# Research Papers — 2026-03 · bundle page 57 of 243 > Full-text markdown of papers #2801–2850 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/57 [← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/56.md) [Next bundle →](https://aipod.dev/papers/archive/2026-03/page/58.md) ### MemCollab: Cross-Model Memory Collaboration via Contrastive Trajectory Distillation Authors: Yurui Chang, Yiran Wu, Qingyun Wu, Lu Lin Published: 2026-03-24 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.23234v2 LLM agents increasingly rely on memory mechanisms to reuse knowledge from past problem-solving experiences. However, existing methods typically construct memory for a single agent and reuse it with the same underlying model, tightly coupling stored knowledge to model-specific reasoning styles. In heterogeneous deployments, where agents may be instantiated with backbone models of different sizes, architectures, or specializations, this raises a key question: can a single memory system be shared across agents with different backbone models? We find that naive cross-model memory transfer can degrade performance, because stored memories often entangle task-relevant knowledge with model-specific biases. To address this challenge, we propose MemCollab, a collaborative memory framework that builds shared cross-model memory by contrasting reasoning trajectories generated by different model-based agents on the same task. Through this contrastive process, MemCollab distills abstract reasoning constraints that capture shared task-level invariants while suppressing model-specific artifacts. We further introduce a task-aware retrieval mechanism that conditions memory access on task category, ensuring that only relevant constraints are retrieved at inference time. Experiments on mathematical reasoning and code generation benchmarks show that MemCollab consistently improves both accuracy and inference-time efficiency across diverse agents, including settings with different model families. These results demonstrate that collaboratively constructed cross-model memory can serve as a shared reasoning resource for heterogeneous LLM-based agents. --- ### GEM: Guided Expectation-Maximization for Behavior-Normalized Candidate Action Selection in Offline RL Authors: Haoyu Wang, Jingcheng Wang, Shunyu Wu, Xinwei Xiao Published: 2026-03-24 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.23232v1 Offline reinforcement learning (RL) can fit strong value functions from fixed datasets, yet reliable deployment still hinges on the action selection interface used to query them. When the dataset induces a branched or multimodal action landscape, unimodal policy extraction can blur competing hypotheses and yield "in-between" actions that are weakly supported by data, making decisions brittle even with a strong critic. We introduce GEM (Guided Expectation-Maximization), an analytical framework that makes action selection both multimodal and explicitly controllable. GEM trains a Gaussian Mixture Model (GMM) actor via critic-guided, advantage-weighted EM-style updates that preserve distinct components while shifting probability mass toward high-value regions, and learns a tractable GMM behavior model to quantify support. During inference, GEM performs candidate-based selection: it generates a parallel candidate set and reranks actions using a conservative ensemble lower-confidence bound together with behavior-normalized support, where the behavior log-likelihood is standardized within each state's candidate set to yield stable, comparable control across states and candidate budgets. Empirically, GEM is competitive across D4RL benchmarks, and offers a simple inference-time budget knob (candidate count) that trades compute for decision quality without retraining. --- ### PERMA: Benchmarking Personalized Memory Agents via Event-Driven Preference and Realistic Task Environments Authors: Shuochen Liu, Junyi Zhu, Long Shu, Junda Lin, Yuhao Chen, Haotian Zhang, Chao Zhang, Derong Xu, Jia Li, Bo Tang, Zhiyu Li, Feiyu Xiong et al. Published: 2026-03-24 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.23231v2 Empowering large language models with long-term memory is crucial for building agents that adapt to users' evolving needs. Existing evaluations of this capability typically interleave preference-related dialogues with irrelevant conversations, reducing the task to needle-in-a-haystack retrieval while ignoring relationships between events driving user preference evolution. Such settings overlook a fundamental characteristic of real-world personalization: preferences emerge gradually and accumulate across interactions within noisy contexts. To bridge this gap, we introduce PERMA, a benchmark designed to evaluate persona consistency over time beyond static preference recall. Additionally, we incorporate (1) text variability and (2) linguistic alignment to simulate erratic user inputs and individual idiolects in real-world data. PERMA consists of temporally ordered interaction events spanning multiple sessions and domains, with preference-related queries inserted over time. We design both multiple-choice and interactive tasks to probe the model's understanding of persona along the interaction timeline. Experiments demonstrate that by linking related interactions, advanced memory systems extract precise preferences and reduce token consumption, outperforming traditional semantic retrieval of raw dialogues. Nevertheless, they still struggle to maintain a coherent persona across temporal depth and cross-domain interference, highlighting the need for more robust personalized memory management in agents. Our code and data are open-sourced at https://github.com/PolarisLiu1/PERMA. --- ### I Came, I Saw, I Explained: Benchmarking Multimodal LLMs on Figurative Meaning in Memes Authors: Shijia Zhou, Saif M. Mohammad, Barbara Plank, Diego Frassinelli Published: 2026-03-24 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.23229v1 Internet memes represent a popular form of multimodal online communication and often use figurative elements to convey layered meaning through the combination of text and images. However, it remains largely unclear how multimodal large language models (MLLMs) combine and interpret visual and textual information to identify figurative meaning in memes. To address this gap, we evaluate eight state-of-the-art generative MLLMs across three datasets on their ability to detect and explain six types of figurative meaning. In addition, we conduct a human evaluation of the explanations generated by these MLLMs, assessing whether the provided reasoning supports the predicted label and whether it remains faithful to the original meme content. Our findings indicate that all models exhibit a strong bias to associate a meme with figurative meaning, even when no such meaning is present. Qualitative analysis further shows that correct predictions are not always accompanied by faithful explanations. --- ### General Machine Learning: Theory for Learning Under Variable Regimes Authors: Aomar Osmani Published: 2026-03-24 Categories: cs.LG, cs.AI, stat.ML arXiv: https://arxiv.org/abs/2603.23220v1 We study learning under regime variation, where the learner, its memory state, and the evaluative conditions may evolve over time. This paper is a foundational and structural contribution: its goal is to define the core learning-theoretic objects required for such settings and to establish their first theorem-supporting consequences. The paper develops a regime-varying framework centered on admissible transport, protected-core preservation, and evaluator-aware learning evolution. It records the immediate closure consequences of admissibility, develops a structural obstruction argument for faithful fixed-ontology reduction in genuinely multi-regime settings, and introduces a protected-stability template together with explicit numerical and symbolic witnesses on controlled subclasses, including convex and deductive settings. It also establishes theorem-layer results on evaluator factorization, morphisms, composition, and partial kernel-level alignment across semantically commensurable layers. A worked two-regime example makes the admissibility certificate, protected evaluative core, and regime-variation cost explicit on a controlled subclass. The symbolic component is deliberately restricted in scope: the paper establishes a first kernel-level compatibility result together with a controlled monotonic deductive witness. The manuscript should therefore be read as introducing a structured learning-theoretic framework for regime-varying learning together with its first theorem-supporting layer, not as a complete quantitative theory of all learning systems. --- ### Decoding AI Authorship: Can LLMs Truly Mimic Human Style Across Literature and Politics? Authors: Nasser A Alsadhan Published: 2026-03-24 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2603.23219v1 Amidst the rising capabilities of generative AI to mimic specific human styles, this study investigates the ability of state-of-the-art large language models (LLMs), including GPT-4o, Gemini 1.5 Pro, and Claude Sonnet 3.5, to emulate the authorial signatures of prominent literary and political figures: Walt Whitman, William Wordsworth, Donald Trump, and Barack Obama. Utilizing a zero-shot prompting framework with strict thematic alignment, we generated synthetic corpora evaluated through a complementary framework combining transformer-based classification (BERT) and interpretable machine learning (XGBoost). Our methodology integrates Linguistic Inquiry and Word Count (LIWC) markers, perplexity, and readability indices to assess the divergence between AI-generated and human-authored text. Results demonstrate that AI-generated mimicry remains highly detectable, with XGBoost models trained on a restricted set of eight stylometric features achieving accuracy comparable to high-dimensional neural classifiers. Feature importance analyses identify perplexity as the primary discriminative metric, revealing a significant divergence in the stochastic regularity of AI outputs compared to the higher variability of human writing. While LLMs exhibit distributional convergence with human authors on low-dimensional heuristic features, such as syntactic complexity and readability, they do not yet fully replicate the nuanced affective density and stylistic variance inherent in the human-authored corpus. By isolating the specific statistical gaps in current generative mimicry, this study provides a comprehensive benchmark for LLM stylistic behavior and offers critical insights for authorship attribution in the digital humanities and social media. --- ### PoseDriver: A Unified Approach to Multi-Category Skeleton Detection for Autonomous Driving Authors: Yasamin Borhani, Taylor Mordan, Yihan Wang, Reyhaneh Hosseininejad, Javad Khoramdel, Alexandre Alahi Published: 2026-03-24 Categories: cs.CV, cs.HC arXiv: https://arxiv.org/abs/2603.23215v2 Object skeletons offer a concise representation of structural information, capturing essential aspects of posture and orientation that are crucial for autonomous driving applications. However, a unified architecture that simultaneously handles multiple instances and categories using only the input image remains elusive. In this paper, we introduce PoseDriver, a unified framework for bottom-up multi-category skeleton detection tailored to common objects in driving scenarios. We model each category as a distinct task to systematically address the challenges of multi-task learning. Specifically, we propose a novel approach for lane detection based on skeleton representations, achieving state-of-the-art performance on the OpenLane dataset. Moreover, we present a new dataset for bicycle skeleton detection and assess the transferability of our framework to novel categories. Experimental results validate the effectiveness of the proposed approach. --- ### Generative Inversion of Spectroscopic Data for Amorphous Structure Elucidation Authors: Jiawei Guo, Daniel Schwalbe-Koda Published: 2026-03-24 Categories: cond-mat.dis-nn, cond-mat.mtrl-sci, cs.LG arXiv: https://arxiv.org/abs/2603.23210v1 Determining atomistic structures from characterization data is one of the most common yet intricate problems in materials science. Particularly in amorphous materials, proposing structures that balance realism and agreement with experiments requires expert guidance, good interatomic potentials, or both. Here, we introduce GLASS, a generative framework that inverts multi-modal spectroscopic measurements into realistic atomistic structures without knowledge of the potential energy surface. A score-based model learns a structural prior from low-fidelity data and samples out-of-distribution structures conditioned on differentiable spectral targets. Reconstructions using pair distribution functions (PDFs), X-ray absorption spectroscopy, and diffraction measurements quantify the complementarity between spectral modalities and demonstrate that PDFs is the most informative probe for our framework. We use GLASS to rationalize three contested experimental problems: paracrystallinity in amorphous silicon, a liquid-liquid phase transition in sulfur, and ball-milled amorphous ice. In each case, generated structures reproduce experimental measurements and reveal mechanisms inaccessible to diffraction analysis alone. --- ### A One-Inclusion Graph Approach to Multi-Group Learning Authors: Noah Bergam, Samuel Deng, Daniel Hsu Published: 2026-03-24 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.23208v2 We prove the tightest-known upper bounds on the sample complexity of multi-group learning. Our algorithm extends the one-inclusion graph prediction strategy using a generalization of bipartite $b$-matching. In the group-realizable setting, we provide a lower bound confirming that our algorithm's $\log n / n$ convergence rate is optimal in general. If one relaxes the learning objective such that the group on which we are evaluated is chosen obliviously of the sample, then our algorithm achieves the optimal $1/n$ convergence rate under group-realizability. --- ### Latency Coding for Efficient and Low-Latency Deep Spiking Neural Networks Authors: Yi Lu, Jianhao Ding, Zhaofei Yu Published: 2026-03-24 Categories: cs.NE arXiv: https://arxiv.org/abs/2603.23206v2 Spiking neural networks (SNNs) offer a biologically inspired computing paradigm with significant potential for energy-efficient neural processing. Among neural coding schemes of SNNs, Time-To-First-Spike (TTFS) coding, which encodes information through the precise timing of a neuron's first spike, provides exceptional activity sparsity and energy efficiency. However, existing TTFS models lack efficient training methods, suffering from high inference latency and limited performance, limiting their practicality on neuromorphic hardware. In this work, we propose latency coding, an extension of TTFS coding, and present a compatible framework that enables the efficient training of deep latency-coded SNNs by leveraging backpropagation through time (BPTT) algorithm. The framework includes: (1) a latency encoding (LE) module with feature extraction and straight-through estimators to address severe information loss in direct intensity-to-latency mapping; (2) relaxation of the strict single-spike constraint in intermediate layers to improve information propagation and gradient flow; and (3) a temporal adaptive decision (TAD) loss function that dynamically weights supervision signals based on the model's confidence, balancing the trade-off between speed and accuracy. Experimental results demonstrate that our method achieves competitive or superior accuracy compared with existing TTFS-coded SNNs with ultra-low inference latency and high energy efficiency. Latency-coded SNNs also demonstrate improved robustness against input perturbations. These findings highlight latency coding as a practical and hardware-friendly approach for fast and energy-efficient neuromorphic processing. --- ### Between Resolution Collapse and Variance Inflation: Weighted Conformal Anomaly Detection in Low-Data Regimes Authors: Oliver Hennhöfer, Christine Preisach Published: 2026-03-24 Categories: stat.ML, cs.LG, stat.ME arXiv: https://arxiv.org/abs/2603.23205v1 Standard conformal anomaly detection provides marginal finite-sample guarantees under the assumption of exchangeability . However, real-world data often exhibit distribution shifts, necessitating a weighted conformal approach to adapt to local non-stationarity. We show that this adaptation induces a critical trade-off between the minimum attainable p-value and its stability. As importance weights localize to relevant calibration instances, the effective sample size decreases. This can render standard conformal p-values overly conservative for effective error control, while the smoothing technique used to mitigate this issue introduces conditional variance, potentially masking anomalies. We propose a continuous inference relaxation that resolves this dilemma by decoupling local adaptation from tail resolution via continuous weighted kernel density estimation. While relaxing finite-sample exactness to asymptotic validity, our method eliminates Monte Carlo variability and recovers the statistical power lost to discretization. Empirical evaluations confirm that our approach not only restores detection capabilities where discrete baselines yield zero discoveries, but outperforms standard methods in statistical power while maintaining valid marginal error control in practice. --- ### Gaze-Regularized Vision-Language-Action Models for Robotic Manipulation Authors: Anupam Pani, Yanchao Yang Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23202v2 Despite advances in Vision-Language-Action (VLA) models, robotic manipulation struggles with fine-grained tasks because current models lack mechanisms for active visual attention allocation. Human gaze naturally encodes intent, planning, and execution patterns -- offering a powerful supervisory signal for guiding robot perception. We introduce a gaze-regularized training framework that aligns VLA models' internal attention with human visual patterns without architectural modifications or inference-time overhead. Our method transforms temporally aggregated gaze heatmaps into patch-level distributions and regularizes the transformer's attention through KL divergence, creating an inductive bias toward task-relevant features while preserving deployment efficiency. When integrated into existing VLA architectures, our approach yields 4-12% improvements across manipulation benchmarks. The gaze-regularized models reach equivalent performance with fewer training steps and maintain robustness under lighting variations and sensor noise. Beyond performance metrics, the learned attention patterns produce interpretable visualizations that mirror human strategies, enhancing trust in robotic systems. Moreover, our framework requires no eye-tracking equipment and applies directly to existing datasets. These results demonstrate that human perceptual priors can significantly accelerate robot learning while improving both task performance and system interpretability. --- ### FDIF: Formula-Driven supervised Learning with Implicit Functions for 3D Medical Image Segmentation Authors: Yukinori Yamamoto, Kazuya Nishimura, Tsukasa Fukusato, Hirokazu Nosato, Tetsuya Ogata, Hirokatsu Kataoka Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23199v2 Deep learning-based 3D medical image segmentation methods relies on large-scale labeled datasets, yet acquiring such data is difficult due to privacy constraints and the high cost of expert annotation. Formula-Driven Supervised Learning (FDSL) offers an appealing alternative by generating training data and labels directly from mathematical formulas. However, existing voxel-based approaches are limited in geometric expressiveness and cannot synthesize realistic textures. We introduce Formula-Driven supervised learning with Implicit Functions (FDIF), a framework that enables scalable pre-training without using any real data and medical expert annotations. FDIF introduces an implicit-function representation based on signed distance functions (SDFs), enabling compact modeling of complex geometries while exploiting the surface representation of SDFs to support controllable synthesis of both geometric and intensity textures. Across three medical image segmentation benchmarks (AMOS, ACDC, and KiTS) and three architectures (SwinUNETR, nnUNet ResEnc-L, and nnUNet Primus-M), FDIF consistently improves over a formula-driven method, and achieves performance comparable to self-supervised approaches pre-trained on large-scale real datasets. We further show that FDIF pre-training also benefits 3D classification tasks, highlighting implicit-function-based formula supervision as a promising paradigm for data-free representation learning. Code is available at https://github.com/yamanoko/FDIF. --- ### Sparser, Faster, Lighter Transformer Language Models Authors: Edoardo Cetin, Stefano Peluchetti, Emilio Castillo, Akira Naruse, Mana Murakami, Llion Jones Published: 2026-03-24 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2603.23198v2 Scaling autoregressive large language models (LLMs) has driven unprecedented progress but comes with vast computational costs. In this work, we tackle these costs by leveraging unstructured sparsity within an LLM's feedforward layers, the components accounting for most of the model parameters and execution FLOPs. To achieve this, we introduce a new sparse packing format and a set of CUDA kernels designed to seamlessly integrate with the optimized execution pipelines of modern GPUs, enabling efficient sparse computation during LLM inference and training. To substantiate our gains, we provide a quantitative study of LLM sparsity, demonstrating that simple L1 regularization can induce over 99% sparsity with negligible impact on downstream performance. When paired with our kernels, we show that these sparsity levels translate into substantial throughput, energy efficiency, and memory usage benefits that increase with model scale. We will release all code and kernels under an open-source license to promote adoption and accelerate research toward establishing sparsity as a practical axis for improving the efficiency and scalability of modern foundation models. --- ### Gaussian mixtures and non-parametric likelihoods through the lens of statistical mechanics Authors: Subhroshekhar Ghosh, Adityanand Guntuboyina, Satyaki Mukherjee, Hoang-Son Tran Published: 2026-03-24 Categories: math.ST, cond-mat.dis-nn, cond-mat.stat-mech, math.PR, stat.ML arXiv: https://arxiv.org/abs/2603.23196v2 In this work, we investigate Gaussian Mixture Models ({\it abbrv} GMM) and the related problem of non parametric maximum likelihood estimation ({\it abbrv} NPMLE) from the perspective of statistical mechanics. In particular, we establish stability guarantees for the NPMLE procedure that extend well beyond the state of the art. Crucially, we obtain guarantees on the Kullback-Leibler divergence between NPMLE estimators and the ground truth, a type of result which has been known to be challenging in the literature on this problem. In particular, we provide high probability upper bounds on the KL divergence between the NPMLE and the true density that are of the order of $\min\big\{\frac{(\log n)^{d+2}}{n} , \frac{\log n}{\sqrt n}\big\}$, which cover a wide range of scenarios for the comparative sizes of $n$ and $d$. We obtain similar guarantees for approximate solutions to the NPMLE problem, addressing realistic situations wherein optimization algorithms need to be stopped in finite time, allowing access only to approximations to the true NPMLE. A cornerstone of our approach is an analysis of the function class complexity of logarithms of gaussian mixture densities, which is able to handle their unboundedness, and could be of wider interest. Our methods lead to novel confidence-interval guarantees for entropy estimation in Gaussian mixtures, demonstrating their wider impact. We also establish correspondences between stability phenomena in the NPMLE problem and concepts such as chaos and multiple valleys in random energy landscapes of statistical mechanics models. While these correspondences are largely of a conceptual nature at this point, we believe that these connections, especially those with concentration phenomena and Langevin dynamics, may be developed into a toolbox for studying a wide variety of random optimization problems in statistics and machine learning. --- ### PhysSkin: Real-Time and Generalizable Physics-Based Animation via Self-Supervised Neural Skinning Authors: Yuanhang Lei, Tao Cheng, Xingxuan Li, Boming Zhao, Siyuan Huang, Ruizhen Hu, Peter Yichen Chen, Hujun Bao, Zhaopeng Cui Published: 2026-03-24 Categories: cs.GR, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2603.23194v2 Achieving real-time physics-based animation that generalizes across diverse 3D shapes and discretizations remains a fundamental challenge. We introduce PhysSkin, a physics-informed framework that addresses this challenge. In the spirit of Linear Blend Skinning, we learn continuous skinning fields as basis functions lifting motion subspace coordinates to full-space deformation, with subspace defined by handle transformations. To generate mesh-free, discretization-agnostic, and physically consistent skinning fields that generalize well across diverse 3D shapes, PhysSkin employs a new neural skinning fields autoencoder which consists of a transformer-based encoder and a cross-attention decoder. Furthermore, we also develop a novel physics-informed self-supervised learning strategy that incorporates on-the-fly skinning-field normalization and conflict-aware gradient correction, enabling effective balancing of energy minimization, spatial smoothness, and orthogonality constraints. PhysSkin shows outstanding performance on generalizable neural skinning and enables real-time physics-based animation. --- ### Gaze-Regularized VLMs for Ego-Centric Behavior Understanding Authors: Anupam Pani, Yanchao Yang Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23190v1 Eye gaze, encompassing fixations and saccades, provides critical insights into human intentions and future actions. This study introduces a gaze-regularized framework that enhances Vision Language Models (VLMs) for egocentric behavior understanding. Unlike existing methods that rely solely on visual data and overlook gaze information, our approach directly incorporates gaze information into the VLM architecture during training. By generating gaze-based queries, the model dynamically focuses on gaze-highlighted regions, while a gaze-regularization mechanism ensures the alignment of model attention with human attention patterns. To better understand how gaze can be effectively integrated into VLMs, we conducted extensive experiments exploring various strategies for incorporating gaze data. These innovations enable the prediction of future events with detailed action descriptions. Experimental results demonstrate a nearly 13 % improvement in semantic scores compared to baseline models not leveraging gaze data, highlighting the effectiveness of our approach. This work establishes a foundation for leveraging the human gaze in VLMs, significantly boosting their predictive capabilities in applications requiring accurate and robust future event prediction. --- ### ViKey: Enhancing Temporal Understanding in Videos via Visual Prompting Authors: Yeonkyung Lee, Dayun Ju, Youngmin Kim, Seil Kang, Seong Jae Hwang Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23186v1 Recent advancements in Video Large Language Models (VideoLLMs) have enabled strong performance across diverse multimodal video tasks. To reduce the high computational cost of processing dense video frames, efficiency-oriented methods such as frame selection have been widely adopted. While effective at minimizing redundancy, these methods often cause notable performance drops on tasks requiring temporal reasoning. Unlike humans, who can infer event progression from sparse visual cues, VideoLLMs frequently misinterpret temporal relations when intermediate frames are omitted. To address this limitation, we explore visual prompting (VP) as a lightweight yet effective way to enhance temporal understanding in VideoLLMs. Our analysis reveals that simply annotating each frame with explicit ordinal information helps the model perceive temporal continuity. This visual cue also supports frame-level referencing and mitigates positional ambiguity within a sparsely sampled sequence. Building on these insights, we introduce ViKey, a training-free framework that combines VP with a lightweight Keyword-Frame Mapping (KFM) module. KFM leverages frame indices as dictionary-like keys to link textual cues to the most relevant frames, providing explicit temporal anchors during inference. Despite its simplicity, our approach substantially improves temporal reasoning and, on some datasets, preserves dense-frame baseline performance with as few as 20% of frames. --- ### ImplicitRM: Unbiased Reward Modeling from Implicit Preference Data for LLM alignment Authors: Hao Wang, Haocheng Yang, Licheng Pan, Lei Shen, Xiaoxi Li, Yinuo Wang, Zhichao Chen, Yuan Lu, Haoxuan Li, Zhouchen Lin Published: 2026-03-24 Categories: cs.CL, cs.AI, stat.AP arXiv: https://arxiv.org/abs/2603.23184v1 Reward modeling represents a long-standing challenge in reinforcement learning from human feedback (RLHF) for aligning language models. Current reward modeling is heavily contingent upon experimental feedback data with high collection costs. In this work, we study \textit{implicit reward modeling} -- learning reward models from implicit human feedback (e.g., clicks and copies) -- as a cost-effective alternative. We identify two fundamental challenges in implicit reward modeling: (1) Implicit preference data lacks definitive negative samples, which makes standard positive-negative classification methods inapplicable; (2) Implicit preference data suffers from user preference bias, where different responses have different propensities to elicit user feedback actions, which exacerbates the difficulty of distinguishing definitive negative samples. To address these challenges, we propose ImplicitRM, which aims to learn unbiased reward models from implicit preference data. ImplicitRM stratifies training samples into four latent groups via a stratification model. Building on this, it derives a learning objective through likelihood maximization, which we prove is theoretically unbiased, effectively resolving both challenges. Experiments demonstrate that ImplicitRM learns accurate reward models across implicit preference datasets. Code is available on our project website. --- ### Reasoning over Semantic IDs Enhances Generative Recommendation Authors: Yingzhi He, Yan Sun, Junfei Tan, Yuxin Chen, Xiaoyu Kong, Chunxu Shen, Xiang Wang, An Zhang, Tat-Seng Chua Published: 2026-03-24 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2603.23183v2 Recent advances in generative recommendation have leveraged pretrained LLMs by formulating sequential recommendation as autoregressive generation over a unified token space comprising language tokens and itemic identifiers, where each item is represented by a compact sequence of discrete tokens, namely Semantic IDs (SIDs). This SID-based formulation enables efficient decoding over large-scale item corpora and provides a natural interface for LLM-based recommenders to leverage rich world knowledge. Meanwhile, breakthroughs in LLM reasoning motivate reasoning-enhanced recommendation, yet effective reasoning over SIDs remains underexplored and challenging. Itemic tokens are not natively meaningful to LLMs; moreover, recommendation-oriented SID reasoning is hard to evaluate, making high-quality supervision scarce. To address these challenges, we propose SIDReasoner, a two-stage framework that elicits reasoning over SIDs by strengthening SID--language alignment to unlock transferable LLM reasoning, rather than relying on large amounts of recommendation-specific reasoning traces. Concretely, SIDReasoner first enhances SID-language alignment via multi-task training on an enriched SID-centered corpus synthesized by a stronger teacher model, grounding itemic tokens in diverse semantic and behavioral contexts. Building on this enhanced alignment, SIDReasoner further improves recommendation reasoning through outcome-driven reinforced optimization, which guides the model toward effective reasoning trajectories without requiring explicit reasoning annotations. Extensive experiments on three real-world datasets demonstrate the effectiveness of our reasoning-augmented SID-based generative recommendation. Beyond accuracy, the results highlight the broader potential of large reasoning models for generative recommendation, including improved interpretability and cross-domain generalization. --- ### Gimbal360: Canonicalizing Planar Diffusion for Spherical Panorama Completion Authors: Yuqin Lu, Haofeng Liu, Yang Zhou, Yihua Dai, Guiqing Li, Shengfeng He, Jun Liang Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23179v2 Diffusion models provide powerful priors for 2D image completion, but these priors are learned on bounded planar images and do not transfer directly to $360^\circ$ panoramas. Perspective observations and spherical panoramas differ in both projective geometry and topology: viewpoint-dependent distortion complicates spatial correspondence, while Equirectangular Projection (ERP) panoramas exhibit intrinsic $S^1$ periodicity that standard Euclidean architectures do not preserve. We present Gimbal360, a unified framework that adapts planar diffusion priors to spherical panoramic completion by standardizing these geometric and topological structures. Our Canonical Viewing Space expresses projective distortion as a fixed function of latitude, providing a consistent interface between perspective inputs and spherical panoramas. To map unposed in-the-wild images into this space, Differentiable Projective Canonicalization projects a dense correspondence field onto a 3-DoF rigid projection manifold without requiring camera parameters at inference. We further introduce Topologically Equivariant Generation, which enforces latent shift equivariance to preserve continuity across the periodic ERP boundary. Together, these designs allow diffusion to operate in a representation whose geometry and topology are explicitly aligned with the spherical domain. We also introduce Horizon360, a curated large-scale dataset of gravity-aligned panoramic environments. Extensive experiments show that Gimbal360 achieves state-of-the-art visual fidelity and seam continuity in $360^\circ$ scene completion. --- ### SAiW: Source-Attributable Invisible Watermarking for Proactive Deepfake Defense Authors: Bibek Das, Chandranath Adak, Soumi Chattopadhyay, Zahid Akhtar, Soumya Dutta Published: 2026-03-24 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.23178v1 Deepfakes generated by modern generative models pose a serious threat to information integrity, digital identity, and public trust. Existing detection methods are largely reactive, attempting to identify manipulations after they occur and often failing to generalize across evolving generation techniques. This motivates the need for proactive mechanisms that secure media authenticity at the time of creation. In this work, we introduce SAiW, a Source-Attributed Invisible watermarking Framework for proactive deepfake defense and media provenance verification. Unlike conventional watermarking methods that treat watermark payloads as generic signals, SAiW formulates watermark embedding as a source-conditioned representation learning problem, where watermark identity encodes the originating source and modulates the embedding process to produce discriminative and traceable signatures. The framework integrates feature-wise linear modulation to inject source identity into the embedding network, enabling scalable multi-source watermark generation. A perceptual guidance module derived from human visual system priors ensures that watermark perturbations remain visually imperceptible while maintaining robustness. In addition, a dual-purpose forensic decoder simultaneously reconstructs the embedded watermark and performs source attribution, providing both automated verification and interpretable forensic evidence. Extensive experiments across multiple deepfake datasets demonstrate that SAiW achieves high perceptual quality while maintaining strong robustness against compression, filtering, noise, geometric transformations, and adversarial perturbations. By binding digital media to its origin through invisible yet verifiable markers, SAiW enables reliable authentication and source attribution, providing a scalable foundation for proactive deepfake defense and trustworthy media provenance. --- ### A Schrödinger Eigenfunction Method for Long-Horizon Stochastic Optimal Control Authors: Louis Claeys, Artur Goldman, Zebang Shen, Niao He Published: 2026-03-24 Categories: cs.LG, math.OC arXiv: https://arxiv.org/abs/2603.23173v1 High-dimensional stochastic optimal control (SOC) becomes harder with longer planning horizons: existing methods scale linearly in the horizon $T$, with performance often deteriorating exponentially. We overcome these limitations for a subclass of linearly-solvable SOC problems-those whose uncontrolled drift is the gradient of a potential. In this setting, the Hamilton-Jacobi-Bellman equation reduces to a linear PDE governed by an operator $\mathcal{L}$. We prove that, under the gradient drift assumption, $\mathcal{L}$ is unitarily equivalent to a Schrödinger operator $\mathcal{S} = -Δ+ \mathcal{V}$ with purely discrete spectrum, allowing the long-horizon control to be efficiently described via the eigensystem of $\mathcal{L}$. This connection provides two key results: first, for a symmetric linear-quadratic regulator (LQR), $\mathcal{S}$ matches the Hamiltonian of a quantum harmonic oscillator, whose closed-form eigensystem yields an analytic solution to the symmetric LQR with \emph{arbitrary} terminal cost. Second, in a more general setting, we learn the eigensystem of $\mathcal{L}$ using neural networks. We identify implicit reweighting issues with existing eigenfunction learning losses that degrade performance in control tasks, and propose a novel loss function to mitigate this. We evaluate our method on several long-horizon benchmarks, achieving an order-of-magnitude improvement in control accuracy compared to state-of-the-art methods, while reducing memory usage and runtime complexity from $\mathcal{O}(Td)$ to $\mathcal{O}(d)$. --- ### From Synthetic to Native: Benchmarking Multilingual Intent Classification in Logistics Customer Service Authors: Haoyu He, Jinyu Zhuang, Haoran Chu, Shuhang Yu, J, T AI Group, Hao Wang, Kunpeng Han Published: 2026-03-24 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.23172v1 Multilingual intent classification is central to customer-service systems on global logistics platforms, where models must process noisy user queries across languages and hierarchical label spaces. Yet most existing multilingual benchmarks rely on machine-translated text, which is typically cleaner and more standardized than native customer requests and can therefore overestimate real-world robustness. We present a public benchmark for hierarchical multilingual intent classification constructed from real logistics customer-service logs. The dataset contains approximately 30K de-identified, stand-alone user queries curated from 600K historical records through filtering, LLM-assisted quality control, and human verification, and is organized into a two-level taxonomy with 13 parent and 17 leaf intents. English, Spanish, and Arabic are included as seen languages, while Indonesian, Chinese, and additional test-only languages support zero-shot evaluation. To directly measure the gap between synthetic and real evaluation, we provide paired native and machine-translated test sets and benchmark multilingual encoders, embedding models, and small language models under flat and hierarchical protocols. Results show that translated test sets substantially overestimate performance on noisy native queries, especially for long-tail intents and cross-lingual transfer, underscoring the need for more realistic multilingual intent benchmarks. --- ### Adaptively Robust LLM Monitoring via Activation Watermarking Authors: Toluwani Aremu, Daniil Ognev, Samuele Poppi, Nils Lukas Published: 2026-03-24 Categories: cs.CR, cs.AI, cs.CY, cs.LG arXiv: https://arxiv.org/abs/2603.23171v3 Providers monitor deployed large language models (LLMs) to detect misuse that they cannot prevent. LLM monitoring is deterministic and often openly available, so $\emph{adaptive}$ attackers with a local copy can search offline for prompts that elicit harmful behavior and evade detection. These attacks are especially concerning because providers never observe the misuse and cannot patch their defenses post-hoc. The core challenge is resisting adaptive attackers while preserving detection rates against non-adaptive ones. We propose $\emph{Activation Watermarking}$ (AWM), which randomizes monitoring through limited fine-tuning that aligns the LLM's hidden states with a secret key-derived direction whenever a response violates a policy. Detection is a similarity test on activations the provider already computes, and attackers who know everything but the key must optimize against differently keyed surrogate detectors. At a matched $1\%$ false-positive rate, such surrogate-based attacks evade every evaluated baseline at least $79\%$ of the time, but AWM more than halves this evasion rate. AWM also achieves the lowest evasion rate on three of four non-adaptive jailbreak families at a small drop in utility measured across seven benchmarks. Because random high-dimensional directions are near-orthogonal, assigning one watermark per policy enables attribution. Across 20 monitored policies, AWM identifies the violated policy with $80\%$ accuracy versus a $5\%$ chance baseline. AWM is (i) efficient, (ii) substantially more robust against adaptive attackers than related work and (iii) can attribute which policies were violated. --- ### GSwap: Realistic Head Swapping with Dynamic Neural Gaussian Field Authors: Jingtao Zhou, Xuan Gao, Dongyu Liu, Junhui Hou, Yudong Guo, Juyong Zhang Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23168v1 We present GSwap, a novel consistent and realistic video head-swapping system empowered by dynamic neural Gaussian portrait priors, which significantly advances the state of the art in face and head replacement. Unlike previous methods that rely primarily on 2D generative models or 3D Morphable Face Models (3DMM), our approach overcomes their inherent limitations, including poor 3D consistency, unnatural facial expressions, and restricted synthesis quality. Moreover, existing techniques struggle with full head-swapping tasks due to insufficient holistic head modeling and ineffective background blending, often resulting in visible artifacts and misalignments. To address these challenges, GSwap introduces an intrinsic 3D Gaussian feature field embedded within a full-body SMPL-X surface, effectively elevating 2D portrait videos into a dynamic neural Gaussian field. This innovation ensures high-fidelity, 3D-consistent portrait rendering while preserving natural head-torso relationships and seamless motion dynamics. To facilitate training, we adapt a pretrained 2D portrait generative model to the source head domain using only a few reference images, enabling efficient domain adaptation. Furthermore, we propose a neural re-rendering strategy that harmoniously integrates the synthesized foreground with the original background, eliminating blending artifacts and enhancing realism. Extensive experiments demonstrate that GSwap surpasses existing methods in multiple aspects, including visual quality, temporal coherence, identity preservation, and 3D consistency. --- ### Dual Contrastive Network for Few-Shot Remote Sensing Image Scene Classification Authors: Zhong Ji, Liyuan Hou, Xuan Wang, Gang Wang, Yanwei Pang Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23161v1 Few-shot remote sensing image scene classification (FS-RSISC) aims at classifying remote sensing images with only a few labeled samples. The main challenges lie in small inter-class variances and large intra-class variances, which are the inherent property of remote sensing images. To address these challenges, we propose a transfer-based Dual Contrastive Network (DCN), which incorporates two auxiliary supervised contrastive learning branches during the training process. Specifically, one is a Context-guided Contrastive Learning (CCL) branch and the other is a Detail-guided Contrastive Learning (DCL) branch, which focus on inter-class discriminability and intra-class invariance, respectively. In the CCL branch, we first devise a Condenser Network to capture context features, and then leverage a supervised contrastive learning on top of the obtained context features to facilitate the model to learn more discriminative features. In the DCL branch, a Smelter Network is designed to highlight the significant local detail information. And then we construct a supervised contrastive learning based on the detail feature maps to fully exploit the spatial information in each map, enabling the model to concentrate on invariant detail features. Extensive experiments on four public benchmark remote sensing datasets demonstrate the competitive performance of our proposed DCN. --- ### UniDial-EvalKit: A Unified Toolkit for Evaluating Multi-Faceted Conversational Abilities Authors: Qi Jia, Haodong Zhao, Dun Pei, Xiujie Song, Ye Shen, Shibo Wang, Zijian Chen, Zicheng Zhang, Xiangyang Zhu, Guangtao Zhai Published: 2026-03-24 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.23160v2 Benchmarking large language models (LLMs) and agents in multi-turn interactive scenarios is essential for understanding their practical capabilities. However, existing evaluation protocols are highly heterogeneous, differing significantly in dataset formats, model interfaces, and evaluation pipelines, which severely impedes systematic comparison. In this work, we present UniDial-EvalKit (UDE), a unified evaluation toolkit for assessing interactive AI systems. The core contribution of UDE lies in its holistic unification: it standardizes heterogeneous data formats into a universal schema, streamlines complex evaluation pipelines through a modular architecture, and aligns metric calculations under a hierarchical scoring aggregation. It also supports efficient large-scale evaluation through parallel generation and scoring, as well as checkpoint resume to eliminate redundant computation. Leveraging UDE, we conduct an extensive evaluation across diverse multi-dimensional benchmarks. Our empirical analysis shows that no single system consistently outperforms others across all benchmarks, while current memory agents often fail to surpass full-context baselines. Further analyses highlight several future directions, including benchmark deduplication and more adaptive memory architectures. --- ### Conformal Cross-Modal Active Learning Authors: Huy Hoang Nguyen, Cédric Jung, Shirin Salehi, Tobias Glück, Anke Schmeink, Andreas Kugi Published: 2026-03-24 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2603.23159v1 Foundation models for vision have transformed visual recognition with powerful pretrained representations and strong zero-shot capabilities, yet their potential for data-efficient learning remains largely untapped. Active Learning (AL) aims to minimize annotation costs by strategically selecting the most informative samples for labeling, but existing methods largely overlook the rich multimodal knowledge embedded in modern vision-language models (VLMs). We introduce Conformal Cross-Modal Acquisition (CCMA), a novel AL framework that bridges vision and language modalities through a teacher-student architecture. CCMA employs a pretrained VLM as a teacher to provide semantically grounded uncertainty estimates, conformally calibrated to guide sample selection for a vision-only student model. By integrating multimodal conformal scoring with diversity-aware selection strategies, CCMA achieves superior data efficiency across multiple benchmarks. Our approach consistently outperforms state-of-the-art AL baselines, demonstrating clear advantages over methods relying solely on uncertainty or diversity metrics. --- ### VoDaSuRe: A Large-Scale Dataset Revealing Domain Shift in Volumetric Super-Resolution Authors: August Leander Høeg, Sophia Wiinberg Bardenfleth, Hans Martin Kjer, Tim Bjørn Dyrby, Vedrana Andersen Dahl, Anders Bjorholm Dahl Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23153v1 Recent advances in volumetric super-resolution (SR) have demonstrated strong performance in medical and scientific imaging, with transformer- and CNN-based approaches achieving impressive results even at extreme scaling factors. In this work, we show that much of this performance stems from training on downsampled data rather than real low-resolution scans. This reliance on downsampling is partly driven by the scarcity of paired high- and low-resolution 3D datasets. To address this, we introduce VoDaSuRe, a large-scale volumetric dataset containing paired high- and low-resolution scans. When training models on VoDaSuRe, we reveal a significant discrepancy: SR models trained on downsampled data produce substantially sharper predictions than those trained on real low-resolution scans, which smooth fine structures. Conversely, applying models trained on downsampled data to real scans preserves more structure but is inaccurate. Our findings suggest that current SR methods are overstated - when applied to real data, they do not recover structures lost in low-resolution scans and instead predict a smoothed average. We argue that progress in deep learning-based volumetric SR requires datasets with paired real scans of high complexity, such as VoDaSuRe. Our dataset and code are publicly available through: https://augusthoeg.github.io/VoDaSuRe/ --- ### Describe-Then-Act: Proactive Agent Steering via Distilled Language-Action World Models Authors: Massimiliano Pappa, Luca Romani, Valentino Sacco, Alessio Palma, Stéphane Lathuilière, Fabio Galasso, Xavier Alameda-Pineda, Indro Spinelli Published: 2026-03-24 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.23149v2 Deploying safety-critical agents requires anticipating the consequences of actions before they are executed. While world models offer a paradigm for this proactive foresight, current approaches relying on visual simulation incur prohibitive latencies, often exceeding several seconds per step. In this work, we challenge the assumption that visual processing is necessary for failure prevention. We show that a trained policy's latent state, combined with its planned actions, already encodes sufficient information to anticipate action outcomes, making visual simulation redundant for failure prevention. To this end, we introduce DILLO (DIstiLLed Language-ActiOn World Model), a fast steering layer that shifts the paradigm from "simulate-then-act" to "describe-then-act." DILLO is trained via cross-modal distillation, where a privileged Vision Language Model teacher annotates offline trajectories and a latent-conditioned Large Language Model student learns to predict semantic outcomes. This creates a text-only inference path, bypassing heavy visual generation entirely, achieving a 14x speedup over baselines. Experiments on MetaWorld and LIBERO demonstrate that DILLO produces high-fidelity descriptions of the next state and is able to steer the policy, improving episode success rate by up to 15 pp and 9.3 pp on average across tasks. Code is available at github.com/MaxPappa/DILLO. --- ### Describe-Then-Act: Proactive Agent Steering via Distilled Language-Action World Models Authors: Massimiliano Pappa, Luca Romani, Valentino Sacco, Alessio Palma, Stéphane Lathuilière, Fabio Galasso, Xavier Alameda-Pineda, Indro Spinelli Published: 2026-03-24 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.23149v1 Deploying safety-critical agents requires anticipating the consequences of actions before they are executed. While world models offer a paradigm for this proactive foresight, current approaches relying on visual simulation incur prohibitive latencies, often exceeding several seconds per step. In this work, we challenge the assumption that visual processing is necessary for failure prevention. We show that a trained policy's latent state, combined with its planned actions, already encodes sufficient information to anticipate action outcomes, making visual simulation redundant for failure prevention. To this end, we introduce DILLO (DIstiLLed Language-ActiOn World Model), a fast steering layer that shifts the paradigm from "simulate-then-act" to "describe-then-act." DILLO is trained via cross-modal distillation, where a privileged Vision Language Model teacher annotates offline trajectories and a latent-conditioned Large Language Model student learns to predict semantic outcomes. This creates a text-only inference path, bypassing heavy visual generation entirely, achieving a 14x speedup over baselines. Experiments on MetaWorld and LIBERO demonstrate that DILLO produces high-fidelity descriptions of the next state and is able to steer the policy, improving episode success rate by up to 15 pp and 9.3 pp on average across tasks. --- ### Why AI-Generated Text Detection Fails: Evidence from Explainable AI Beyond Benchmark Accuracy Authors: Shushanta Pudasaini, Luis Miralles-Pechuán, David Lillis, Marisa Llorens Salvador Published: 2026-03-24 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.23146v2 The widespread adoption of Large Language Models (LLMs) has made the detection of AI-Generated text a pressing and complex challenge. Although many detection systems report high benchmark accuracy, their reliability in real-world settings remains uncertain, and their interpretability is often unexplored. In this work, we investigate whether contemporary detectors genuinely identify machine authorship or merely exploit dataset-specific artefacts. We propose an interpretable detection framework that integrates linguistic feature engineering, machine learning, and explainable AI techniques. When evaluated on two prominent benchmark corpora, namely PAN CLEF 2025 and COLING 2025, our model trained on 30 linguistic features achieves leaderboard-competitive performance, attaining an F1 score of 0.9734. However, systematic cross-domain and cross-generator evaluation reveals substantial generalisation failure: classifiers that excel in-domain degrade significantly under distribution shift. Using SHAP- based explanations, we show that the most influential features differ markedly between datasets, indicating that detectors often rely on dataset-specific stylistic cues rather than stable signals of machine authorship. Further investigation with in-depth error analysis exposes a fundamental tension in linguistic-feature-based AI text detection: the features that are most discriminative on in-domain data are also the features most susceptible to domain shift, formatting variation, and text-length effects. We believe that this knowledge helps build AI detectors that are robust across different settings. To support replication and practical use, we release an open-source Python package that returns both predictions and instance-level explanations for individual texts. --- ### DAK-UCB: Diversity-Aware Prompt Routing for LLMs and Generative Models Authors: Donya Jafari, Farzan Farnia Published: 2026-03-24 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.23140v2 The expansion of generative AI and LLM services underscores the growing need for adaptive mechanisms to select an appropriate available model to respond to a user's prompts. Recent works have proposed offline and online learning formulations to identify the optimal generative AI model for an input prompt, based solely on maximizing prompt-based fidelity evaluation scores, e.g., CLIP-Score in text-to-image generation. However, such fidelity-based selection methods overlook the diversity of generated outputs, and hence, they can fail to address potential diversity shortcomings in the generated responses. In this paper, we introduce the Diversity-Aware Kernelized Upper Confidence Bound (DAK-UCB) method as a contextual bandit algorithm for the online selection of generative models with diversity considerations. The proposed DAK-UCB method incorporates both fidelity and diversity-related metrics into the selection process. We design this framework based on prompt-aware diversity score functions that decompose to a two-sample-based expectation over prompt-output pairs in the previous generation rounds. Specifically, we illustrate the application of our framework using joint kernel distance and kernel entropy measures. Our experimental results demonstrate the effectiveness of DAK-UCB in promoting diversity-aware model selection while maintaining fidelity in the generations for a sequence of prompts. The code is available at https://github.com/Donya-Jafari/DAK-UCB. --- ### HGNet: Scalable Foundation Model for Automated Knowledge Graph Generation from Scientific Literature Authors: Devvrat Joshi, Islem Rekik Published: 2026-03-24 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2603.23136v1 Automated knowledge graph (KG) construction is essential for navigating the rapidly expanding body of scientific literature. However, existing approaches struggle to recognize long multi-word entities, often fail to generalize across domains, and typically overlook the hierarchical nature of scientific knowledge. While general-purpose large language models (LLMs) offer adaptability, they are computationally expensive and yield inconsistent accuracy on specialized tasks. As a result, current KGs are shallow and inconsistent, limiting their utility for exploration and synthesis. We propose a two-stage framework for scalable, zero-shot scientific KG construction. The first stage, Z-NERD, introduces (i) Orthogonal Semantic Decomposition (OSD), which promotes domain-agnostic entity recognition by isolating semantic "turns" in text, and (ii) a Multi-Scale TCQK attention mechanism that captures coherent multi-word entities through n-gram-aware attention heads. The second stage, HGNet, performs relation extraction with hierarchy-aware message passing, explicitly modeling parent, child, and peer relations. To enforce global consistency, we introduce two complementary objectives: a Differentiable Hierarchy Loss to discourage cycles and shortcut edges, and a Continuum Abstraction Field (CAF) Loss that embeds abstraction levels along a learnable axis in Euclidean space. This is the first approach to formalize hierarchical abstraction as a continuous property within standard Euclidean embeddings, offering a simpler alternative to hyperbolic methods. We release SPHERE (https://github.com/basiralab/SPHERE), a multi-domain benchmark for hierarchical relation extraction. Our framework establishes a new state of the art on SciERC, SciER, and SPHERE, improving NER by 8.08% and RE by 5.99% on out-of-distribution tests. In zero-shot settings, gains reach 10.76% for NER and 26.2% for RE. --- ### A Bayesian Learning Approach for Drone Coverage Network: A Case Study on Cardiac Arrest in Scotland Authors: Tathagata Basu, Edoardo Patelli, Gianluca Filippi, Ben Parsonage, Christy Maddock, Massimiliano Vasile, Marco Fossati, Adam Loyd, Shaun Marshall, Paul Gowens Published: 2026-03-24 Categories: cs.LG, stat.AP arXiv: https://arxiv.org/abs/2603.23134v2 Drones are becoming popular as a complementary system for Emergency Medical Services (EMS). Although several pilot studies and flight trials have shown the feasibility of drone-assisted Automated External Defibrillator (AED) delivery, running a full-scale operational network remains challenging due to high capital expenditure and environmental uncertainties. In this paper, we formulate a reliability-informed Bayesian learning framework for designing drone-assisted AED delivery networks under environmental and operational uncertainty. We propose our objective function based on the survival probability of Out-off Hospital Cardiac Arrest (OHCA) patients to identify the ideal locations of drone stations. Moreover, we consider the coverage of existing EMS infrastructure to improve the response reliability in remote areas. We illustrate our proposed method using geographically referenced cardiac arrest data from Scotland. The result shows how environmental variability and spatial demand patterns influence optimal drone station placement across urban and rural regions. In addition, we assess the robustness of the network and evaluate its economic viability using a cost-effectiveness analysis based on expected Quality Adjusted Life Year (QALY). The findings suggest that drone-assisted AED delivery is expected to be cost-effective and has the potential to significantly improve the emergency response coverage in rural and urban areas with longer ambulance response times. --- ### A Bayesian Learning Approach for Drone Coverage Network: A Case Study on Cardiac Arrest in Scotland Authors: Tathagata Basu, Edoardo Patelli, Gianluca Filippi, Ben Parsonage, Christy Maddock, Massimiliano Vasile, Marco Fossati, Adam Loyd, Shaun Marshall, Paul Gowens Published: 2026-03-24 Categories: cs.LG, stat.AP arXiv: https://arxiv.org/abs/2603.23134v1 Drones are becoming popular as a complementary system for \ac{ems}. Although several pilot studies and flight trials have shown the feasibility of drone-assisted \ac{aed} delivery, running a full-scale operational network remains challenging due to high capital expenditure and environmental uncertainties. In this paper, we formulate a reliability-informed Bayesian learning framework for designing drone-assisted \ac{aed} delivery networks under environmental and operational uncertainty. We propose our objective function based on the survival probability of \ac{ohca} patients to identify the ideal locations of drone stations. Moreover, we consider the coverage of existing \ac{ems} infrastructure to improve the response reliability in remote areas. We illustrate our proposed method using geographically referenced cardiac arrest data from Scotland. The result shows how environmental variability and spatial demand patterns influence optimal drone station placement across urban and rural regions. In addition, we assess the robustness of the network and evaluate its economic viability using a cost-effectiveness analysis based on expected \ac{qaly}. The findings suggest that drone-assisted \ac{aed} delivery is expected to be cost-effective and has the potential to significantly improve the emergency response coverage in rural and urban areas with longer ambulance response times. --- ### InterDyad: Interactive Dyadic Speech-to-Video Generation by Querying Intermediate Visual Guidance Authors: Dongwei Pan, Longwei Guo, Jiazhi Guan, Luying Huang, Yiding Li, Haojie Liu, Haocheng Feng, Wei He, Kaisiyuan Wang, Hang Zhou Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23132v1 Despite progress in speech-to-video synthesis, existing methods often struggle to capture cross-individual dependencies and provide fine-grained control over reactive behaviors in dyadic settings. To address these challenges, we propose InterDyad, a framework that enables naturalistic interactive dynamics synthesis via querying structural motion guidance. Specifically, we first design an Interactivity Injector that achieves video reenactment based on identity-agnostic motion priors extracted from reference videos. Building upon this, we introduce a MetaQuery-based modality alignment mechanism to bridge the gap between conversational audio and these motion priors. By leveraging a Multimodal Large Language Model (MLLM), our framework is able to distill linguistic intent from audio to dictate the precise timing and appropriateness of reactions. To further improve lip-sync quality under extreme head poses, we propose Role-aware Dyadic Gaussian Guidance (RoDG) for enhanced lip-synchronization and spatial consistency. Finally, we introduce a dedicated evaluation suite with novelly designed metrics to quantify dyadic interaction. Comprehensive experiments demonstrate that InterDyad significantly outperforms state-of-the-art methods in producing natural and contextually grounded two-person interactions. Please refer to our project page for demo videos: https://interdyad.github.io/. --- ### Polaris: A Gödel Agent Framework for Small Language Models through Experience-Abstracted Policy Repair Authors: Aditya Kakade, Vivek Srivastava, Shirish Karande Published: 2026-03-24 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.23129v4 Gödel agent realize recursive self-improvement: an agent inspects its own policy and traces and then modifies that policy in a tested loop. We introduce Polaris, Gödel agent for compact models that performs policy repair via experience abstraction, turning failures into policy updates through a structured cycle of analysis, strategy formation, abstraction, and minimal code patch repair with conservative checks. Unlike response level self correction or parameter tuning, Polaris makes policy level changes with small, auditable patches that persist in the policy and are reused on unseen instances within each benchmark. As part of the loop, the agent engages in meta reasoning: it explains its errors, proposes concrete revisions to its own policy, and then updates the policy. To enable cumulative policy refinement, we introduce experience abstraction, which distills failures into compact, reusable strategies that transfer to unseen instances. On MGSM, DROP, GPQA, and LitBench (covering arithmetic reasoning, compositional inference, graduate-level problem solving, and creative writing evaluation), a 7-billion-parameter model equipped with Polaris achieves consistent gains over the base policy and competitive baselines. --- ### 3rd Place of MeViS-Audio Track of the 5th PVUW: VIRST-Audio Authors: Jihwan Hong, Jaeyoung Do Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23126v1 Audio-based Referring Video Object Segmentation (ARVOS) requires grounding audio queries into pixel-level object masks over time, posing challenges in bridging acoustic signals with spatio-temporal visual representations. In this report, we present VIRST-Audio, a practical framework built upon a pretrained RVOS model integrated with a vision-language architecture. Instead of relying on audio-specific training, we convert input audio into text using an ASR module and perform segmentation using text-based supervision, enabling effective transfer from text-based reasoning to audio-driven scenarios. To improve robustness, we further incorporate an existence-aware gating mechanism that estimates whether the referred target object is present in the video and suppresses predictions when it is absent, reducing hallucinated masks and stabilizing segmentation behavior. We evaluate our approach on the MeViS-Audio track of the 5th PVUW Challenge, where VIRST-Audio achieves 3rd place, demonstrating strong generalization and reliable performance in audio-based referring video segmentation. --- ### PiCo: Active Manifold Canonicalization for Robust Robotic Visual Anomaly Detection Authors: Teng Yan, Binkai Liu, Shuai Liu, Yue Yu, Bingzhuo Zhong Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23122v1 Industrial deployment of robotic visual anomaly detection (VAD) is fundamentally constrained by passive perception under diverse 6-DoF pose configurations and unstable operating conditions such as illumination changes and shadows, where intrinsic semantic anomalies and physical disturbances coexist and interact. To overcome these limitations, a paradigm shift from passive feature learning to Active Canonicalization is proposed. PiCo (Pose-in-Condition Canonicalization) is introduced as a unified framework that actively projects observations onto a condition-invariant canonical manifold. PiCo operates through a cascaded mechanism. The first stage, Active Physical Canonicalization, enables a robotic agent to reorient objects in order to reduce geometric uncertainty at its source. The second stage, Neural Latent Canonicalization, adopts a three-stage denoising hierarchy consisting of photometric processing at the input level, latent refinement at the feature level, and contextual reasoning at the semantic level, progressively eliminating nuisance factors across representational scales. Extensive evaluations on the large-scale M2AD benchmark demonstrate the superiority of this paradigm. PiCo achieves a state-of-the-art 93.7% O-AUROC, representing a 3.7% improvement over prior methods in static settings, and attains 98.5% accuracy in active closed-loop scenarios. These results demonstrate that active manifold canonicalization is critical for robust embodied perception. --- ### SMSP: A Plug-and-Play Strategy of Multi-Scale Perception for MLLMs to Perceive Visual Illusions Authors: Jinzhe Tu, Ruilei Guo, Zihan Guo, Junxiao Yang, Shiyao Cui, Minlie Huang Published: 2026-03-24 Categories: cs.CV, cs.MM arXiv: https://arxiv.org/abs/2603.23118v2 Recent works have shown that multimodal large language models (MLLMs) are highly vulnerable to hidden-pattern visual illusions, where the hidden content is imperceptible to models but obvious to humans. This deficiency highlights a perceptual misalignment between current MLLMs and humans, and also introduces potential safety concerns. To systematically investigate this failure, we introduce IlluChar, a comprehensive and challenging illusion dataset, and uncover a key underlying mechanism for the models' failure: high-frequency attention bias, where the models are easily distracted by high-frequency background textures in illusion images, causing them to overlook hidden patterns. To address the issue, we propose the Strategy of Multi-Scale Perception (SMSP), a plug-and-play framework that aligns with human visual perceptual strategies. By suppressing distracting high-frequency background signals, SMSP generates images closer to human perception. Our experiments demonstrate that SMSP significantly improves the performance of all evaluated MLLMs on illusion images, for instance, increasing the accuracy of Qwen3-VL-8B-Instruct from 13.0% to 84.0%. Our work provides novel insights into MLLMs' visual perception, and offers a practical and robust solution to enhance it. Our code is publicly available at https://github.com/Tujz2023/SMSP. --- ### TRAP: Hijacking VLA CoT-Reasoning via Adversarial Patches Authors: Zhengxian Huang, Wenjun Zhu, Haoxuan Qiu, Xiaoyu Ji, Wenyuan Xu Published: 2026-03-24 Categories: cs.CR, cs.AI, cs.RO arXiv: https://arxiv.org/abs/2603.23117v2 By integrating Chain-of-Thought (CoT) reasoning, Vision-Language-Action (VLA) models have demonstrated strong capabilities in robotic manipulation, particularly by improving generalization and interpretability. However, the security of CoT-based reasoning mechanisms remains largely unexplored. In this paper, we show that CoT reasoning introduces a novel attack vector for targeted behavior hijacking--for example, causing a robot to mistakenly deliver a knife to a person instead of an apple--without modifying the user's instruction. We first provide empirical evidence that CoT strongly governs action generation, even when it is semantically misaligned with the input instructions. Building on this observation, we propose TRAP, the first targeted behavior-hijacking adversarial attack against CoT-reasoning VLA models. By targeting the reasoning-to-action pathway, TRAP uses an adversarial patch (e.g., a tablecloth placed on the table) to steer intermediate CoT reasoning and downstream actions toward adversary-defined behaviors. Extensive evaluations on three representative reasoning VLAs, spanning distinct CoT reasoning mechanisms, demonstrate the effectiveness of TRAP. Notably, we implemented the patch by printing it on paper in a real-world setting. Our findings highlight the urgent need to secure CoT reasoning in VLA systems. The project page is available at https://zhengxian-huang.github.io/TRAP-website/. --- ### Automatic Segmentation of 3D CT scans with SAM2 using a zero-shot approach Authors: Miquel Lopez Escoriza, Pau Amargant Alvarez Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23116v1 Foundation models for image segmentation have shown strong generalization in natural images, yet their applicability to 3D medical imaging remains limited. In this work, we study the zero-shot use of Segment Anything Model 2 (SAM2) for automatic segmentation of volumetric CT data, without any fine-tuning or domain-specific training. We analyze how SAM2 should be applied to CT volumes and identify its main limitation: the lack of inherent volumetric awareness. To address this, we propose a set of inference-alone architectural and procedural modifications that adapt SAM2's video-based memory mechanism to 3D data by treating CT slices as ordered sequences. We conduct a systematic ablation study on a subset of 500 CT scans from the TotalSegmentator dataset to evaluate prompt strategies, memory propagation schemes and multi-pass refinement. Based on these findings, we select the best-performing configuration and report final results on a bigger sample of the TotalSegmentator dataset comprising 2,500 CT scans. Our results show that, even with frozen weights, SAM2 can produce coherent 3D segmentations when its inference pipeline is carefully structured, demonstrating the feasibility of a fully zero-shot approach for volumetric medical image segmentation. --- ### AgentFoX: LLM Agent-Guided Fusion with eXplainability for AI-Generated Image Detection Authors: Yangxin Yu, Yue Zhou, Bin Li, Kaiqing Lin, Haodong Li, Jiangqun Ni, Bo Cao Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23115v2 The realism of AI-generated images (AIGI) poses increasing challenges for reliable forensic detection, where heterogeneous expert detectors may produce conflicting predictions across diverse generative sources and post-processing conditions. Existing multi-expert fusion methods rely on fixed rules or learned fusion strategies, offering limited ability to assess sample-specific reliability, execute rigorous adjudication of conflicts, and provide evidence-grounded explanations. We propose AgentFoX, an LLM-driven agentic multi-expert framework for AIGI detection that employs a command-and-reasoning core to perform evidence fusion. Following predefined guidelines, the core coordinates designated subtasks to collect semantic and signal-level evidence, reason over structured contexts to determine authenticity, and generate an auditable report for explainability. During this process, Expert Profiles are constructed for model-centric reliability assessment, while Clustering Profiles are built for data-centric contextual analysis, jointly establishing evidence contexts for conflict resolution. Extensive evaluations across diverse benchmarks demonstrate the robustness and generalizability of AgentFoX under complex conditions. --- ### On the Context Sensitivity of LLM Moral Judgment Authors: Adrian Sauter, Mona Schirmer Published: 2026-03-24 Categories: cs.AI, cs.CL, cs.CY, cs.HC arXiv: https://arxiv.org/abs/2603.23114v2 A human's moral decision depends heavily on the context. Yet research on LLM morality has largely studied fixed scenarios. We address this gap by introducing Contextual MoralChoice, a dataset of moral dilemmas with systematic contextual variations known from moral psychology to shift human judgment: consequentialist, emotional, and relational. Evaluating 22 LLMs, we find that nearly all models are context-sensitive, shifting their judgments towards rule-violating behavior. Comparing with a human survey, we find that models and humans are most triggered by different contextual variations, and that a model aligned with human judgments in the base case is not necessarily aligned in its contextual sensitivity. This raises the question of controlling contextual sensitivity, which we address with an activation steering approach that can reliably increase or decrease a model's contextual sensitivity. Code and data: https://github.com/adrian-sauter/contextual_moralchoice. --- ### Between Rules and Reality: On the Context Sensitivity of LLM Moral Judgment Authors: Adrian Sauter, Mona Schirmer Published: 2026-03-24 Categories: cs.AI, cs.CL, cs.CY, cs.HC arXiv: https://arxiv.org/abs/2603.23114v1 A human's moral decision depends heavily on the context. Yet research on LLM morality has largely studied fixed scenarios. We address this gap by introducing Contextual MoralChoice, a dataset of moral dilemmas with systematic contextual variations known from moral psychology to shift human judgment: consequentialist, emotional, and relational. Evaluating 22 LLMs, we find that nearly all models are context-sensitive, shifting their judgments toward rule-violating behavior. Comparing with a human survey, we find that models and humans are most triggered by different contextual variations, and that a model aligned with human judgments in the base case is not necessarily aligned in its contextual sensitivity. This raises the question of controlling contextual sensitivity, which we address with an activation steering approach that can reliably increase or decrease a model's contextual sensitivity. --- ### High-Resolution Tensor-Network Fourier Methods for Exponentially Compressed Non-Gaussian Aggregate Distributions Authors: Juan José Rodríguez-Aldavero, Juan José García-Ripoll Published: 2026-03-24 Categories: stat.ML, cs.LG, math.NA, quant-ph arXiv: https://arxiv.org/abs/2603.23106v1 Characteristic functions of weighted sums of independent random variables exhibit low-rank structure in the quantized tensor train (QTT) representation, also known as matrix product states (MPS), enabling up to exponential compression of their fully non-Gaussian probability distributions. Under variable independence, the global characteristic function factorizes into local terms. Its low-rank QTT structure arises from intrinsic spectral smoothness in continuous models, or from spectral energy concentration as the number of components $D$ grows in discrete models. We demonstrate this on weighted sums of Bernoulli and lognormal random variables. In the former, despite an adversarial, incompressible small-$D$ regime, the characteristic function undergoes a sharp bond-dimension collapse for $D \gtrsim 300$ components, enabling polylogarithmic time and memory scaling. In the latter, the approach reaches high-resolution discretizations of $N = 2^{30}$ frequency modes on standard hardware, far beyond the $N = 2^{24}$ ceiling of dense implementations. These compressed representations enable efficient computation of Value at Risk (VaR) and Expected Shortfall (ES), supporting applications in quantitative finance and beyond. --- ### NeuroSeg Meets DINOv3: Transferring 2D Self-Supervised Visual Priors to 3D Neuron Segmentation via DINOv3 Initialization Authors: Yik San Cheng, Runkai Zhao, Weidong Cai Published: 2026-03-24 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.23104v1 2D visual foundation models, such as DINOv3, a self-supervised model trained on large-scale natural images, have demonstrated strong zero-shot generalization, capturing both rich global context and fine-grained structural cues. However, an analogous 3D foundation model for downstream volumetric neuroimaging remains lacking, largely due to the challenges of 3D image acquisition and the scarcity of high-quality annotations. To address this gap, we propose to adapt the 2D visual representations learned by DINOv3 to a 3D biomedical segmentation model, enabling more data-efficient and morphologically faithful neuronal reconstruction. Specifically, we design an inflation-based adaptation strategy that inflates 2D filters into 3D operators, preserving semantic priors from DINOv3 while adapting to 3D neuronal volume patches. In addition, we introduce a topology-aware skeleton loss to explicitly enforce structural fidelity of graph-based neuronal arbor reconstruction. Extensive experiments on four neuronal imaging datasets, including two from BigNeuron and two public datasets, NeuroFly and CWMBS, demonstrate consistent improvements in reconstruction accuracy over SoTA methods, with average gains of 2.9% in Entire Structure Average, 2.8% in Different Structure Average, and 3.8% in Percentage of Different Structure. Code: https://github.com/yy0007/NeurINO. --- ### SpecXMaster Technical Report Authors: Yutang Ge, Yaning Cui, Hanzheng Li, Jun-Jie Wang, Fanjie Xu, Jinhan Dong, Yongqi Jin, Dongxu Cui, Peng Jin, Guojiang Zhao, Hengxing Cai, Tianci Yangfeng et al. Published: 2026-03-24 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.23101v4 Intelligent spectroscopy serves as a pivotal element in AI-driven closed-loop scientific discovery, functioning as the critical bridge between matter structure and artificial intelligence. However, conventional expert-dependent spectral interpretation encounters substantial hurdles, including susceptibility to human bias and error, dependence on limited specialized expertise, and variability across interpreters. To address these challenges, we propose SpecXMaster, an intelligent framework leveraging Agentic Reinforcement Learning (RL) for NMR molecular spectral interpretation. SpecXMaster enables automated extraction of multiplicity information from both 1H and 13C spectra directly from raw FID (free induction decay) data. This end-to-end pipeline enables fully automated interpretation of NMR spectra into chemical structures. It demonstrates superior performance across multiple public NMR interpretation benchmarks and has been refined through iterative evaluations by professional chemical spectroscopists. We believe that SpecXMaster, as a novel methodological paradigm for spectral interpretation, will have a profound impact on the organic chemistry community. ---