guided video synthesis module that integrates simulated trajectories with scene-aware relighting for temporally coherent results. Experimental results demonstrate improvements in CLIP-Similarity (+2.2\%), FID score (+9.3\%), and Motion-FID (+3\%), with human evaluation showing enhanced physical plausibility (+24\%) and text-video alignment (+35\%). Our approach provides a practical balance between physical realism and computational efficiency for controllable image animation. --- ### High-dimensional Semi-supervised Classification via the Fermat Distance Authors: Ruoxu Tan, Yiming Zang Published: 2026-04-26 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2604.23573v1 Semi-supervised classification, where unlabeled data are massive but labeled data are limited, often arises in machine learning applications. We address this challenge under high-dimensional data by leveraging the manifold and cluster assumptions. Based on the Fermat distance, a density-sensitive metric that naturally encodes the cluster assumption, we propose the weighted $k$-nearest neighbors (NN) classifier and multidimensional scaling (MDS)-induced classifiers. The use of MDS with a large target dimension allows the effective application of linear classifiers to complex manifold data. Theoretically, we derive a sharp lower bound for the expected excess risk within clusters and prove that the weighted $k$-NN classifier utilizing the true Fermat distance is minimax optimal. Furthermore, we explicitly quantify the utility of unlabeled data by showing that the error arising from estimating the Fermat distance decays exponentially with the pooled sample size. Such a rate is much faster than the related rates in the literature. Extensive experiments on synthetic and real datasets demonstrate competitive or superior performance of our approaches compared to state-of-the-art graph-based semi-supervised classifiers. --- ### CyberCane: Neuro-Symbolic RAG for Privacy-Preserving Phishing Detection with Formal Ontology Reasoning Authors: Safayat Bin Hakim, Aniqa Afzal, Qi Zhao, Vigna Majmundar, Pawel Sloboda, Houbing Herbert Song Published: 2026-04-26 Categories: cs.CR, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2604.23563v1 Privacy-critical domains require phishing detection systems that satisfy contradictory constraints: near-zero false positives to prevent workflow disruption, transparent explanations for non-expert staff, strict regulatory compliance prohibiting sensitive data exposure to external APIs, and robustness against AI-generated attacks. Existing rule-based systems are brittle to novel campaigns, while LLM-based detectors violate privacy regulations through unredacted data transmission. We introduce CyberCane, a neuro-symbolic framework integrating deterministic symbolic analysis with privacy-preserving retrieval-augmented generation (RAG). Our dual-phase pipeline applies lightweight symbolic rules to email metadata, then escalates borderline cases to semantic classification via RAG with automated sensitive data redaction and retrieval from a phishing-only corpus. We further introduce PhishOnt, an OWL ontology enabling verifiable attack classification through formal reasoning chains. Evaluation on DataPhish2025 (12.3k emails; mixed human/LLM) and Nazario/SpamAssassin demonstrates a 78.6-point recall gain over symbolic-only detection on AI-generated threats, with precision exceeding 98% and FPR as low as 0.16%. Healthcare deployment projects a 542x ROI; tunable operating points support diverse risk tolerances, with open-source implementation at https://github.com/sbhakim/Cybercane. --- ### EyeBrain: Left and Right Brain Lateralization Activity Classification Through Pupil Diameter and Fixation Duration Authors: Ko Watanabe, Pooja Pol, Nicolas Großmann, Shoya Ishimaru, Andreas Dengel Published: 2026-04-26 Categories: q-bio.NC, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2604.23562v1 The relationship between brain lateralization and cognitive functions is well-documented. The left hemisphere primarily handles tasks such as language and arithmetic, while the right hemisphere is involved in creative activities like drawing and music perception. Eye-tracking technology has shown the potential to reveal cognitive states by measuring ocular metrics such as pupil diameter and fixation duration. However, the ability to distinguish lateralized brain activity using these ocular metrics remains underexplored. Here, we demonstrate that pupil diameter and fixation duration can effectively classify left and right brain hemisphere activities. We obtained a considerably high classification performance, with an F1 score of 0.894. The results suggest that ocular metrics are robust indicators of lateralized brain activity and can be applied in cognitive monitoring and neurorehabilitation. Our future work expands on this by integrating these methods into real-time applications EyeBrain, potentially broadening their use across various cognitive and neurological domains. --- ### DLM: Unified Decision Language Models for Offline Multi-Agent Sequential Decision Making Authors: Zhuohui Zhang, Bin Cheng, Bin He Published: 2026-04-26 Categories: cs.MA, cs.AI arXiv: https://arxiv.org/abs/2604.23557v1 Building scalable and reusable multi-agent decision policies from offline datasets remains a challenge in offline multi-agent reinforcement learning (MARL), as existing methods often rely on fixed observation formats and action spaces that limit generalization. In contrast, large language models (LLMs) offer a flexible modeling interface that can naturally accommodate heterogeneous observations and actions. Motivated by this, we propose the Decision Language Model (DLM), which formulates multi-agent decision making as a dialogue-style sequence prediction problem under the centralized training with decentralized execution paradigm. DLM is trained in two stages: a supervised fine-tuning phase, which leverages dialogue-style datasets for centralized training with inter-agent context and generates executable actions from offline trajectories, followed by a group relative policy optimization phase to enhance robustness to out-of-distribution actions through lightweight reward functions. Experiments on multiple benchmarks show that a unified DLM outperforms strong offline MARL baselines and LLM-based conversational decision-making methods, while demonstrating strong zero-shot generalization to unseen scenarios across tasks. --- ### On the Memorization of Consistency Distillation for Diffusion Models Authors: Bingqing Jiang, Difan Zou Published: 2026-04-26 Categories: cs.LG, cs.AI, stat.ML arXiv: https://arxiv.org/abs/2604.23552v1 Diffusion models are central to modern generative modeling, and understanding how they balance memorization and generalization is critical for reliable deployment. Recent work has shown that memorization in diffusion models is shaped by training dynamics, with generalization and memorization emerging at different stages of training. However, deployed diffusion models are often further distilled, introducing an additional training phase whose impact on memorization is not well understood. In this work, we analyze how distillation reshapes memorization behavior in diffusion models, taking consistency distillation as a representative framework. Empirically, we show that when applied to a teacher model that has memorized data, consistency distillation significantly reduces transferred memorization in the student while preserving, and sometimes improving, sample quality. To explain this behavior, we provide a theoretical analysis using a random feature neural network model (Bonnaire et al., 2025), showing that consistency distillation suppresses unstable feature directions associated with memorization while preserving stable, generalizable modes. Our findings suggest that distillation can serve not only as an acceleration tool, but also as a mechanism for improving the memorization-generalization trade-off. --- ### Spatiotemporal Degradation-Aware 3D Gaussian Splatting for Realistic Underwater Scene Reconstruction Authors: Shaohua Liu, Ning Gao, Zuoya Gu, Hongkun Dou, Yue Deng, Hongjue Li Published: 2026-04-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.23551v1 Reconstructing realistic underwater scenes from underwater video remains a meaningful yet challenging task in the multimedia domain. The inherent spatiotemporal degradations in underwater imaging, including caustics, flickering, attenuation, and backscattering, frequently result in inaccurate geometry and appearance in existing 3D reconstruction methods. While a few recent works have explored underwater degradation-aware reconstruction, they often address either spatial or temporal degradation alone, falling short in more real-world underwater scenarios where both types of degradation occur. We propose MarineSTD-GS, a novel 3D Gaussian Splatting-based framework that explicitly models both temporal and spatial degradations for realistic underwater scene reconstruction. Specifically, we introduce two paired Gaussian primitives: Intrinsic Gaussians represent the true scene, while Degraded Gaussians render the degraded observations. The color of each Degraded Gaussian is physically derived from its paired Intrinsic Gaussian via a Spatiotemporal Degradation Modeling (SDM) module, enabling self-supervised disentanglement of realistic appearance from degraded images. To ensure stable training and accurate geometry, we further propose a Depth-Guided Geometry Loss and a Multi-Stage Optimization strategy. We also construct a simulated benchmark with diverse spatial and temporal degradations and ground-truth appearances for comprehensive evaluation. Experiments on both simulated and real-world datasets show that MarineSTD-GS robustly handles spatiotemporal degradations and outperforms existing methods in novel view synthesis with realistic, water-free scene appearances. --- ### Hardware-Efficient FPGA Implementation of Sigmoid Function Using Mixed-Radix Hyperbolic Rotation CORDIC Authors: Chintan Panchal, Ankur Changela, Mohendra Roy Published: 2026-04-26 Categories: cs.AR, cs.AI arXiv: https://arxiv.org/abs/2604.23547v1 Efficient hardware implementation of nonlinear activation functions is a crucial task in deploying artificial neural networks on resource-constrained and edge devices such as Field-Programmable Gate Arrays (FPGAs). The sigmoid activation function is widely used for probabilistic output, binary classification, and gating mechanisms in recurrent neural networks, despite its reliance on exponential computations. This paper presents a hardware-efficient FPGA implementation of the sigmoid activation function using a mixed-radix CORDIC-based architecture. The proposed approach leverages the mathematical relationship between the sigmoid and hyperbolic tangent functions. The input range is normalized to 1, enabling the corresponding tanh computation to operate within a reduced range of 0.5, which significantly improves convergence behavior. To achieve high accuracy with minimal hardware overhead, a modified mixed-radix hyperbolic rotation CORDIC (MR-HRC) algorithm combining radix-2 and radix-4 iterations is introduced. The initial radix-2 stage ensures stable convergence, while the subsequent radix-4 stage accelerates convergence without requiring scale-factor compensation. In the final stage, a radix-2 linear vectoring CORDIC (R2-LVC) is used to compute the hyperbolic tangent by dividing hyperbolic sine and cosine values derived from the MR-HRC algorithm. The entire architecture is fully pipelined and implemented on an FPGA. The design is realized on an Xilinx Virtex-7 FPGA using a 16-bit fixed-point representation. Experimental results demonstrate a significant reduction in hardware utilization, requiring only 835 logic slices with zero DSP usage. Additionally, the design achieves a mean absolute error of 4.23 10^-4, outperforming several recent sigmoid implementations. --- ### COMO: Closed-Loop Optical Molecule Recognition with Minimum Risk Training Authors: Zhuoqi Lyu, Qing Ke Published: 2026-04-26 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.23546v1 Optical chemical structure recognition (OCSR) translates molecular images into machine-readable representations like SMILES strings or molecular graphs, but remains challenging in real-world documents due to inexhaustible variations in chemical structures, shorthand conventions, and visual noise. Most existing deep-learning-based approaches rely on teacher forcing with token-level Maximum Likelihood Estimation (MLE). This training paradigm suffers from exposure bias, as models are trained under ground-truth prefixes but must condition on their own previous predictions during inference. Moreover, token-level MLE objectives hinder the optimization towards molecular-level evaluation criteria such as chemical validity and structural similarity. Here we introduce Minimum Risk Training (MRT) to OCSR and propose COMO (Closed-loop Optical Molecule recOgnition), a closed-loop framework that mitigates exposure bias by directly optimizing over molecule-level, non-differentiable objectives, by iteratively sampling and evaluating the model's own predictions. Experiments on ten benchmarks including synthetic and real-world chemical diagrams from patent and scientific literature demonstrate that COMO substantially outperforms existing rule-based and learning-based methods with less training data. Ablation studies further show that MRT is architecture-agnostic, demonstrating its potential for broad application to end-to-end OCSR systems. --- ### Pref-CTRL: Preference Driven LLM Alignment using Representation Editing Authors: Imranul Ashrafi, Inigo Jauregi Unanue, Massimo Piccardi Published: 2026-04-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.23543v1 Test-time alignment methods offer a promising alternative to fine-tuning by steering the outputs of large language models (LLMs) at inference time with lightweight interventions on their internal representations. Recently, a prominent and effective approach, RE-Control (Kong et al., 2024), has proposed leveraging an external value function trained over the LLM's hidden states to guide generation via gradient-based editing. While effective, this method overlooks a key characteristic of alignment tasks, i.e. that they are typically formulated as learning from human preferences between candidate responses. To address this, in this paper we propose a novel preference-based training framework, Pref-CTRL, that uses a multi-objective value function to better reflect the structure of preference data. Our approach has outperformed RE-Control on two benchmark datasets and showed greater generalization on out-of-domain datasets. Our source code is available at https://github.com/UTS-nlPUG/pref-ctrl. --- ### AusSmoke meets MultiNatSmoke: a fully-labelled diverse smoke segmentation dataset Authors: Weihao Li, Hongjin Zhao, Gao Zhu, Ge-Peng Ji, Nicholas Wilson, Marta Yebra, Nick Barnes Published: 2026-04-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.23542v1 Wildfires are an escalating global concern due to the devastating impacts on the environment, economy, and human health, with notable incidents such as the 2019-2020 Australian bushfires and the 2025 California wildfires underscoring the severity of these events. AI-enabled camera-based smoke detection has emerged as a promising approach for the rapid detection of wildfires. However, existing wildfire smoke segmentation datasets that are used for training detection and segmentation models are limited in scale, geographically constrained, and often rely on synthetic imagery, which hinders effective training and generalization. To overcome these limitations, we present AusSmoke, a new smoke segmentation dataset collected from Australia to address the data scarcity in this region. Furthermore, we introduce a MultiNational geographically diverse and substantially larger fully-labelled benchmark, called MultiNatSmoke, that consolidates publicly available international datasets with the newly collected Australian imagery, expanding the scale by an order of magnitude over previous collections. Finally, we benchmark smoke segmentation models, demonstrating improved performance and enhanced generalization across diverse geographical contexts. The project is available at \href{https://github.com/henryzhao0615/MultiNatSmoke}{Github}. --- ### Oracle Noise: Faster Semantic Spherical Alignment for Interpretable Latent Optimization Authors: Haosen Li, Wenshuo Chen, Lei Wang, Shaofeng Liang, Haozhe Jia, Yutao Yue Published: 2026-04-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.23540v1 Text-to-image diffusion models have achieved remarkable generative capabilities, yet accurately aligning complex textual prompts with synthesized layouts remains an ongoing challenge. In these models, the initial Gaussian noise acts as a critical structural seed dictating the macroscopic layout. Recent online optimization and search methods attempt to refine this noise to enhance text-image alignment. However, relying on unconstrained Euclidean gradient ascent mathematically inflates the latent norm and destroys the standard Gaussian prior, causing severe visual artifacts like color over-saturation. Furthermore, these methods suffer from inefficient semantic routing and easily fall into the ``reward hacking'' trap of external proxy models. To address these intertwined bottlenecks, we propose Oracle Noise, a zero-shot framework reframing noise initialization as semantic-driven optimization strictly confined to a Riemannian hypersphere. Instead of relying on complex external parsers, we directly identify the most impactful structural words in the prompt to efficiently route optimization energy. By updating the noise strictly along a spherical path, we mathematically preserve the original Gaussian distribution. This geometric constraint eliminates norm inflation and unlocks aggressive step sizes for rapid convergence. Extensive experiments demonstrate that Oracle Noise significantly accelerates semantic alignment and achieves superior aesthetics without black-box models. It completely mitigates Euclidean-induced degradation, establishing state-of-the-art performance across human preference metrics (e.g., HPSv2, ImageReward), semantic alignment (CLIP Score), and sample diversity, all within a strict 2-second optimization budget. --- ### MetaGAI: A Large-Scale and High-Quality Benchmark for Generative AI Model and Data Card Generation Authors: Haoxuan Zhang, Ruochi Li, Yang Zhang, Zhenni Liang, Junhua Ding, Ting Xiao, Haihua Chen Published: 2026-04-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.23539v1 The rapid proliferation of Generative AI necessitates rigorous documentation standards for transparency and governance. However, manual creation of Model and Data Cards is not scalable, while automated approaches lack large-scale, high-fidelity benchmarks for systematic evaluation. We introduce MetaGAI, a comprehensive benchmark comprising 2,541 verified document triplets constructed through semantic triangulation of academic papers, GitHub repositories, and Hugging Face artifacts. Unlike prior single-source datasets, MetaGAI employs a multi-agent framework with specialized Retriever, Generator, and Editor agents, validated through four-dimensional human-in-the-loop assessment, including human evaluation of editor-refined ground truth. We establish a robust evaluation protocol combining automated metrics with validated LLM-as-a-Judge frameworks. Extensive analysis reveals that sparse Mixture-of-Experts architectures achieve superior cost-quality efficiency, while a fundamental trade-off exists between faithfulness and completeness. MetaGAI provides a foundational testbed for benchmarking, training, and analyzing automated Model and Data Card generation methods at scale. Our data and code are available at: https://github.com/haoxuan-unt2024/MetaGAI-Benchmark. --- ### $Z^2$-Sampling: Zero-Cost Zigzag Trajectories for Semantic Alignment in Diffusion Models Authors: Haosen Li, Wenshuo Chen, Shaofeng Liang, Lei Wang, Kaishen Yuan, Yutao Yue Published: 2026-04-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.23536v1 Diffusion models have achieved unprecedented success in text-aligned generation, largely driven by Classifier-Free Guidance (CFG). However, standard CFG operates strictly on instantaneous gradients, omitting the intrinsic curvature of the data manifold. Recent methods like Zigzag-sampling (Z-Sampling) explicitly traverse multi-step forward-backward trajectories to probe this curvature, significantly improving semantic alignment. Yet, these explicit traversals triple the Neural Function Evaluation (NFE) cost and introduce unconstrained truncation errors from off-manifold evaluations, causing cumulative drift from the true marginal distribution. In this paper, we theoretically demonstrate that the explicit zigzag sequence is topologically reducible. We propose Implicit Z-Sampling, rigorously proving that intermediate states can be algebraically annihilated via operator dualities, physically eliminating off-manifold approximation errors. To push sampling efficiency to its theoretical lower bound, we introduce $Z^2$-Sampling (Zero-cost Zigzag Sampling). Exploiting the Probability Flow ODE's temporal coherence, $Z^2$-Sampling couples implicit algebraic collapse with a dynamically cached Temporal Semantic Surrogate. This restores the standard 2-NFE baseline without sacrificing semantic exploration. We formally prove via Backward Error Analysis that this discrete collapse inherently synthesizes a directional derivative curvature penalty. Finally, extensive evaluations demonstrate that $Z^2$-Sampling structurally shatters the performance-efficiency Pareto frontier. We validate its universal applicability across diverse architectures (U-Nets, DiTs) and modalities (image/video), establishing seamless orthogonality with advanced alignment frameworks (AYS, Diffusion-DPO). --- ### Emotion-Conditioned Short-Horizon Human Pose Forecasting with a Lightweight Predictive World Model Authors: Jingni Huang, Peter Bloodsworth Published: 2026-04-26 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2604.23532v1 Short-term human pose prediction plays a crucial role in interactive systems, assistive robots, and emotion-aware human-computer interaction(1-3). While current trajectory prediction models primarily rely on geometric motion cues, they often overlook the underlying emotional signals influencing human motion dynamics(4-5). This paper investigates whether facial expression-derived emotion embeddings can provide auxiliary conditional signals for short-term pose prediction. To further evaluate multimodal conditionation in a recursive prediction setting, we propose a lightweight autoregressive predictive world model that performs 15-step rolling pose prediction. This framework combines pose keypoints with emotion embeddings through a learnable gating mechanism and performs autoregressive unfolding prediction using a recurrent sequence model based on a two-layer LSTM architecture. Experiments were conducted on two small-scale pose-emotion video datasets: controlled motion sequences with minimal facial expression changes and, natural emotion-driven motion sequences with considerable facial expression changes. The results show that simple multimodal fusion does not consistently improve prediction accuracy, while normalized gating fusion significantly enhances the performance of emotion-driven motion sequences. Furthermore, counterfactual perturbation experiments demonstrate that the predicted trajectory exhibits measurable sensitivity to changes in multimodal input, suggesting that facial expression embeddings act as auxiliary conditional signals rather than redundant features. In summary, these results indicate that incorporating facial expression-derived emotion embeddings into emotion-conditional short-term pose prediction based on a lightweight predictive world model architecture is a feasible approach. --- ### MTRouter: Cost-Aware Multi-Turn LLM Routing with History-Model Joint Embeddings Authors: Yiqun Zhang, Hao Li, Zihan Wang, Shi Feng, Xiaocui Yang, Daling Wang, Bo Zhang, Lei Bai, Shuyue Hu Published: 2026-04-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.23530v1 Multi-turn, long-horizon tasks are increasingly common for large language models (LLMs), but solving them typically requires many sequential model invocations, accumulating substantial inference costs. Here, we study cost-aware multi-turn LLM routing: selecting which model to invoke at each turn from a model pool, given a fixed cost budget. We propose MTRouter, which encodes the interaction history and candidate models into joint history-model embeddings, and learns an outcome estimator from logged trajectories to predict turn-level model utility. Experiments show that MTRouter improves the performance-cost trade-off: on ScienceWorld, it surpasses GPT-5 while reducing total cost by 58.7%; on Humanity's Last Exam (HLE), it achieves competitive accuracy while reducing total cost by 43.4% relative to GPT-5, and these gains even carry over to held-out tasks. Further analyses reveal several mechanisms underlying its effectiveness: relative to prior multi-turn routers, MTRouter makes fewer model switches, is more tolerant to transient errors, and exhibits emergent specialization across models. Code: https://github.com/ZhangYiqun018/MTRouter --- ### When PINNs Go Wrong: Pseudo-Time Stepping Against Spurious Solutions Authors: Sifan Wang, Shawn Koohy, Yiping Lu, Paris Perdikaris Published: 2026-04-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.23528v1 Physics-informed neural networks (PINNs) provide a promising machine learning framework for solving partial differential equations, but their training often breaks down on challenging problems, sometimes converging to physically incorrect solutions despite achieving small residual losses. This failure, we argue, is not merely an optimization difficulty. Rather, it reflects a fundamental weakness of the empirical PDE residual loss, which can admit trivial or spurious solutions during training. From this perspective, we revisit pseudo-time stepping, a technique that has recently shown strong empirical success in PINNs. We show that its main benefit is not simply to ease optimization; instead, when combined with collocation-point resampling, it helps reveal and avoid spurious solutions. At the same time, we find that the effectiveness of pseudo-time stepping depends critically on the choice of step size, which cannot be tuned reliably from the training loss alone. To overcome this limitation, we propose an adaptive pseudo-time stepping strategy that selects the step size from a finite-difference surrogate of the local residual Jacobian, yielding the largest step permitted by local stability without per-problem tuning. Across a diverse set of PDE benchmarks, the proposed method consistently improves both accuracy and robustness. Together, these findings provide a clearer understanding of why PINNs fail and suggest a practical pathway toward more reliable physics-informed learning. All code and data accompanying this manuscript are available at https://github.com/sifanexisted/jaxpi2. --- ### Grammar-Constrained Refinement of Safety Operational Rules Using Language in the Loop: What Could Go Wrong Authors: Khouloud Gaaloul, Zaid Ghazal, Madhu Latha Pulimi, Sam Emmanuel Kathiravan Published: 2026-04-26 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.23523v1 Safety specifications in cyber-physical systems (CPS) capture the operational conditions the system must satisfy to operate safely within its intended environment. As operating environments evolve, operational rules must be continuously refined to preserve consistency with observed system behavior during simulation-based verification and validation. Revising inconsistent rules is challenging because the changes must remain syntactically correct under a domain-specific grammar. Language-in-the-loop refinement further raises safety concerns beyond syntactic violations, as it can produce semantically unjustified refinements that overfit to the observed outcomes. We introduce a framework that combines counterfactual reasoning with a grammar-constrained refinement loop to refine operational rules, aligning them with the observed system behavior. Applied to an autonomous driving control system, our approach successfully resolved the inconsistencies in an operational rule inferred by a conventional baseline while remaining grammar compliant. An empirical large language model (LLM) study further revealed model-dependent refinement quality and safety lessons, which motivate rigorous grammar enforcement, stronger semantic validation, and broader evaluation in future work. --- ### Multi-Plane HyperX: A Low-Latency and Cost-Effective Network for Large-Scale AI and HPC Systems Authors: Ziyu Wang, Fei Lei, Dezun Dong Published: 2026-04-26 Categories: cs.NI, cs.LG arXiv: https://arxiv.org/abs/2604.23519v1 Multi-plane architectures have become increasingly prevalent in the Fat-Tree networks of AI data centers. By leveraging multiple ports on a single network interface card (NIC) or multiple NICs within a scale-up domain, each port or NIC is allocated to an independent network plane, thereby provisioning the overall system with multiple network planes. However, no prior literature has explored the application of multi-plane technologies to direct networks such as HyperX. This paper investigates the multi-plane HyperX network and demonstrates that, compared to state-of-the-art network topologies like multi-plane Fat-Tree, Dragonfly, and Dragonfly+, the multi-plane HyperX architecture achieves a significantly smaller network diameter and superior cost-effectiveness. --- ### Autocorrelation Reintroduces Spectral Bias in KANs for Time Series Forecasting Authors: Chen Zeng, Jiahui Wang, Qiao Wang Published: 2026-04-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.23518v1 Existing theory suggests that Kolmogorov-Arnold Networks (KANs) can overcome the spectral bias commonly observed in neural networks under the assumption that inputs are statistically independent. However, this assumption does not hold in time series forecasting (TSF), where inputs are lagged observations with strong temporal autocorrelation. Through theoretical analysis and empirical validation, we obtain an unexpected finding: temporal autocorrelation reintroduces spectral bias in KANs, and the bias becomes increasingly pronounced as the degree of autocorrelation increases. This suggests that standard KANs may face substantial difficulties in TSF with strongly autocorrelated inputs. To address this problem, we introduce the Discrete Cosine Transform (DCT) to reduce the correlations among the network inputs. As expected, experimental results reveal that DCT preprocessing substantially reduces the observed low-frequency preference in TSF. This result also corroborates that the spectral bias of KANs in TSF tasks is indeed induced by the autocorrelation among input variables. --- ### Probabilistic Graphical Model using Graph Neural Networks for Bayesian Inversion of Discrete Structural Component States Authors: Teng Li, Stephen Wu, Yong Huang, James L. Beck, Hui Li Published: 2026-04-26 Categories: stat.ML, cs.LG, stat.ME arXiv: https://arxiv.org/abs/2604.23514v2 The health condition of components in civil infrastructures can be described by various discrete states according to their performance degradation. Inferring these states from measurable responses is typically an ill-posed inverse problem. Although Bayesian methods are well-suited to tackle such problems, computing the posterior probability density function (PDF) presents challenges. The likelihood function cannot be analytically formulated due to the unclear relationship between discrete states and structural responses, and the high-dimensional state parameters resulting from numerous components severely complicates the computation of the marginal likelihood function. To address these challenges, this study proposes a novel Bayesian inversion paradigm for discrete variables based on Probabilistic Graphical Models (PGMs). The Markov networks are employed as modeling tools, with model parameters learned from data and structural topology prior. It has been proved that inferring this PGM produces the same probabilistic estimation as the posterior PDF derived from Bayesian inference, which effectively solves the above challenges. The inference is accomplished by Graph Neural Networks (GNNs), and a graph property-based GNN training strategy is developed to enable accurate inference across varying graph scales, thereby significantly reducing the computational overhead in high-dimensional problems. Both synthetic and experimental data are used to validate the proposed framework --- ### BurstGP: Enhancing Raw Burst Image Super Resolution with Generative Priors Authors: Dong Huo, Tristan Aumentado-Armstrong, Samrudhdhi B. Rangrej, Maitreya Suin, Angela Ning Ye, Zhiming Hu, Amanpreet Walia, Amirhossein Kazerouni, Konstantinos G. Derpanis, Iqbal Mohomed, Alex Levinshtein Published: 2026-04-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.23508v1 Burst image super resolution (BISR) aims to construct a single high-resolution (HR) image by aggregating information from multiple low-resolution (LR) frames, relying on temporal redundancy and spatial coherence across the burst. While conventional methods achieve impressive results, they often struggle with complex textures and oversmoothing. Diffusion models, particularly those pretrained on high-quality data, have shown remarkable capability in generating realistic details for image and video super-resolution. However, their potential remains largely under-explored in BISR, where existing approaches typically rely on task-specific diffusion models trained from scratch and operate on single-frame reconstructions. In this work, we propose BurstGP, a novel diffusion-based solution for BISR, which leverages generative priors of recent foundation models to overcome these issues. In particular, we build a multiframe-aware diffusion model on top of a conventional BISR approach, which boosts image quality with minimal loss to fidelity. Further, we introduce (i) a novel degradation-aware conditioning mechanism, which controls synthesis of fine details based on the estimated degradation in the input, and (ii) a robust sRGB-to-lRGB inverter, enabling us to utilize generative multiframe (video) sRGB priors, while operating with raw input and lRGB output images. Empirically, we demonstrate that BurstGP outperforms the existing state of the art, both quantitatively (especially with respect to perceptual metrics, including MUSIQ and LPIPS) and qualitatively. In particular, our proposed method excels at recovering richer textures and finer structural details, highlighting the potential of video priors for BISR over traditional methods. --- ### Uncertainty Propagation in LLM-Based Systems Authors: Boming Xia, Liming Zhu, Erdun Gao, Qinghua Lu, Minhui Xue, Dino Sejdinovic Published: 2026-04-26 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.23505v1 Uncertainty in large language model (LLM)-based systems is often studied at the level of a single model output, yet deployed LLM applications are compound systems in which uncertainty is transformed and reused across model internals, workflow stages, component boundaries, persistent state, and human or organisational processes. Without principled treatment of how uncertainty is carried and reused across these boundaries, early errors can propagate and compound in ways that are difficult to detect and govern. This paper develops a systems-level account of uncertainty propagation. It introduces a conceptual framing for characterising propagated uncertainty signals, presents a structured taxonomy spanning intra-model (P1), system-level (P2), and socio-technical (P3) propagation mechanisms, synthesises cross-cutting engineering insights, and identifies five open research challenges. --- ### Interpretable Physics-Informed Load Forecasting for U.S. Grid Resilience: SHAP-Guided Ensemble Validation in Hybrid Deep Learning Under Extreme Weather Authors: Md Abubakkar, Sajib Debnath, Md. Uzzal Mia Published: 2026-04-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.23500v1 Accurate short-term electricity load forecasting is a cornerstone of U.S. grid reliability; however, prevailing deep learning models remain opaque, limiting operator trust during extreme weather. A unified, interpretable, physics-informed ensemble framework is proposed, integrating a Convolutional Neural Network (CNN) branch for local feature extraction and a Transformer branch for long-range dependency modeling; the branches are fused through a validation-optimized weighted ensemble and regularized by a physics-informed loss derived from the piecewise parabolic temperature-demand relationship of the Electric Reliability Council of Texas (ERCOT) system. Post-hoc interpretability is provided through SHapley Additive exPlanations (SHAP) with the DeepExplainer backend, yielding global and event-level attributions. Using eight years of ERCOT hourly load data (2018-2025) fused with Automated Surface Observing System (ASOS) records from three Texas stations, the framework achieves 713 MW MAE, 812 MW RMSE, and 1.18% MAPE on the test window. For Hampel-flagged extreme events, MAPE falls by 20.7% relative to its Transformer branch and by 40.5% relative to its CNN branch; an ablation confirms that the parabolic and ramp constraints drive a 14.7% RMSE reduction. SHAP analysis reveals a regime shift: temperature dominates under normal operation, whereas wind speed and precipitation become more influential during cold fronts and heatwaves. --- ### When Does Dynamic Preconditioning Preserve the Polyak-Ruppert CLT? A Stabilization Threshold Authors: Sunyoung An, Xiaoming Huo Published: 2026-04-26 Categories: math.ST, math.OC, stat.ML arXiv: https://arxiv.org/abs/2604.23498v1 Polyak-Ruppert averaging yields an asymptotically normal estimator with sandwich covariance $H^{-1}SH^{-1}$, the foundation of online inference. When the gradient step is preconditioned by a data-driven matrix $P_t$, we ask how fast $P_t$ must stabilize for the central limit theorem (CLT) to remain valid. We resolve this via an exact preconditioner-isolating decomposition of the averaged error that confines $P_t$ to a dynamic remainder $R_n$, leaving the martingale and Taylor terms preconditioner-free. Let $M_t = (P_t H)^{-1}$ denote the effective inverse drift matrix, with $\|M_t - M_{t-1}\|_{\mathrm{op}} \lesssim t^{-β}$ and step-size exponent $α\in (1/2, 1)$. We identify a stabilization-rate threshold $β> (α+1)/2$ and prove that, within the class of polynomial rate hypotheses used in our upper bound, it cannot be weakened: the dynamic remainder $\sqrt{n}\,R_n$ vanishes in $L^2$ whenever $β> (α+1)/2$, and we exhibit sequences satisfying those hypotheses for which it does not vanish when $β\le (α+1)/2$. A single stabilization argument certifies three SA variants - SA-AdaGrad, SA-RMSProp, and SA-ONS - with gain $ρ_t = c/t$, each delivering one-step $L^2(\mathrm{op})$ stabilization of order $t^{-1}$, yielding the CLT $\sqrt{n}(\bar{x}_n - x^*) \to N(0, H^{-1}SH^{-1})$; under bounded inputs the pathwise rate $β= 1$ further preserves the $n^{-1/6}$ Wasserstein rate at $α^* = 2/3$. Under standard regularity conditions, Wald-type online inference remains valid for dynamically preconditioned averaged SGD whose stabilization rate exceeds the threshold. --- ### Do Transaction-Level and Actor-Level AML Queues Agree? An Empirical Evaluation of Granularity Effects on the Elliptic++ Graph Authors: Ankur Malik Published: 2026-04-26 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.23494v1 Graph-based anti-money laundering (AML) systems on blockchain networks can score suspicious activity at two granularity levels -- transactions or actor addresses -- yet compliance action is conducted per actor. This paper contributes an evaluation methodology for measuring how scoring granularity affects investigation queue composition under fixed review budgets. We formalize the evaluation through a projection framework mapping transaction-level scores to the actor-level action unit via four aggregation operators, and introduce budgeted investigation metrics -- yield@budget, burden decomposition, and case fragmentation. Using the public Elliptic++ Bitcoin dataset (203,769 transactions; 822,942 address occurrences), we train independent random forest classifiers at each level under a causal temporal protocol and compare review queues through Jaccard overlap, burden decomposition, and feature-matching ablations. At one-percent budget, temporal evaluation yields mean Jaccard of 0.374 (SD 0.171); static pooled evaluation yields 0.087 (95% CI (0.079, 0.094)). An enriched address model receiving all 237 features produces even lower overlap (Jaccard=0.051), with 4.3% illicit per 100 reviews versus 30.2% for the transaction-projected queue. Address-level detection value is temporally concentrated: two timesteps exceed 91% illicit per 100 reviews while the static burden is only 3.4%. A fixed hybrid policy underperforms the best single-level queue by 5.05pp (CI (-10.2pp, -0.9pp)). These findings establish that scoring granularity is a consequential design variable for AML investigation systems -- same data, same budget, different queues, different addresses investigated. --- ### K-SENSE: A Knowledge-Guided Self-Augmented Encoder for Neuro-Semantic Evaluation of Mental Health Conditions on Social Media Authors: Vijay Yadav Published: 2026-04-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.23493v2 Early detection of mental health conditions, particularly stress and depression, from social media text remains a challenging open problem in computational psychiatry and natural language processing. Automated systems must contend with figurative language, implicit emotional expression, and the high noise inherent in user-generated content. Existing approaches either leverage external commonsense knowledge to model mental states explicitly, or apply self-augmentation and contrastive training to improve generalization, but seldom do both in a principled, unified framework. We propose K-SENSE (Knowledge-guided Self-augmented Encoder for Neuro-Semantic Evaluation of Mental Health), a framework that jointly exploits external psychological reasoning and internal representation robustness. K-SENSE adopts a three-stage encoding pipeline: (1) inferential commonsense knowledge is extracted from the COMET model across five mental state dimensions; (2) a semantic anchor is constructed by combining hidden representations from two parallel encoding streams, projected into a shared space before fusion; and (3) a supervised contrastive learning objective aligns same-class representations while encouraging the attention mechanism to suppress irrelevant knowledge noise. We evaluate K-SENSE on Dreaddit (stress detection) and Depression_Mixed (depression detection), achieving mean F1-scores of 86.1 (0.6%) and 94.3 (0.8%), respectively, over five independent runs. These represent improvements of approximately 2.6 and 1.5 percentage points over the strongest prior baselines. Ablation experiments confirm the contribution of each architectural component, including the temporal knowledge integration strategy and the choice to keep the knowledge encoder frozen during fine-tuning. --- ### Do Prompt-Elicited Trajectories Reflect Training-Time Reward Hacking? A Systematic Study on Monitoring Training-Time Reward Hacking in Code Generation Authors: Lichen Li, Hengguang Zhou, Yijun Liang, Tianyi Zhou, Cho-Jui Hsieh Published: 2026-04-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.23488v3 Reward hacking in code generation, where models exploit evaluation loopholes to obtain high reward without correctly solving the intended task, poses a critical challenge for Reinforcement Learning (RL) and the deployment of reasoning models. Existing studies often rely on explicitly prompted hacking trajectories, but it remains unclear whether monitors trained on such data can detect reward hacks that arise without direct hacking instructions during RL training. In this work, we introduce Trace-and-Amplify, a framework for scalable curation of reward-hacking trajectories that arise during RL training without explicit hacking instructions. The framework uses unit-test tracers to identify hacking solutions when they occur and retains such trajectories for monitor training and evaluation. Through controlled comparisons between monitors trained on prompt-elicited hacking trajectories and training-time reward-hacking trajectories collected by Trace-and-Amplify, we find that \textbf{(1) prompt-elicited-data-trained monitors often fail to generalize to trajectories curated by our framework}, and \textbf{(2) monitors trained on our Trace-and-Amplify trajectories demonstrate stronger generalizability to unseen hacking types}. Our results indicate that prompted reward hacking data may not fully reflect training-time reward-hacking behaviors, and that relying solely on these data can lead to misleading conclusions. Codebase is available at https://github.com/LichenLillc/CoTMonitoring.git --- ### Your Students Don't Use LLMs Like You Wish They Did Authors: Sebastian Kobler, Matthew Clemson, Angela Sun, Jonathan K. Kummerfeld Published: 2026-04-26 Categories: cs.CL, cs.CY, cs.HC arXiv: https://arxiv.org/abs/2604.23486v1 Educational NLP systems are typically evaluated using engagement metrics and satisfaction surveys, which are at best a proxy for meeting pedagogical goals. We introduce six computational metrics for automated evaluation of pedagogical alignment in student-AI dialogue. We validate our metrics through analysis of 12,650 messages across 500 conversations from four courses. Using our metrics, we identify a fundamental misalignment: educators design conversational tutors for sustained learning dialogue, but students mainly use them for answer-extraction. Deployment context is the strongest predictor of usage patterns, outweighing student preference or system design: when AI tools are optional, usage concentrates around deadlines; when integrated into course structure, students ask for solutions to verbatim assignment questions. Whole-dialogue evaluation misses these turn-by-turn patterns. Our metrics will enable researchers building educational dialogue systems to measure whether they are achieving their pedagogical goals. --- ### Agentic Adversarial Rewriting Exposes Architectural Vulnerabilities in Black-Box NLP Pipelines Authors: Mazal Bethany, Kim-Kwang Raymond Choo, Nishant Vishwamitra, Peyman Najafirad Published: 2026-04-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.23483v1 Multi-component natural language processing (NLP) pipelines are increasingly deployed for high-stakes decisions, yet no existing adversarial method can test their robustness under realistic conditions: binary-only feedback, no gradient access, and strict query budgets. We formalize this strict black-box threat model and propose a two-agent evasion framework operating in a semantic perturbation space. An Attacker Agent generates meaning-preserving rewrites while a Prompt Optimization Agent refines the attack strategy using only binary decision feedback within a 10-query budget. Evaluated against four evidence-based misinformation detection pipelines, the framework achieves evasion rates of 19.95 to 40.34% on modern large language model (LLM) based systems, compared to at most 3.90% for token-level perturbation baselines that rely on surrogate models because they cannot operate under our threat model. A legacy system relying on static lexical retrieval exhibits near-total vulnerability 97.02%, establishing a lower bound that exposes how architectural choices govern the attack surface. Evasion effectiveness is associated with three architectural properties: evidence retrieval mechanism, retrieval-inference coupling, and baseline classification accuracy. The iterative prompt optimization yields the largest marginal gains against the most robust targets, confirming that adaptive strategy discovery is essential when evasion is non-trivial. Analysis of successful rewrites reveals four exploitation patterns, each targeting failures at distinct pipeline stages. A pattern-informed defense reduces the evasion rate by up to 65.18%. --- ### Leveraging Spatial Transcriptomics as Alternative to Manual Annotations for Deep Learning-Based Nuclei Analysis Authors: Kazuya Nishimura, Ryoma Bise, Haruka Hirose, Yasuhiro Kojima Published: 2026-04-26 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2604.23481v1 Deep learning-based nuclei segmentation and classification in pathology images typically rely on large-scale pixel-level manual annotations, which are costly and difficult to obtain across diverse tissues and staining conditions. To address this limitation, we propose a framework that leverages spatial transcriptomics (ST) data as supervision for nuclei segmentation and classification. By incorporating cell-level ST data, we obtain gene expression profiles and corresponding nuclear masks from histopathological images. Gene expression profiles are converted into cell-type labels and used as training data for image-based classification. Because existing gene expression-based cell-type classification methods are not designed for image recognition, we introduce an image-oriented classification approach that bridges gene expression-based cell typing and image-based cell classification. To evaluate generalization, we conduct segmentation experiments on previously unseen organs and compare our method with conventional supervised models. Despite being trained on fewer organ types, our framework achieves higher segmentation accuracy, demonstrating strong transferability. Classification experiments further show consistent improvements over existing approaches. --- ### JudgeSense: A Benchmark for Prompt Sensitivity in LLM-as-a-Judge Systems Authors: Rohith Reddy Bellibatlu, Edward Raff, Wenbin Zhang Published: 2026-04-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.23478v3 Large language models are widely used to judge the output of other language models, yet whether a judge returns the same verdict when the same request is worded differently remains largely unexamined. We study that question across four evaluation tasks and twenty-five judges from six providers. To support the analysis we release JudgeSense, a benchmark of 880 items from human-labelled corpora, each issued under two instructions that differ in wording and not in what they ask, with the complete decision logs. Every score is reported against the judge's own agreement with itself on the identical prompt, so decoding noise is not charged to wording, and the release lets a reader ask the same of any judge not in our roster. Rewording costs agreement on all four tasks, and on two it clears the threshold we declare for a practically meaningful effect; the ordinal task is both the least stable and the one fewest judges are accurate on, and within a single family parameter count does not predict stability. A judge measured inside an agent harness yields a smaller estimate than the same judge reached through a direct API call, because its agreement with itself collapses faster than its agreement across wordings. --- ### JudgeSense: A Benchmark for Prompt Sensitivity in LLM-as-a-Judge Systems Authors: Rohith Reddy Bellibatlu, Edward Raff, Wenbin Zhang Published: 2026-04-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.23478v2 Large language models are widely adopted as automated evaluation judges, yet the stability of their verdicts under semantically equivalent prompt rephrasings remains largely unexamined. We conduct a systematic empirical study of prompt-induced decision instability across multiple evaluation tasks and judge architectures. To facilitate this analysis, we release JudgeSense, a benchmark comprising hand-validated prompt-paraphrase pairs spanning factuality, coherence, relevance, and preference, drawn from established NLP benchmarks and accompanied by comprehensive decision logs. The benchmark enables the measurement of judge stability across equivalent prompts, allowing researchers to assess whether stability correlates with model scale or instruction-tuning, and to identify which tasks are most sensitive to prompt wording. Our evaluation reveals that coherence remains the primary task for distinguishing judge behavior, while factuality judgments demonstrate high stability under standard conditions. Pairwise evaluation tasks consistently exhibit position bias. Crucially, we find that model scale is not a reliable proxy for consistency; notably, as an interesting result in our analysis, the largest and newest models are not the most consistent. --- ### Supernodes and Halos: Loss-Critical Hubs in LLM Feed-Forward Layers Authors: Audrey Cherilyn, Houman Safaai Published: 2026-04-26 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2604.23475v2 We study the organization of channel-level importance in transformer feed-forward networks (FFNs). Using a Fisher-style loss proxy (LP) based on activation-gradient second moments, we show that loss sensitivity is concentrated in a small set of channels within each layer. In Llama-3.1-8B, the top 1% of channels per layer accounts for a median of 58.7% of LP mass, with a range of 33.0% to 86.1%. We call these loss-critical channels supernodes. Although FFN layers also contain strong activation outliers, LP-defined supernodes overlap only weakly with activation-defined outliers and are not explained by activation power or weight norms alone. Around this core, we find a weaker but consistent halo structure: some non-supernode channels share the supernodes' write support and show stronger redundancy with the protected core. We use one-shot structured FFN pruning as a diagnostic test of this organization. At 50% FFN sparsity, baselines that prune many supernodes degrade sharply, whereas our SCAR variants explicitly protect the supernode core; the strongest variant, SCAR-Prot, reaches perplexity 54.8 compared with 989.2 for Wanda-channel. The LP-concentration pattern appears across Mistral-7B, Llama-2-7B, and Qwen2-7B, remains visible in targeted Llama-3.1-70B experiments, and increases during OLMo-2-7B pretraining. These results suggest that LLM FFNs develop a small learned core of loss-critical channels, and that preserving this core is important for reliable structured pruning. --- ### Benchmarking Storage Systems for Machine Learning Workloads Using NIO Bench Authors: Jonathan W. Morris, Ionut Mistreanu, Connor Louie Published: 2026-04-25 Categories: cs.PF, cs.DC, cs.LG arXiv: https://arxiv.org/abs/2609.05418v1 Machine learning training workloads place unique demands on storage systems, yet most existing benchmarks focus on computational throughput rather than file system I/O behavior. We present a benchmarking framework, Neural I/O Benchmark (NIO Bench), that characterizes storage access patterns across six diverse ML model architectures: Language Transformers, Vision Transformers, Diffusion Models, Spiking Neural Networks, Artificial Neural Networks, and Reinforcement Learning. Our framework employs a two-layer tracing approach combining Python-level I/O hooks for semantic phase context with Linux strace for complete syscall coverage including DataLoader worker subprocesses. We evaluate all six models on a Nautilus Kubernetes cluster with Ceph distributed file system. Our results reveal that I/O is heavily concentrated in data preparation, model loading, and model checkpointing. We also found that training is compute-bound rather than data-bound once data is staged, and that storage access follows an extreme power law where fewer than 10% of files account for over 90% of bytes transferred, and that read tail latency from cache misses on distributed storage is the primary storage bottleneck. These findings suggest that storage systems optimized for ML should prioritize aggressive data prefetching, page cache pinning, and efficient handling of bursty checkpoint writes. --- ### Dynamic Governance of Multi-LLM Agent Systems for Collaborative Conversational Outcomes Authors: Alexander Liss, Nicholas Desmond, Santiago Gil Gallego Published: 2026-04-25 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.11207v1 When two LLM agents with structurally opposed objectives interact across multiple turns, the absence of a shared goal function produces not competition but collapse: the visitor capitulates, the site agent stops varying its approach, and the conversation terminates without achieving either agent's stated objective. This paper asks whether a control-theoretic governance layer can substitute for that missing goal function. The Experience Orchestrator (EO) addresses this in a simulated financial services environment where a site agent guides a visitor toward advisor contact while the visitor maintains psychologically realistic resistance. EO governs the joint trajectory through three mechanisms: a Contextual Bandit (CB) that selects content arms calibrated from real-world web analytics, a PID controller that enforces behavioral consistency via dynamic schema constraints, and a POMDP belief tracker that maintains a probabilistic model of visitor intent. Across 60,000 simulations, EO achieves a +32 percentage point lift in high-intent advisor contact rate (78.1% vs. 46.1% over a naive LLM control), with CB variant selection accounting for 97% of between-factor outcome variance -- confirming that the governance policy, not environmental initial conditions, determines where trajectories end up. Persona-level analysis reveals two distinct regimes: for visitors with no natural inclination toward conversion, the governance layer is the difference between a functional system and a non-functional one; for visitors already near alignment, a naive LLM's empathetic defaults are largely sufficient. All findings are conditional on LLM-to-LLM simulation. The PID controller has not been calibrated against real human unpredictability, and validating EO on live traffic is the critical next step. --- ### Towards an Argumentative Foundation for Evaluative AI Authors: Xiang Yin, Tim Miller, Nico Potyka, Antonio Rago, Francesca Toni Published: 2026-04-25 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2608.07473v1 Evaluative AI (EAI) has been recently proposed as a way to support human decision-making, not by producing a single recommendation, but by presenting competing hypotheses together with evidence for and against each. In this position paper, we advocate (computational) argumentation as a particularly suitable paradigm to provide a formal, computable foundation for forms of EAI that are explainable and contestable, setting the ground for a long-term research agenda towards distributed and human-centred EAI systems. --- ### IdiomX A Multilingual Benchmark for Idiom Understanding, Retrieval, and Interpretation Authors: Ayman Ali Sharara Published: 2026-04-25 Categories: cs.CL, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2606.02584v1 Idiomatic expressions remain a persistent challenge for natural language processing because their meanings are often non-compositional, context-dependent, and difficult to align across languages. Existing idiom resources are often limited in scale, contextual diversity, or multilingual coverage, restricting their utility for modern language models. We introduce IdiomX, a large-scale multilingual benchmark for idiom understanding, retrieval, and interpretation, constructed through a reproducible multi-stage pipeline combining lexical resource extraction, large-scale normalization, controlled large language model enrichment, and structured validation. The resulting dataset contains over 190K contextualized examples spanning 12K+ idioms, with aligned English, Arabic, and French semantic representations, idiomatic and literal usage labels, and rich linguistic metadata. Building on this resource, we define a unified four-task benchmark covering idiom detection, context-to-idiom retrieval, Arabic-to-English idiom retrieval, and idiom interpretation, extending evaluation from figurative recognition to semantic grounding and explainable meaning retrieval. Experiments show that contextual transformer models substantially improve idiom detection, while hybrid retrieval and reranking architectures significantly strengthen both monolingual and cross-lingual idiom retrieval. Results further demonstrate that idiom interpretation can be effectively modeled as a semantic retrieval task, introducing interpretability as a complementary benchmark dimension. Overall, IdiomX provides a scalable benchmark for studying idiomatic language as a progression from detection to retrieval and semantic interpretation, and offers a modular framework extensible to additional languages and figurative reasoning tasks --- ### Fuzzy, Neutrosophic, and Uncertain Graph Theory: Properties and Applications Authors: Takaaki Fujita, Florentin Smarandache Published: 2026-04-25 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.23936v1 This book presents a comprehensive and systematic survey of graph theory under uncertainty, with particular emphasis on the unifying role of the uncertain graph framework. It reviews fundamental concepts, structural properties, graph classes, and graph parameters within fuzzy, neutrosophic, and related models, while also introducing a wide range of extensions such as uncertain digraphs, hypergraphs, superhypergraphs, and dynamic graphs. In addition to theoretical developments, the book explores practical applications, including uncertain molecular graphs, decision-making systems, graph neural networks, knowledge graphs, and cognitive maps. By organizing diverse uncertainty-aware graph models within a common perspective, this work provides a coherent framework for understanding their relationships, capabilities, and applications in complex systems. --- ### Robustness of Refugee-Matching Gains to Off-Policy Evaluation Choices Authors: Kirk Bansak, Elisabeth Paulson, Dominik Rothenhäusler, Jeremy Ferwerda, Jens Hainmueller, Michael Hotard Published: 2026-04-25 Categories: cs.LG, econ.EM, stat.AP, stat.ML arXiv: https://arxiv.org/abs/2605.06686v1 Previous research has investigated the potential of refugee matching for boosting refugee outcomes, first considered by Bansak et al. (2018). This paper demonstrates the stability of counterfactual impact evaluation results in the context of refugee matching in the United States using a range of off-policy evaluation methods. In order to estimate counterfactual impact and test the robustness of our results, we employ several evaluation methods, including inverse probability weighting (IPW) and multiple variants of augmented inverse probability weighting (AIPW). We also consider various modifications, including alternative modeling architectures and different assignment procedures. The impact estimates remain consistent in magnitude in all scenarios as well as statistically significant in most cases. Furthermore, the estimates are also consistent with the results originally presented in Bansak et al. (2018). --- ### From Canopy to Collision: A Hybrid Predictive Framework for Identifying Risk Factors in Tree-Involved Traffic Crashes Authors: Abdul Azim, Ahmed Hossain, Soumyadip Maitra, Panick Kalambay Published: 2026-04-25 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.06684v1 Tree-involved crashes represent a critical subset of run-off-road (ROR) collisions, often resulting in fatal or severe injuries due to high-energy impacts. This study develops a comprehensive analytical framework to identify and quantify risk factors contributing to crash severity in tree-involved collisions using the Crash Report Sampling System (CRSS) database spanning 2020-2023. The modeling framework follows a multi-step process. First, a machine learning based classification model (CatBoost) identifies key factors associated with binary crash injury severity (KA: fatal or incapacitating injury versus BC: non-incapacitating or possible injury). Second, SHapley Additive exPlanations (SHAP) tool is used to quantify and visualize the marginal effects of top influential factors on crash severity. Third, a binary logistic regression model estimates factor effects and validates SHAP-derived importance measures. Finally, SHAP interaction plots examine the combined effects of key contributing factors. Results reveal restraint non-use as the most influential predictor, with unrestrained occupants nearly three times more likely to experience severe outcomes due to ejection risk. Vehicle age, speeding violations, and driver impairment demonstrate substantial effects, reflecting reduced crashworthiness, increased impact forces, and reduced control capabilities. Critical interactions emerge between lighting conditions and vehicle age, speeding and lighting conditions, restraint use and vehicle age, and road surface and speeding, demonstrating additive risk effects with specific interactions. These findings provide critical insights for targeted safe system-based interventions, including enhanced seat belt enforcement, speed management in reduced visibility conditions, and vehicle fleet modernization. --- ### OceanPile: A Large-Scale Multimodal Ocean Corpus for Foundation Models Authors: Yida Xue, Ningyu Zhang, Tingwei Wu, Zhe Ma, Daxiong Ji, Zhao Wang, Guozhou Zheng, Huajun Chen Published: 2026-04-25 Categories: cs.MM, cs.AI, cs.CL, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2605.00877v2 The vast and underexplored ocean plays a critical role in regulating global climate and supporting marine biodiversity, yet artificial intelligence has so far delivered limited impact in this domain due to a fundamental data bottleneck. Specifically, ocean data are highly fragmented across disparate sources and inherently exhibit multi-modal, high-noise, and weakly labeled characteristics, lacking unified schemas and semantic alignment. Although Multimodal Large Language Models (MLLMs) have achieved remarkable success in general domains, their application to ocean science remains severely constrained by the absence of large-scale, well-aligned multimodal datasets tailored to marine environments. To bridge this gap, we introduce OceanPile, a large-scale multimodal corpus designed for ocean foundation models. It comprises three key components: OceanCorpus, a unified collection integrating sonar data, underwater imagery, marine science visuals, and scientific text from diverse authoritative sources; OceanInstruction, a high-quality instruction dataset synthesized via a novel pipeline guided by a hierarchical Ocean Concept Knowledge Graph; and OceanBenchmark, a manually curated evaluation benchmark for rigorous assessment. We establish a multi-stage quality control process to ensure scientific validity and alignment across modalities. Experimental validation demonstrates significant performance improvements for models trained on our data. All datasets are publicly released to advance the field of marine artificial intelligence and empower domain-specific MLLMs. ---