# Research Papers — 2026-03 · bundle page 233 of 243 > Full-text markdown of papers #11601–11650 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/233 [← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/232.md) [Next bundle →](https://aipod.dev/papers/archive/2026-03/page/234.md) ### Intrinsic Task Symmetry Drives Generalization in Algorithmic Tasks Authors: Hyeonbin Hwang, Yeachan Park Published: 2026-03-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.01968v1 Grokking, the sudden transition from memorization to generalization, is characterized by the emergence of low-dimensional representations, yet the mechanism underlying this organization remains elusive. We propose that intrinsic task symmetries primarily drive grokking and shape the geometry of the model's representation space. We identify a consistent three-stage training dynamic underlying grokking: (i) memorization, (ii) symmetry acquisition, and (iii) geometric organization. We show that generalization emerges during the symmetry acquisition phase, after which representations reorganize into a structured, task-aligned geometry. We validate this symmetry-driven account across diverse algorithmic domains, including algebraic, structural, and relational reasoning tasks. Building on these findings, we introduce a symmetry-based diagnostic that anticipates the onset of generalization and propose strategies to accelerate it. Together, our results establish intrinsic symmetry as the key factor enabling neural networks to move beyond memorization and achieve robust algorithmic reasoning. --- ### AMemGym: Interactive Memory Benchmarking for Assistants in Long-Horizon Conversations Authors: Cheng Jiayang, Dongyu Ru, Lin Qiu, Yiyang Li, Xuezhi Cao, Yangqiu Song, Xunliang Cai Published: 2026-03-02 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.01966v1 Long-horizon interactions between users and LLM-based assistants necessitate effective memory management, yet current approaches face challenges in training and evaluation of memory. Existing memory benchmarks rely on static, off-policy data as context, limiting evaluation reliability and scalability. To address these gaps, we introduce AMemGym, an interactive environment enabling on-policy evaluation and optimization for memory-driven personalization. AMemGym employs structured data sampling to predefine user profiles, state-dependent questions, and state evolution trajectories, enabling cost-effective generation of high-quality, evaluation-aligned interactions. LLM-simulated users expose latent states through role-play while maintaining structured state consistency. Comprehensive metrics based on structured data guide both assessment and optimization of assistants. Extensive experiments reveal performance gaps in existing memory systems (e.g., RAG, long-context LLMs, and agentic memory) and corresponding reasons. AMemGym not only enables effective selection among competing approaches but also can potentially drive the self-evolution of memory management strategies. By bridging structured state evolution with free-form interactions, our framework provides a scalable, diagnostically rich environment for advancing memory capabilities in conversational agents. --- ### CoVAE: correlated multimodal generative modeling Authors: Federico Caretti, Guido Sanguinetti Published: 2026-03-02 Categories: cs.LG, q-bio.QM arXiv: https://arxiv.org/abs/2603.01965v1 Multimodal Variational Autoencoders have emerged as a popular tool to extract effective representations from rich multimodal data. However, such models rely on fusion strategies in latent space that destroy the joint statistical structure of the multimodal data, with profound implications for generation and uncertainty quantification. In this work, we introduce Correlated Variational Autoencoders (CoVAE), a new generative architecture that captures the correlations between modalities. We test CoVAE on a number of real and synthetic data sets demonstrating both accurate cross-modal reconstruction and effective quantification of the associated uncertainties. --- ### TiledAttention: a CUDA Tile SDPA Kernel for PyTorch Authors: Taimur Khan Published: 2026-03-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.01960v2 TiledAttention is a scaled dot-product attention (SDPA) forward operator for SDPA research on NVIDIA GPUs. Implemented in cuTile Python (TileIR) and exposed as a PyTorch-callable function, it is easier to modify than low-level CUDA templates while retaining realistic behavior via online softmax and tiled $K,V$ streaming. Algorithmically, TiledAttention follows the established FlashAttention-style online-softmax formulation; our novelty is the cuTile/TileIR implementation strategy, schedule-level modifiability, and reproducible benchmarking/profiling workflow. The approach is both performant and directly editable at the schedule level from Python (tile shapes, staging, shared-memory layout), enabling rapid, reproducible kernel research without template-heavy CUDA/CUTLASS rewrites. We benchmark TiledAttention on an NVIDIA DGX GB10 node with a reproducible harness and compare against PyTorch SDPA (auto-dispatch), explicit unfused baselines (torch_sdpa_math, standard eager attention), and forced backend probes (FlashAttention2, EffecientAttention, CuDNN Attention) across sequence length, head dimension, and precision (FP16/BF16). While production fused baselines remain stronger overall, TiledAttention delivers large speedups over standard eager attention paths and is available for direct use within PyTorch workflows, providing a practical balance between performance and customizability. --- ### The Expressive Limits of Diagonal SSMs for State-Tracking Authors: Mehran Shakerinava, Behnoush Khavari, Siamak Ravanbakhsh, Sarath Chandar Published: 2026-03-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.01959v2 State-Space Models (SSMs) have recently been shown to achieve strong empirical performance on a variety of long-range sequence modeling tasks while remaining efficient and highly-parallelizable. However, the theoretical understanding of their expressive power remains limited. In this work, we study the expressivity of input-Dependent Complex-valued Diagonal (DCD) SSMs on sequential state-tracking tasks. We show that single-layer DCD SSMs cannot express state-tracking of any non-Abelian group at finite precision. More generally, we show that $k$-layer DCD SSMs can express state-tracking of a group if and only if that group has a subnormal series of length $k$, with Abelian factors. That is, we identify the precise expressivity range of $k$-layer DCD SSMs within the solvable groups. Empirically, we find that multi-layer models often fail to learn state-tracking for non-Abelian groups, highlighting a gap between expressivity and learnability. --- ### The Expressive Limits of Diagonal SSMs for State-Tracking Authors: Mehran Shakerinava, Behnoush Khavari, Siamak Ravanbakhsh, Sarath Chandar Published: 2026-03-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.01959v1 State-Space Models (SSMs) have recently been shown to achieve strong empirical performance on a variety of long-range sequence modeling tasks while remaining efficient and highly-parallelizable. However, the theoretical understanding of their expressive power remains limited. In this work, we study the expressivity of input-Dependent Complex-valued Diagonal (DCD) SSMs on sequential state-tracking tasks. We show that single-layer DCD SSMs cannot express state-tracking of any non-Abelian group at finite precision. More generally, we show that $k$-layer DCD SSMs can express state-tracking of a group if and only if that group has a subnormal series of length $k$, with Abelian factors. That is, we identify the precise expressivity range of $k$-layer DCD SSMs within the solvable groups. Empirically, we find that multi-layer models often fail to learn state-tracking for non-Abelian groups, highlighting a gap between expressivity and learnability. --- ### Closed-Loop Action Chunks with Dynamic Corrections for Training-Free Diffusion Policy Authors: Pengyuan Wu, Pingrui Zhang, Zhigang Wang, Dong Wang, Bin Zhao, Xuelong Li Published: 2026-03-02 Categories: cs.RO, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2603.01953v2 Diffusion-based policies have achieved remarkable results in robotic manipulation but often struggle to adapt rapidly in dynamic scenarios, leading to delayed responses or task failures. We present DCDP, a Dynamic Closed-Loop Diffusion Policy framework that integrates chunk-based action generation with real-time correction. DCDP integrates a self-supervised dynamic feature encoder, cross-attention fusion, and an asymmetric action encoder-decoder to inject environmental dynamics before action execution, achieving real-time closed-loop action correction and enhancing the system's adaptability in dynamic scenarios. In dynamic PushT simulations, DCDP improves adaptability by 19\% without retraining while requiring only 5\% additional computation. Its modular design enables plug-and-play integration, achieving both temporal coherence and real-time responsiveness in dynamic robotic scenarios, including real-world manipulation tasks. The project page is at: https://github.com/wupengyuan/dcdp --- ### LiveCultureBench: a Multi-Agent, Multi-Cultural Benchmark for Large Language Models in Dynamic Social Simulations Authors: Viet-Thanh Pham, Lizhen Qu, Thuy-Trang Vu, Gholamreza Haffari, Dinh Phung Published: 2026-03-02 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.01952v1 Large language models (LLMs) are increasingly deployed as autonomous agents, yet evaluations focus primarily on task success rather than cultural appropriateness or evaluator reliability. We introduce LiveCultureBench, a multi-cultural, dynamic benchmark that embeds LLMs as agents in a simulated town and evaluates them on both task completion and adherence to socio-cultural norms. The simulation models a small city as a location graph with synthetic residents having diverse demographic and cultural profiles. Each episode assigns one resident a daily goal while others provide social context. An LLM-based verifier generates structured judgments on norm violations and task progress, which we aggregate into metrics capturing task-norm trade-offs and verifier uncertainty. Using LiveCultureBench across models and cultural profiles, we study (i) cross-cultural robustness of LLM agents, (ii) how they balance effectiveness against norm sensitivity, and (iii) when LLM-as-a-judge evaluation is reliable for automated benchmarking versus when human oversight is needed. --- ### Accelerating Single-Pass SGD for Generalized Linear Prediction Authors: Qian Chen, Shihong Ding, Cong Fang Published: 2026-03-02 Categories: cs.LG, math.OC, stat.ML arXiv: https://arxiv.org/abs/2603.01951v1 We study generalized linear prediction under a streaming setting, where each iteration uses only one fresh data point for a gradient-level update. While momentum is well-established in deterministic optimization, a fundamental open question is whether it can accelerate such single-pass non-quadratic stochastic optimization. We propose the first algorithm that successfully incorporates momentum via a novel data-dependent proximal method, achieving dual-momentum acceleration. Our derived excess risk bound decomposes into three components: an improved optimization error, a minimax optimal statistical error, and a higher-order model-misspecification error. The proof handles mis-specification via a fine-grained stationary analysis of inner updates, while localizing statistical error through a two-phase outer-loop analysis. As a result, we resolve the open problem posed by Jain et al. (2018a) and demonstrate that momentum acceleration is more effective than variance reduction for generalized linear prediction in the streaming setting. --- ### Semantic Similarity is a Spurious Measure of Comic Understanding: Lessons Learned from Hallucinations in a Benchmarking Experiment Authors: Christopher Driggers-Ellis, Nachiketh Tibrewal, Rohit Bogulla, Harsh Khanna, Sangpil Youm, Christan Grant, Bonnie Dorr Published: 2026-03-02 Categories: cs.LG, cs.CL, cs.CV arXiv: https://arxiv.org/abs/2603.01950v1 A system that enables blind or visually impaired users to access comics/manga would introduce a new medium of storytelling to this community. However, no such system currently exists. Generative vision-language models (VLMs) have shown promise in describing images and understanding comics, but most research on comic understanding is limited to panel-level analysis. To fully support blind and visually impaired users, greater attention must be paid to page-level understanding and interpretation. In this work, we present a preliminary benchmark of VLM performance on comic interpretation tasks. We identify and categorize hallucinations that emerge during this process, organizing them into generalized object-hallucination taxonomies. We conclude with guidance on future research, emphasizing hallucination mitigation and improved data curation for comic interpretation. --- ### Probabilistic Retrofitting of Learned Simulators Authors: Cristiana Diaconu, Miles Cranmer, Richard E. Turner, Tanya Marwah, Payel Mukhopadhyay Published: 2026-03-02 Categories: cs.LG, cs.AI, cs.CE arXiv: https://arxiv.org/abs/2603.01949v2 Dominant approaches for modelling Partial Differential Equations (PDEs) rely on deterministic predictions, yet many physical systems of interest are inherently chaotic and uncertain. While training probabilistic models from scratch is possible, it is computationally expensive and fails to leverage the significant resources already invested in high-performing deterministic backbones. In this work, we adopt a training-efficient strategy to transform pre-trained deterministic models into probabilistic ones via retrofitting with a proper scoring rule: the Continuous Ranked Probability Score (CRPS). Crucially, this approach is architecture-agnostic: it applies the same adaptation mechanism across distinct model backbones with minimal code modifications. The method proves highly effective across different scales of pre-training: for models trained on single dynamical systems, we achieve 20-54% reductions in rollout CRPS and up to 30% improvements in variance-normalised RMSE (VRMSE) relative to compute-matched deterministic fine-tuning. We further validate our approach on a PDE foundation model, trained on multiple systems and retrofitted on the dataset of interest, to show that our probabilistic adaptation yields an improvement of up to 40% in CRPS and up to 15% in VRMSE compared to deterministic fine-tuning. Validated across diverse architectures and dynamics, our results show that probabilistic PDE modelling need not require retraining from scratch, but can be unlocked from existing deterministic backbones with modest additional training cost. --- ### PreSight: Preoperative Outcome Prediction for Parkinson's Disease via Region-Prior Morphometry and Patient-Specific Weighting Authors: Yand Wang, Chen Zhang, Lanyun Zhu, Yixin Chen, Qunbo Wang, Yutong Bai, Jurgen Germann, Yinghong Wen, Shuai Shao Published: 2026-03-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.01948v1 Preoperative improvement rate prediction for Parkinson's disease surgery is clinically important yet difficult because imaging signals are subtle and patients are heterogeneous. We address this setting, where only information available before surgery is used, and the goal is to predict patient-specific postoperative motor benefit. We present PreSight, a presurgical outcome model that fuses clinical priors with preoperative MRI and deformation-based morphometry (DBM) and adapts regional importance through a patient-specific weighting module. The model produces end-to-end, calibrated, decision-ready predictions with patient-level explanations. We evaluate PreSight on a real-world two-center cohort of 400 subjects with multimodal presurgical inputs and postoperative improvement labels. PreSight outperforms strong clinical, imaging-only, and multimodal baselines. It attains 88.89% accuracy on internal validation and 85.29% on an external-center test for responder classification and shows better probability calibration and higher decision-curve net benefit. Ablations and analyses confirm the contribution of DBM and the patient-specific weighting module and indicate that the model emphasizes disease-relevant regions in a patient-specific manner. These results demonstrate that integrating clinical prior knowledge with region-adaptive morphometry enables reliable presurgical decision support in routine practice. --- ### physfusion: A Transformer-based Dual-Stream Radar and Vision Fusion Framework for Open Water Surface Object Detection Authors: Yuting Wan, Liguo Sun, Jiuwu Hao, Zao Zhang, Pin LV Published: 2026-03-02 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.01947v1 Detecting water-surface targets for Unmanned Surface Vehicles (USVs) is challenging due to wave clutter, specular reflections, and weak appearance cues in long-range observations. Although 4D millimeter-wave radar complements cameras under degraded illumination, maritime radar point clouds are sparse and intermittent, with reflectivity attributes exhibiting heavy-tailed variations under scattering and multipath, making conventional fusion designs struggle to exploit radar cues effectively. We propose PhysFusion, a physics-informed radar-image detection framework for water-surface perception. The framework integrates: (1) a Physics-Informed Radar Encoder (PIR Encoder) with an RCS Mapper and Quality Gate, transforming per-point radar attributes into compact scattering priors and predicting point-wise reliability for robust feature learning under clutter; (2) a Radar-guided Interactive Fusion Module (RIFM) performing query-level radar-image fusion between semantically enriched radar features and multi-scale visual features, with the radar branch modeled by a dual-stream backbone including a point-based local stream and a transformer-based global stream using Scattering-Aware Self-Attention (SASA); and (3) a Temporal Query Aggregation module (TQA) aggregating frame-wise fused queries over a short temporal window for temporally consistent representations. Experiments on WaterScenes and FLOW demonstrate that PhysFusion achieves 59.7% mAP50:95 and 90.3% mAP50 on WaterScenes (T=5 radar history) using 5.6M parameters and 12.5G FLOPs, and reaches 94.8% mAP50 and 46.2% mAP50:95 on FLOW under radar+camera setting. Ablation studies quantify the contributions of PIR Encoder, SASA-based global reasoning, and RIFM. --- ### When Numbers Tell Half the Story: Human-Metric Alignment in Topic Model Evaluation Authors: Thibault Prouteau, Francis Lareau, Nicolas Dugué, Jean-Charles Lamirel, Christophe Malaterre Published: 2026-03-02 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.01945v1 Topic models uncover latent thematic structures in text corpora, yet evaluating their quality remains challenging, particularly in specialized domains. Existing methods often rely on automated metrics like topic coherence and diversity, which may not fully align with human judgment. Human evaluation tasks, such as word intrusion, provide valuable insights but are costly and primarily validated on general-domain corpora. This paper introduces Topic Word Mixing (TWM), a novel human evaluation task assessing inter-topic distinctness by testing whether annotators can distinguish between word sets from single or mixed topics. TWM complements word intrusion's focus on intra-topic coherence and provides a human-grounded counterpart to diversity metrics. We evaluate six topic models - both statistical and embedding-based (LDA, NMF, Top2Vec, BERTopic, CFMF, CFMF-emb) - comparing automated metrics with human evaluation methods based on nearly 4,000 annotations from a domain-specific corpus of philosophy of science publications. Our findings reveal that word intrusion and coherence metrics do not always align, particularly in specialized domains, and that TWM captures human-perceived distinctness while appearing to align with diversity metrics. We release the annotated dataset and task generation code. This work highlights the need for evaluation frameworks bridging automated and human assessments, particularly for domain-specific corpora. --- ### MobileMold: A Smartphone-Based Microscopy Dataset for Food Mold Detection Authors: Dinh Nam Pham, Leonard Prokisch, Bennet Meyer, Jonas Thumbs Published: 2026-03-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.01944v1 Smartphone clip-on microscopes turn everyday devices into low-cost, portable imaging systems that can even reveal fungal structures at the microscopic level, enabling mold inspection beyond unaided visual checks. In this paper, we introduce MobileMold, an open smartphone-based microscopy dataset for food mold detection and food classification. MobileMold contains 4,941 handheld microscopy images spanning 11 food types, 4 smartphones, 3 microscopes, and diverse real-world conditions. Beyond the dataset release, we establish baselines for (i) mold detection and (ii) food-type classification, including a multi-task setting that predicts both attributes. Across multiple pretrained deep learning architectures and augmentation strategies, we obtain near-ceiling performance (accuracy = 0.9954, F1 = 0.9954, MCC = 0.9907), validating the utility of our dataset for detecting food spoilage. To increase transparency, we complement our evaluation with saliency-based visual explanations highlighting mold regions associated with the model's predictions. MobileMold aims to contribute to research on accessible food-safety sensing, mobile imaging, and exploring the potential of smartphones enhanced with attachments. --- ### Ignore All Previous Instructions: Jailbreaking as a de-escalatory peace building practise to resist LLM social media bots Authors: Huw Day, Adrianna Jezierska, Jessica Woodgate Published: 2026-03-02 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2603.01942v1 Large Language Models have intensified the scale and strategic manipulation of political discourse on social media, leading to conflict escalation. The existing literature largely focuses on platform-led moderation as a countermeasure. In this paper, we propose a user-centric view of "jailbreaking" as an emergent, non-violent de-escalation practice. Online users engage with suspected LLM-powered accounts to circumvent large language model safeguards, exposing automated behaviour and disrupting the circulation of misleading narratives. --- ### BAED: a New Paradigm for Few-shot Graph Learning with Explanation in the Loop Authors: Chao Chen, Xujia Li, Dongsheng Hong, Shanshan Lin, Xiangwen Liao, Chuanyi Liu, Lei Chen Published: 2026-03-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.01941v1 The challenges of training and inference in few-shot environments persist in the area of graph representation learning. The quality and quantity of labels are often insufficient due to the extensive expert knowledge required to annotate graph data. In this context, Few-Shot Graph Learning (FSGL) approaches have been developed over the years. Through sophisticated neural architectures and customized training pipelines, these approaches enhance model adaptability to new label distributions. However, compromises in \textcolor{black}{the model's} robustness and interpretability can result in overfitting to noise in labeled data and degraded performance. This paper introduces the first explanation-in-the-loop framework for the FSGL problem, called BAED. We novelly employ the belief propagation algorithm to facilitate label augmentation on graphs. Then, leveraging an auxiliary graph neural network and the gradient backpropagation method, our framework effectively extracts explanatory subgraphs surrounding target nodes. The final predictions are based on these informative subgraphs while mitigating the influence of redundant information from neighboring nodes. Extensive experiments on seven benchmark datasets demonstrate superior prediction accuracy, training efficiency, and explanation quality of BAED. As a pioneer, this work highlights the potential of the explanation-based research paradigm in FSGL. --- ### CoVe: Training Interactive Tool-Use Agents via Constraint-Guided Verification Authors: Jinpeng Chen, Cheng Gong, Hanbo Li, Ziru Liu, Zichen Tian, Xinyu Fu, Shi Wu, Chenyang Zhang, Wu Zhang, Suiyun Zhang, Dandan Tu, Rui Liu Published: 2026-03-02 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.01940v1 Developing multi-turn interactive tool-use agents is challenging because real-world user needs are often complex and ambiguous, yet agents must execute deterministic actions to satisfy them. To address this gap, we introduce \textbf{CoVe} (\textbf{Co}nstraint-\textbf{Ve}rification), a post-training data synthesis framework designed for training interactive tool-use agents while ensuring both data complexity and correctness. CoVe begins by defining explicit task constraints, which serve a dual role: they guide the generation of complex trajectories and act as deterministic verifiers for assessing trajectory quality. This enables the creation of high-quality training trajectories for supervised fine-tuning (SFT) and the derivation of accurate reward signals for reinforcement learning (RL). Our evaluation on the challenging $τ^2$-bench benchmark demonstrates the effectiveness of the framework. Notably, our compact \textbf{CoVe-4B} model achieves success rates of 43.0\% and 59.4\% in the Airline and Retail domains, respectively; its overall performance significantly outperforms strong baselines of similar scale and remains competitive with models up to $17\times$ its size. These results indicate that CoVe provides an effective and efficient pathway for synthesizing training data for state-of-the-art interactive tool-use agents. To support future research, we open-source our code, trained model, and the full set of 12K high-quality trajectories used for training. --- ### Explanation-Guided Adversarial Training for Robust and Interpretable Models Authors: Chao Chen, Yanhui Chen, Shanshan Lin, Dongsheng Hong, Shu Wu, Xiangwen Liao, Chuanyi Liu Published: 2026-03-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.01938v1 Deep neural networks (DNNs) have achieved remarkable performance in many tasks, yet they often behave as opaque black boxes. Explanation-guided learning (EGL) methods steer DNNs using human-provided explanations or supervision on model attributions. These approaches improve interpretability but typically assume benign inputs and incur heavy annotation costs. In contrast, both predictions and saliency maps of DNNs could dramatically alter facing imperceptible perturbations or unseen patterns. Adversarial training (AT) can substantially improve robustness, but it does not guarantee that model decisions rely on semantically meaningful features. In response, we propose Explanation-Guided Adversarial Training (EGAT), a unified framework that integrates the strength of AT and EGL to simultaneously improve prediction performance, robustness, and explanation quality. EGAT generates adversarial examples on the fly while imposing explanation-based constraints on the model. By jointly optimizing classification performance, adversarial robustness, and attributional stability, EGAT is not only more resistant to unexpected cases, including adversarial attacks and out-of-distribution (OOD) scenarios, but also offer human-interpretable justifications for the decisions. We further formalize EGAT within the Probably Approximately Correct learning framework, demonstrating theoretically that it yields more stable predictions under unexpected situations compared to standard AT. Empirical evaluations on OOD benchmark datasets show that EGAT consistently outperforms competitive baselines in both clean accuracy and adversarial accuracy +37% while producing more semantically meaningful explanations, and requiring only a limited increase +16% in training time. --- ### Dream2Learn: Structured Generative Dreaming for Continual Learning Authors: Salvatore Calcagno, Matteo Pennisi, Federica Proietto Salanitri, Amelia Sorrenti, Simone Palazzo, Concetto Spampinato, Giovanni Bellitto Published: 2026-03-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.01935v1 Continual learning requires balancing plasticity and stability while mitigating catastrophic forgetting. Inspired by human dreaming as a mechanism for internal simulation and knowledge restructuring, we introduce Dream2Learn (D2L), a framework in which a model autonomously generates structured synthetic experiences from its own internal representations and uses them for self-improvement. Rather than reconstructing past data as in generative replay, D2L enables a classifier to create novel, semantically distinct dreamed classes that are coherent with its learned knowledge yet do not correspond to previously observed data. These dreamed samples are produced by conditioning a frozen diffusion model through soft prompt optimization driven by the classifier itself. The generated data are not used to replace memory, but to expand and reorganize the representation space, effectively allowing the network to self-train on internally synthesized concepts. By integrating dreamed classes into continual training, D2L proactively structures latent features to support forward knowledge transfer and adaptation to future tasks. This prospective self-training mechanism mirrors the role of sleep in consolidating and reorganizing memory, turning internal simulations into a tool for improved generalization. Experiments on Mini-ImageNet, FG-ImageNet, and ImageNet-R demonstrate that D2L consistently outperforms strong rehearsal-based baselines and achieves positive forward transfer, confirming its ability to enhance adaptability through internally generated training signals. --- ### BAWSeg: A UAV Multispectral Benchmark for Barley Weed Segmentation Authors: Haitian Wang, Xinyu Wang, Muhammad Ibrahim, Dustin Severtson, Ajmal Mian Published: 2026-03-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.01932v2 Accurate weed mapping in cereal fields requires pixel-level segmentation from UAV imagery that remains reliable across fields, seasons, and illumination. Existing multispectral pipelines often depend on thresholded vegetation indices, which are brittle under radiometric drift and mixed crop--weed pixels, or on single-stream CNN and Transformer backbones that ingest stacked bands and indices, where radiance cues and normalized index cues interfere and reduce sensitivity to small weed clusters embedded in crop canopy. We propose VISA, a two-stream segmentation network that decouples these cues and fuses them at native resolution. The radiance stream learns from calibrated five-band reflectance using local residual convolutions, channel recalibration, spatial gating, and skip-connected decoding, which preserve fine textures, row boundaries, and small weed structures that are often weakened after ratio-based index compression. The index stream operates on vegetation-index maps with windowed self-attention to model local structure efficiently, state-space layers to propagate field-scale context without quadratic attention cost, and Slot Attention to form stable region descriptors that improve discrimination of sparse weeds under canopy mixing. To support supervised training and deployment-oriented evaluation, we introduce BAWSeg, a four-year UAV multispectral dataset collected over commercial barley paddocks in Western Australia, providing radiometrically calibrated blue, green, red, red edge, and near-infrared orthomosaics, derived vegetation indices, and dense crop, weed, and other labels with leakage-free block splits. On BAWSeg, VISA achieves 75.6% mIoU and 63.5% weed IoU with 22.8 M parameters, outperforming a multispectral SegFormer-B1 baseline by 1.2 mIoU and 1.9 weed IoU. Under cross-plot and cross-year protocols, VISA maintains 71.2% and 69.2% mIoU, respectively. --- ### From Variance to Invariance: Qualitative Content Analysis for Narrative Graph Annotation Authors: Junbo Huang, Max Weinig, Ulrich Fritsche, Ricardo Usbeck Published: 2026-03-02 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.01930v2 Narratives in news discourse play a critical role in shaping public understanding of economic events, such as inflation. Annotating and evaluating these narratives in a structured manner remains a key challenge for Natural Language Processing (NLP). In this work, we introduce a narrative graph annotation framework that integrates principles from qualitative content analysis (QCA) to prioritize annotation quality by reducing annotation errors. We present a dataset of inflation narratives annotated as directed acyclic graphs (DAGs), where nodes represent events and edges encode causal relations. To evaluate annotation quality, we employed a $6\times3$ factorial experimental design to examine the effects of narrative representation (six levels) and distance metric type (three levels) on inter-annotator agreement (Krippendorrf's $α$), capturing the presence of human label variation (HLV) in narrative interpretations. Our analysis shows that (1) lenient metrics (overlap-based distance) overestimate reliability, and (2) locally-constrained representations (e.g., one-hop neighbors) reduce annotation variability. Our annotation and implementation of graph-based Krippendorrf's $α$ are open-sourced. The annotation framework and evaluation results provide practical guidance for NLP research on graph-based narrative annotation under HLV. --- ### LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving Authors: Yuechen Luo, Fang Li, Shaoqing Xu, Yang Ji, Zehan Zhang, Bing Wang, Yuannan Shen, Jianwei Cui, Long Chen, Guang Chen, Hangjun Ye, Zhi-Xin Yang et al. Published: 2026-03-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.01928v2 While Vision-Language-Action (VLA) models have revolutionized autonomous driving by unifying perception and planning, their reliance on explicit textual Chain-of-Thought (CoT) leads to semantic-perceptual decoupling and perceptual-symbolic conflicts. Recent shifts toward latent reasoning attempt to bypass these bottlenecks by thinking in continuous hidden space. However, without explicit intermediate constraints, standard latent CoT often operates as a physics-agnostic representation. To address this, we propose the Latent Spatio-Temporal VLA (LaST-VLA), a framework shifting the reasoning paradigm from discrete symbolic processing into a physically grounded Latent Spatio-Temporal CoT. By implementing a dual-feature alignment mechanism, we distill geometric constraints from 3D foundation models and dynamic foresight from world models directly into the latent space. Coupled with a progressive SFT training strategy that transitions from feature alignment to trajectory generation, and refined via Reinforcement Learning with Group Relative Policy Optimization (GRPO) to ensure safety and rule compliance. \method~setting a new record on NAVSIM v1 (91.3 PDMS) and NAVSIM v2 (87.1 EPDMS), while excelling in spatial-temporal reasoning on SURDS and NuDynamics benchmarks. --- ### MealRec: Multi-granularity Sequential Modeling via Hierarchical Diffusion Models for Micro-Video Recommendation Authors: Xinxin Dong, Haokai Ma, Yuze Zheng, Yongfu Zha, Yonghui Yang, Xiaodong Wang Published: 2026-03-02 Categories: cs.IR, cs.CV arXiv: https://arxiv.org/abs/2603.01926v1 Micro-video recommendation aims to capture user preferences from the collaborative and context information of the interacted micro-videos, thereby predicting the appropriate videos. This target is often hindered by the inherent noise within multimodal content and unreliable implicit feedback, which weakens the correspondence between behaviors and underlying interests. While conventional works have predominantly approached such scenario through behavior-augmented modeling and content-centric multimodal analysis, these paradigms can inadvertently give rise to two non-trivial challenges: preference-irrelative video representation extraction and inherent modality conflicts. To address these issues, we propose a Multi-granularity sequential modeling method via hierarchical diffusion models for micro-video Recommendation (MealRec), which simultaneously considers temporal correlations during preference modeling from intra- and inter-video perspectives. Specifically, we first propose Temporal-guided Content Diffusion (TCD) to refine video representations under intra-video temporal guidance and personalized collaborative signals to emphasize salient content while suppressing redundancy. To achieve the semantically coherent preference modeling, we further design the Noise-unconditional Preference Denoising (NPD) to recovers informative user preferences from corrupted states under the blind denoising. Extensive experiments and analyses on four micro-video datasets from two platforms demonstrate the effectiveness, universality, and robustness of our MealRec, further uncovering the effective mechanism of our proposed TCD and NPD. The source code and corresponding dataset will be available upon acceptance. --- ### Bound Propagation meets Constraint Simplification: Improving Logic-based XAI for Neural Networks Authors: Ronaldo Gomes, Jairo Ribeiro, Luiz Queiroz, Thiago Alves Rocha Published: 2026-03-02 Categories: cs.LO, cs.LG arXiv: https://arxiv.org/abs/2603.01923v1 Logic-based methods for explaining neural network decisions offer formal guarantees of correctness and non-redundancy, but they often suffer from high computational costs, especially for large networks. In this work, we improve the efficiency of such methods by combining bound propagation with constraint simplification. These simplifications, derived from the propagation, tighten neuron bounds and eliminate unnecessary binary variables, making the explanation process more efficient. Our experiments suggest that combining these techniques reduces explanation time by up to 89.26\%, particularly for larger neural networks. --- ### Real Money, Fake Models: Deceptive Model Claims in Shadow APIs Authors: Yage Zhang, Yukun Jiang, Zeyuan Chen, Michael Backes, Xinyue Shen, Yang Zhang Published: 2026-03-02 Categories: cs.CR, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2603.01919v3 Access to frontier large language models (LLMs), such as GPT-5 and Gemini-2.5, is often hindered by high pricing, payment barriers, and regional restrictions. These limitations drive the proliferation of $\textit{shadow APIs}$, third-party services that claim to provide access to official model services without regional limitations via indirect access. Despite their widespread use, it remains unclear whether shadow APIs deliver outputs consistent with those of the official APIs, raising concerns about the reliability of downstream applications and the validity of research findings that depend on them. In this paper, we present the first systematic audit between official LLM APIs and corresponding shadow APIs. We first identify 17 shadow APIs that have been utilized in 187 academic papers, with the most popular one reaching more than 5,900 citations and 58,000 GitHub stars by December 6, 2025. Through multidimensional auditing of three representative shadow APIs across utility, safety, and model verification, we uncover widespread behavioral inconsistency and fingerprint-based evidence consistent with deceptive model claims in a subset of audited endpoints. Specifically, we reveal performance divergence reaching up to 47.21%, significant unpredictability in safety behaviors, and identity verification failures in 45.83% of fingerprint tests. These practices critically undermine the reproducibility and validity of scientific research, harm the interests of shadow API users, and damage the reputation of official model providers. By the time of writing, 4 of the 17 providers have already ceased operations, underscoring the operational volatility of this market. Meanwhile, unverifiable compliance claims and independent model-substitution testing platforms have emerged in the ecosystem, reflecting growing community awareness of this risk. --- ### Real Money, Fake Models: Deceptive Model Claims in Shadow APIs Authors: Yage Zhang, Yukun Jiang, Zeyuan Chen, Michael Backes, Xinyue Shen, Yang Zhang Published: 2026-03-02 Categories: cs.CR, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2603.01919v2 Access to frontier large language models (LLMs), such as GPT-5 and Gemini-2.5, is often hindered by high pricing, payment barriers, and regional restrictions. These limitations drive the proliferation of $\textit{shadow APIs}$, third-party services that claim to provide access to official model services without regional limitations via indirect access. Despite their widespread use, it remains unclear whether shadow APIs deliver outputs consistent with those of the official APIs, raising concerns about the reliability of downstream applications and the validity of research findings that depend on them. In this paper, we present the first systematic audit between official LLM APIs and corresponding shadow APIs. We first identify 17 shadow APIs that have been utilized in 187 academic papers, with the most popular one reaching 5,966 citations and 58,639 GitHub stars by December 6, 2025. Through multidimensional auditing of three representative shadow APIs across utility, safety, and model verification, we uncover both indirect and direct evidence of deception practices in shadow APIs. Specifically, we reveal performance divergence reaching up to $47.21\%$, significant unpredictability in safety behaviors, and identity verification failures in $45.83\%$ of fingerprint tests. These deceptive practices critically undermine the reproducibility and validity of scientific research, harm the interests of shadow API users, and damage the reputation of official model providers. --- ### AdaPonderLM: Gated Pondering Language Models with Token-Wise Adaptive Depth Authors: Shixiang Song, He Li, Zitong Wang, Boyi Zeng, Feichen Song, Yixuan Wang, Zhiqin John Xu, Ziwei He, Zhouhan Lin Published: 2026-03-02 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.01914v2 Test-time scaling via recurrent/iterative Transformers enables large language models to spend more computation at inference, but most pretrained recurrent LMs run a fixed number of iterations, wasting compute on easy tokens and lacking token-wise adaptivity. Following the core idea of Adaptive Computation Time(ACT) and Early Exit(EE), we propose AdaPonderLM, a self-supervised recurrent language model that learns token-wise early exiting during pretraining without manually tuned per-token/per-layer pruning ratios. AdaPonderLM uses iteration-specific MLP gates with a monotonic halting mask to decide when each token stops recurring, and introduces a KV reuse mechanism that reuses cached key/value states for halted tokens, ensuring train--test consistency and practical acceleration. Across Pythia backbones from 70M to 410M (pretraining) and up to 2.8B (continued pretraining), AdaPonderLM reduces inference compute at about 10% while maintaining comparable language modeling perplexity and competitive downstream accuracy. Our analysis shows the learned gates allocate more computation to high-NLL (hard) tokens, exhibiting adaptive computation time behavior in a fully self-supervised setting. Meanwhile, under iso-FLOPs, the learned halting policy consistently outperforms fixed pruning, showing AdaPonderLM allocates compute to the right tokens rather than just reducing average depth. --- ### Zero-shot Low-Field MRI Enhancement via Diffusion-Based Adaptive Contrast Transport Authors: Muyu Liu, Chenhe Du, Xuanyu Tian, Qing Wu, Xiao Wang, Haonan Zhang, Hongjiang Wei, Yuyao Zhang Published: 2026-03-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.01913v1 Low-field (LF) magnetic resonance imaging (MRI) democratizes access to diagnostic imaging but is fundamentally limited by low signal-to-noise ratio and significant tissue contrast distortion due to field-dependent relaxation dynamics. Reconstructing high-field (HF) quality images from LF data is a blind inverse problem, severely challenged by the scarcity of paired training data and the unknown, non-linear contrast transformation operator. Existing zero-shot methods, which assume simplified linear degradation, often fail to recover authentic tissue contrast. In this paper, we propose DACT(Diffusion-Based Adaptive Contrast Transport), a novel zero-shot framework that restores HF-quality images without paired supervision. DACT synergizes a pre-trained HF diffusion prior to ensure anatomical fidelity with a physically-informed adaptive forward model. Specifically, we introduce a differentiable Sinkhorn optimal transport module that explicitly models and corrects the intensity distribution shift between LF and HF domains during the reverse diffusion process. This allows the framework to dynamically learn the intractable contrast mapping while preserving topological consistency. Extensive experiments on simulated and real clinical LF datasets demonstrate that DACT achieves state-of-the-art performance, yielding reconstructions with superior structural detail and correct tissue contrast. --- ### Demonstrating ViviDoc: Generating Interactive Documents through Human-Agent Collaboration Authors: Yinghao Tang, Yupeng Xie, Yingchaojie Feng, Tingfeng Lan, Wei Chen Published: 2026-03-02 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.01912v1 Interactive articles help readers engage with complex ideas through exploration, yet creating them remains costly, requiring both domain expertise and web development skills. Recent LLM-based agents can automate content creation, but naively applying them yields uncontrollable and unverifiable outputs. We present ViviDoc, a human-agent collaborative system that generates interactive educational documents from a single topic input. ViviDoc introduces a multi-agent pipeline (Planner, Executor, Evaluator) and the Document Specification (DocSpec), a human-readable intermediate representation that decomposes each interactive visualization into State, Render, Transition, and Constraint components. The DocSpec enables educators to review and refine generation plans before code is produced, bridging the gap between pedagogical intent and executable output. Expert evaluation and a user study show that ViviDoc substantially outperforms naive agentic generation and provides an intuitive editing experience. Our project homepage is available at https://vividoc-homepage.vercel.app/. --- ### FLANS at SemEval-2026 Task 7: RAG with Open-Sourced Smaller LLMs for Everyday Knowledge Across Diverse Languages and Cultures Authors: Liliia Bogdanova, Shiran Sun, Lifeng Han, Natalia Amat Lefort, Flor Miriam Plaza-del-Arco Published: 2026-03-02 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.01910v1 This system paper describes our participation in the SemEval-2025 Task-7 ``Everyday Knowledge Across Diverse Languages and Cultures''. We attended two subtasks, i.e., Track 1: Short Answer Questions (SAQ), and Track 2: Multiple-Choice Questions (MCQ). The methods we used are retrieval augmented generation (RAGs) with open-sourced smaller LLMs (OS-sLLMs). To better adapt to this shared task, we created our own culturally aware knowledge base (CulKBs) by extracting Wikipedia content using keyword lists we prepared. We extracted both culturally-aware wiki-text and country-specific wiki-summary. In addition to the local CulKBs, we also have one system integrating live online search output via DuckDuckGo. Towards better privacy and sustainability, we aimed to deploy smaller LLMs (sLLMs) that are open-sourced on the Ollama platform. We share the prompts we developed using refinement techniques and report the learning curve of such prompts. The tested languages are English, Spanish, and Chinese for both tracks. Our resources and codes are shared via https://github.com/aaronlifenghan/FLANS-2026 --- ### Efficient RLVR Training via Weighted Mutual Information Data Selection Authors: Xinyu Zhou, Boyu Zhu, Haotian Zhang, Huiming Wang, Zhijiang Guo Published: 2026-03-02 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2603.01907v1 Reinforcement learning (RL) plays a central role in improving the reasoning and alignment of large language models, yet its efficiency critically depends on how training data are selected. Existing online selection strategies predominantly rely on difficulty-based heuristics, favouring datapoints with intermediate success rates, implicitly equating difficulty with informativeness and neglecting epistemic uncertainty arising from limited evidence. We introduce InSight, an INformation-guided data SamplInG metHod for RL Training, grounded in a weighted mutual information objective. By modeling data outcomes with Bayesian latent success rates, we show that expected uncertainty reduction decomposes into complementary difficulty- and evidence-dependent components, revealing a fundamental limitation of difficulty-only selection. Leveraging this observation, InSight constructs a stable acquisition score based on the mean belief of datapoints' success rather than noisy sampled outcomes, and naturally extends to multi-rollout settings common in reinforcement learning with verifiable rewards (RLVR). Extensive experiments demonstrate that InSight consistently achieves state-of-the-art performance and improves training efficiency, including a +1.41 average gain on Planning & Mathmatics benchmarks, +1.01 improvement on general reasoning, and up to ~2.2x acceleration, with negligible additional computational overhead. --- ### Agentic Code Reasoning Authors: Shubham Ugare, Satish Chandra Published: 2026-03-02 Categories: cs.SE, cs.AI, cs.PL arXiv: https://arxiv.org/abs/2603.01896v2 Can LLM agents explore codebases and reason about code semantics without executing the code? We study this capability, which we call agentic code reasoning, and introduce semi-formal reasoning: a structured prompting methodology that requires agents to construct explicit premises, trace execution paths, and derive formal conclusions. Unlike unstructured chain-of-thought, semi-formal reasoning acts as a certificate: the agent cannot skip cases or make unsupported claims. We evaluate across three tasks (patch equivalence verification, fault localization, and code question answering) and show that semi-formal reasoning consistently improves accuracy on all of them. For patch equivalence, accuracy improves from 78% to 88% on curated examples and reaches 93% on real-world agent-generated patches, approaching the reliability needed for execution-free RL reward signals. For code question answering on RubberDuckBench Mohammad et al. (2026), semi-formal reasoning achieves 87% accuracy. For fault localization on Defects4J Just et al. (2014), semi-formal reasoning improves Top-5 accuracy by 5 percentage points over standard reasoning. These results demonstrate that structured agentic reasoning enables meaningful semantic code analysis without execution, opening practical applications in RL training pipelines, code review, and static program analysis. --- ### Generative Visual Chain-of-Thought for Image Editing Authors: Zijin Yin, Tiankai Hang, Yiji Cheng, Shiyi Zhang, Runze He, Yu Xu, Chunyu Wang, Bing Li, Zheng Chang, Kongming Liang, Qinglin Lu, Zhanyu Ma Published: 2026-03-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.01893v2 Existing image editing methods struggle to perceive where to edit, especially under complex scenes and nuanced spatial instructions. To address this issue, we propose Generative Visual Chain-of-Thought (GVCoT), a unified framework that performs native visual reasoning by first generating spatial cues to localize the target region and then executing the edit. Unlike prior text-only CoT or tool-dependent visual CoT paradigms, GVCoT jointly optimizes visual tokens generated during the reasoning and editing phases in an end-to-end manner. This way fosters the emergence of innate spatial reasoning ability and enables more effective utilization of visual-domain cues. The main challenge of training GCVoT lies in the scarcity of large-scale editing data with precise edit region annotations; to this end, we construct GVCoT-Edit-Instruct, a dataset of 1.8M high-quality samples spanning 19 tasks. We adopt a progressive training strategy: supervised fine-tuning to build foundational localization ability in reasoning trace before final editing, followed by reinforcement learning to further improve reasoning and editing quality. Finally, we introduce SREdit-Bench, a new benchmark designed to comprehensively stress-test models under sophisticated scenes and fine-grained referring expressions. Experiments demonstrate that GVCoT consistently outperforms state-of-the-art models on SREdit-Bench and ImgEdit. We hope our GVCoT will inspire future research toward interpretable and precise image editing. --- ### SEAR: Sample Efficient Action Chunking Reinforcement Learning Authors: C. F. Maximilian Nagy, Onur Celik, Emiliyan Gospodinov, Florian Seligmann, Weiran Liao, Aryan Kaushik, Gerhard Neumann Published: 2026-03-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.01891v2 Action chunking improves exploration and accelerates value propagation in long-horizon reinforcement learning, but naively applying off-policy methods to the temporally extended action space at reduced decision frequency offsets these gains, leading to poor sample efficiency. Existing online action chunking methods address these issues through computationally expensive critic-only approaches or by relying on offline data. We introduce SEAR, a sample-efficient off-policy algorithm that enables online reinforcement learning with action chunks by addressing both challenges. To handle high-dimensional action sequences, SEAR employs a causal transformer critic trained with multi-horizon targets that provide a training signal for every prefix of an action chunk, effectively increasing the useful gradients per sample. To restore decision frequency, SEAR operates with a receding horizon and random replanning, ensuring uniform state coverage while combining the fast value propagation of large chunks with the reactivity of small ones. SEAR outperforms state-of-the-art online reinforcement learning methods including SimbaV2 on challenging Metaworld manipulation tasks. Beyond online RL, applying SEAR to the offline-to-online method QC improves its performance on the demanding OGBench cube-triple tasks. --- ### SEAR: Sample Efficient Action Chunking Reinforcement Learning Authors: C. F. Maximilian Nagy, Onur Celik, Emiliyan Gospodinov, Florian Seligmann, Weiran Liao, Aryan Kaushik, Gerhard Neumann Published: 2026-03-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.01891v1 Action chunking can improve exploration and value estimation in long horizon reinforcement learning, but makes learning substantially harder since the critic must evaluate action sequences rather than single actions, greatly increasing approximation and data efficiency challenges. As a result, existing action chunking methods, primarily designed for the offline and offline-to-online settings, have not achieved strong performance in purely online reinforcement learning. We introduce SEAR, an off policy online reinforcement learning algorithm for action chunking. It exploits the temporal structure of action chunks and operates with a receding horizon, effectively combining the benefits of small and large chunk sizes. SEAR outperforms state of the art online reinforcement learning methods on Metaworld, training with chunk sizes up to 20. --- ### Resolving Blind Inverse Problems under Dynamic Range Compression via Structured Forward Operator Modeling Authors: Muyu Liu, Xuanyu Tian, Chenhe Du, Qing Wu, Hongjiang Wei, Yuyao Zhang Published: 2026-03-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.01890v1 Recovering radiometric fidelity from unknown dynamic range compression (UDRC), such as low-light enhancement and HDR reconstruction, is a challenging blind inverse problem, due to the unknown forward model and irreversible information loss introduced by compression. To address this challenge, we first identify monotonicity as the fundamental physical invariant shared across UDRC tasks. Leveraging this insight, we introduce the \textbf{cascaded monotonic Bernstein} (CaMB) operator to parameterize the unknown forward model. CaMB enforces monotonicity as a hard architectural inductive bias, constraining optimization to physically consistent mappings and enabling robust and stable operator estimation. We further integrate CaMB with a plug-and-play diffusion framework, proposing \textbf{CaMB-Diff}. Within this framework, the diffusion model serves as a powerful geometric prior for structural and semantic recovery, while CaMB explicitly models and corrects radiometric distortions through a physically grounded forward operator. Extensive experiments on a variety of zero-shot UDRC tasks, including low-light enhancement, low-field MRI enhancement, and HDR reconstruction, demonstrate that CaMB-Diff significantly outperforms state-of-the-art zero-shot baselines in terms of both signal fidelity and physical consistency. Moreover, we empirically validate the effectiveness of the proposed CaMB parameterization in accurately modeling the unknown forward operator. --- ### Diagnosing Generalization Failures from Representational Geometry Markers Authors: Chi-Ning Chou, Artem Kirsanov, Yao-Yuan Yang, SueYeon Chung Published: 2026-03-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.01879v1 Generalization, the ability to perform well beyond the training context, is a hallmark of biological and artificial intelligence, yet anticipating unseen failures remains a central challenge. Conventional approaches often take a ``bottom-up'' mechanistic route by reverse-engineering interpretable features or circuits to build explanatory models. While insightful, these methods often struggle to provide the high-level, predictive signals for anticipating failure in real-world deployment. Here, we propose using a ``top-down'' approach to studying generalization failures inspired by medical biomarkers: identifying system-level measurements that serve as robust indicators of a model's future performance. Rather than mapping out detailed internal mechanisms, we systematically design and test network markers to probe structure, function links, identify prognostic indicators, and validate predictions in real-world settings. In image classification, we find that task-relevant geometric properties of in-distribution (ID) object manifolds consistently forecast poor out-of-distribution (OOD) generalization. In particular, reductions in two geometric measures, effective manifold dimensionality and utility, predict weaker OOD performance across diverse architectures, optimizers, and datasets. We apply this finding to transfer learning with ImageNet-pretrained models. We consistently find that the same geometric patterns predict OOD transfer performance more reliably than ID accuracy. This work demonstrates that representational geometry can expose hidden vulnerabilities, offering more robust guidance for model selection and AI interpretability. --- ### CTForensics: A Comprehensive Dataset and Method for AI-Generated CT Image Detection Authors: Yiheng Li, Zichang Tan, Guoqing Xu, Yichun Yeh, Yang Yang, Zhen Lei Published: 2026-03-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.01878v2 Recent advances in generative AI have made synthetic Computed Tomography (CT) images increasingly realistic, enabling promising applications in medical data augmentation while raising serious concerns about clinical safety and data trustworthiness. Detecting AI-generated CT images remains challenging for two key reasons: existing benchmarks cover only limited generation sources, and many detectors are adapted from natural-image forensics without explicitly modeling CT-specific imaging properties. In this paper, we introduce CTForensics, a dataset for detecting AI-generated CT images. CTForensics contains 75,990 2D CT images, including a dedicated test benchmark of 29,990 balanced authentic and generated samples from ten representative CT generative models spanning GAN-based and diffusion-based paradigms. We further propose the Enhanced Spatial-Frequency CT Forgery Detector (ESF-CTFD), a CT-oriented CNN framework built around a Wavelet-Enhanced Central Stem, Multi-Scale Spatial Aggregation, and a Frequency-Aware Prediction Block. The Wavelet-Enhanced Central Stem enhances local intensity correlations and high-frequency residuals, Multi-Scale Spatial Aggregation aligns anatomical features across resolutions with lightweight residual units, and the Frequency-Aware Prediction Block models global spectral artifacts. Extensive experiments on CTForensics show that ESF-CTFD achieves 96.01% mAcc and 99.96% mAP, outperforming existing methods and maintaining strong robustness under realistic perturbations with only a 0.99% average drop. Codes will be available at https://github.com/liyih/CTForensics. --- ### KDFlow: A User-Friendly and Efficient Knowledge Distillation Framework for Large Language Models Authors: Songming Zhang, Xue Zhang, Tong Zhang, Bojie Hu, Yufeng Chen, Jinan Xu Published: 2026-03-02 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.01875v4 Knowledge distillation (KD) is widely used to compress and post-train large language models (LLMs), yet many existing frameworks execute teacher inference with the same training-oriented backend as student optimization, leading to suboptimal efficiency. In this paper, we propose KDFlow, a novel framework for LLM distillation that features a decoupled architecture and employs SGLang for teacher inference. KDFlow combines SGLang for teacher inference with PyTorch FSDP2 for student optimization, allowing each model to run on a backend tailored to its workload. To enable efficient full-vocabulary distillation in this decoupled architecture, KDFlow transfers the teacher's final hidden states via Ray's object store and recomputes teacher logits on each student worker using a frozen copy of the teacher's output head. Furthermore, our framework supports both off-policy and on-policy distillation and incorporates cross-tokenizer algorithms through highly extensible and user-friendly APIs. Experiments show that KDFlow achieves a 1.44$\times$ to 6.36$\times$ speedup over MS-SWIFT in off-policy distillation and a 1.43$\times$ to 1.75$\times$ speedup over verl in on-policy distillation. KDFlow further scales to 64 GPUs, achieving 3.68$\times$ and 2.52$\times$ strong-scaling speedups in two representative model configurations. The code and documentation are publicly available. --- ### KDFlow: A User-Friendly and Efficient Knowledge Distillation Framework for Large Language Models Authors: Songming Zhang, Xue Zhang, Tong Zhang, Bojie Hu, Yufeng Chen, Jinan Xu Published: 2026-03-02 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.01875v3 Knowledge distillation (KD) is an essential technique to compress large language models (LLMs) into smaller ones. However, despite the distinct roles of the student model and the teacher model in KD, most existing frameworks still use a homogeneous training backend (e.g., FSDP and DeepSpeed) for both models, leading to suboptimal training efficiency. In this paper, we present a novel framework for LLM distillation, termed \textbf{KDFlow}, which features a decoupled architecture and employs SGLang for teacher inference. By bridging the training efficiency of FSDP2 and the inference efficiency of SGLang, KDFlow achieves full utilization of both advantages in a unified system. Moreover, instead of transferring full logits across different processes, our framework only transmits the teacher's hidden states using zero-copy data transfer and recomputes the logits on the student side, effectively balancing the communication cost and KD performance. Furthermore, our framework supports both off-policy and on-policy distillation and incorporates KD algorithms for cross-tokenizer KD through highly extensible and user-friendly APIs. Experiments show that KDFlow can achieve \textbf{1.44$\times$ to 6.36$\times$} speedup compared to current KD frameworks, enabling researchers to rapidly prototype and scale LLM distillation with minimal engineering overhead. Code is available at: https://github.com/songmzhang/KDFlow --- ### Phishing the Phishers with SpecularNet: Hierarchical Graph Autoencoding for Reference-Free Web Phishing Detection Authors: Tailai Song, Pedro Casas, Michela Meo Published: 2026-03-02 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2603.01874v1 Phishing remains the most pervasive threat to the Web, enabling large-scale credential theft and financial fraud through deceptive webpages. While recent reference-based and generative-AI-driven phishing detectors achieve strong accuracy, their reliance on external knowledge bases, cloud services, and complex multimodal pipelines fundamentally limits practicality, scalability, and reproducibility. In contrast, conventional deep learning approaches often fail to generalize to evolving phishing campaigns. We introduce SpecularNet, a novel lightweight framework for reference-free web phishing detection that demonstrates how carefully designed compact architectures can rival heavyweight systems. SpecularNet operates solely on the domain name and HTML structure, modeling the Document Object Model (DOM) as a tree and leveraging a hierarchical graph autoencoding architecture with directional, level-wise message passing. This design captures higher-order structural invariants of phishing webpages while enabling fast, end-to-end inference on standard CPUs. Extensive evaluation against 13 state of the art phishing detectors, including leading reference-based systems, shows that SpecularNet achieves competitive detection performance with dramatically lower computational cost. On benchmark datasets, it reaches an F1 score of 93.9%, trailing the best reference-based method slightly while reducing inference time from several seconds to approximately 20 milliseconds per webpage. Field and robustness evaluations further validate SpecularNet in real-world deployments, on a newly collected 2026 open-world dataset, and against adversarial attacks. --- ### Generalizing Logic-based Explanations for Machine Learning Classifiers via Optimization Authors: Francisco Mateus Rocha Filho, Ajalmar Rêgo da Rocha Neto, Thiago Alves Rocha Published: 2026-03-02 Categories: cs.LO, cs.LG arXiv: https://arxiv.org/abs/2603.01870v1 Machine learning models support decision-making, yet the reasons behind their predictions are opaque. Clear and reliable explanations help users make informed decisions and avoid blindly trusting model outputs. However, many existing explanation methods fail to guarantee correctness. Logic-based approaches ensure correctness but often offer overly constrained explanations, limiting coverage. Recent work addresses this by incrementally expanding explanations while maintaining correctness. This process is performed separately for each feature, adjusting both its upper and lower bounds. However, this approach faces a trade-off: smaller increments incur high computational costs, whereas larger ones may lead to explanations covering fewer instances. To overcome this, we propose two novel methods. Onestep builds upon this prior work, generating explanations in a single step for each feature and each bound, eliminating the overhead of an iterative process. \textit{Twostep} takes a gradual approach, improving coverage. Experimental results show that Twostep significantly increases explanation coverage (by up to 72.60\% on average across datasets) compared to Onestep and, consequently, to prior work. --- ### Sovereign AI-based Public Services are Viable and Affordable Authors: António Branco, Luís Gomes, Rodrigo Santos, Eduardo Santos, João Silva, Nuno Marques, Madalena Rodrigues Published: 2026-03-02 Categories: cs.CL, cs.CY arXiv: https://arxiv.org/abs/2603.01869v1 The rapid expansion of AI-based remote services has intensified debates about the long-term implications of growing structural concentration in infrastructure and expertise. As AI capabilities become increasingly intertwined with geopolitical interests, the availability and reliability of foundational AI services can no longer be taken for granted. This issue is particularly pressing for AI-enabled public services for citizens, as governments and public agencies are progressively adopting 24/7 AI-driven support systems typically operated through commercial offerings from a small oligopoly of global technology providers. This paper challenges the prevailing assumption that general-purpose architectures, offered by these providers, are the optimal choice for all application contexts. Through practical experimentation, we demonstrate that viable and cost-effective alternatives exist. Alternatives that align with principles of digital and cultural sovereignty. Our findings provide an empirical illustration that sovereign AI-based public services are both technically feasible and economically sustainable, capable of operating effectively on premises with modest computational and financial resources while maintaining cultural and digital autonomy. The technical insights and deployment lessons reported here are intended to inform the adoption of similar sovereign AI public services by national agencies and governments worldwide. --- ### CyclicJudge: Mitigating Judge Bias Efficiently in LLM-based Evaluation Authors: Ziyi Zhu, Olivier Tieleman, Alexey Bukhtiyarov, Jinghong Chen Published: 2026-03-02 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.01865v3 LLM-as-judge evaluation has become standard practice for open-ended model assessment; however, judges exhibit systematic biases that cannot be averaged out by increasing the number of scenarios or generations. These biases are often similar in magnitude to the model differences that benchmarks are designed to detect, resulting in unreliable rankings when single-judge evaluations are used. We introduce a variance decomposition that partitions benchmark score variance into scenario, generation, judge, and residual components. Based on this analysis, CyclicJudge, a round-robin assignment of judges to scenarios, is demonstrated to be the optimal strategy for a fixed judge panel and judge-call budget: the score recovers the panel mean exactly while matching the cost of single-judge evaluation. Empirical results on MT-Bench and MindEval validate the effectiveness of CyclicJudge as predicted, across both general-purpose and domain-specific evaluation settings. --- ### Streaming Real-Time Trajectory Prediction Using Endpoint-Aware Modeling Authors: Alexander Prutsch, David Schinagl, Horst Possegger Published: 2026-03-02 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2603.01864v1 Future trajectories of neighboring traffic agents have a significant influence on the path planning and decision-making of autonomous vehicles. While trajectory forecasting is a well-studied field, research mainly focuses on snapshot-based prediction, where each scenario is treated independently of its global temporal context. However, real-world autonomous driving systems need to operate in a continuous setting, requiring real-time processing of data streams with low latency and consistent predictions over successive timesteps. We leverage this continuous setting to propose a lightweight yet highly accurate streaming-based trajectory forecasting approach. We integrate valuable information from previous predictions with a novel endpoint-aware modeling scheme. Our temporal context propagation uses the trajectory endpoints of the previous forecasts as anchors to extract targeted scenario context encodings. Our approach efficiently guides its scene encoder to extract highly relevant context information without needing refinement iterations or segment-wise decoding. Our experiments highlight that our approach effectively relays information across consecutive timesteps. Unlike methods using multi-stage refinement processing, our approach significantly reduces inference latency, making it well-suited for real-world deployment. We achieve state-of-the-art streaming trajectory prediction results on the Argoverse~2 multi-agent and single-agent benchmarks, while requiring substantially fewer resources. --- ### Tide: A Customisable Dataset Generator for Anti-Money Laundering Research Authors: Montijn van den Beukel, Jože Martin Rožanec, Ana-Lucia Varbanescu Published: 2026-03-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.01863v1 The lack of accessible transactional data significantly hinders machine learning research for Anti-Money Laundering (AML). Privacy and legal concerns prevent the sharing of real financial data, while existing synthetic generators focus on simplistic structural patterns and neglect the temporal dynamics (timing and frequency) that characterise sophisticated laundering schemes. We present Tide, an open-source synthetic dataset generator that produces graph-based financial networks incorporating money laundering patterns defined by both structural and temporal characteristics. Tide enables reproducible, customisable dataset generation tailored to specific research needs. We release two reference datasets with varying illicit ratios (LI: 0.10\%, HI: 0.19\%), alongside the implementation of state-of-the-art detection models. Evaluation across these datasets reveals condition-dependent model rankings: LightGBM achieves the highest PR-AUC (78.05) in the low illicit ratio condition, while XGBoost performs best (85.12) at higher fraud prevalence. These divergent rankings demonstrate that the reference datasets can meaningfully differentiate model capabilities across operational conditions. Tide provides the research community with a configurable benchmark that exposes meaningful performance variation across model architectures, advancing the development of robust AML detection methods. --- ### Let the Agent Search: Autonomous Exploration Beats Rigid Workflows in Temporal Question Answering Authors: Xufei Lv, Jiahui Yang, Haoyuan Sun, Xialin Su, Zhiliang Tian, Yifu Gao, Linbo Qiao, Houde Liu Published: 2026-03-02 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.01853v2 Temporal Knowledge Graph Question Answering (TKGQA) is challenging because it requires multi-hop reasoning under complex temporal constraints. Recent LLM-based approaches have improved semantic modeling for this task, but many still rely on fixed reasoning workflows or costly post-training, which can limit adaptability and make error recovery difficult. We show that enabling an off-the-shelf Large Language Model (LLM) to determine its next action is already effective in a zero-shot setting. Based on this insight, we propose AT2QA, an Autonomous and Training-free Agent for TKG Question Answering. AT2QA empowers the LLM to iteratively interact with the TKG via a generic search tool, inherently enabling autonomous exploration and dynamic self-correction during reasoning. To further elicit the LLM's potential for complex temporal reasoning, we introduce a training-free experience mining mechanism that distills a compact few-shot demonstration library from successful self-generated trajectories. AT2QA also yields a transparent audit trail for every prediction. Experiments on three challenging benchmarks -- MultiTQ, Timeline-CronQuestion, and Timeline-ICEWS-Actor -- show that AT2QA achieves new state-of-the-art performance, surpassing the strongest baselines by 10.7, 4.9, and 11.2 absolute points, respectively. Our code is available at https://github.com/AT2QA-Official-Code/AT2QA-Official-Code --- ### Tiny-DroNeRF: Tiny Neural Radiance Fields aboard Federated Learning-enabled Nano-drones Authors: Ilenia Carboni, Elia Cereda, Lorenzo Lamberti, Daniele Malpetti, Francesco Conti, Daniele Palossi Published: 2026-03-02 Categories: cs.RO, cs.CV, eess.SY arXiv: https://arxiv.org/abs/2603.01850v1 Sub-30g nano-sized aerial robots can leverage their agility and form factor to autonomously explore cluttered and narrow environments, like in industrial inspection and search and rescue missions. However, the price for their tiny size is a strong limit in their resources, i.e., sub-100 mW microcontroller units (MCUs) delivering $\sim$100 GOps/s at best, and memory budgets well below 100 MB. Despite these strict constraints, we aim to enable complex vision-based tasks aboard nano-drones, such as dense 3D scene reconstruction: a key robotic task underlying fundamental capabilities like spatial awareness and motion planning. Top-performing 3D reconstruction methods leverage neural radiance fields (NeRF) models, which require GBs of memory and massive computation, usually delivered by high-end GPUs consuming 100s of Watts. Our work introduces Tiny-DroNeRF, a lightweight NeRF model, based on Instant-NGP, and optimized for running on a GAP9 ultra-low-power (ULP) MCU aboard our nano-drones. Then, we further empower our Tiny-DroNeRF by leveraging a collaborative federated learning scheme, which distributes the model training among multiple nano-drones. Our experimental results show a 96% reduction in Tiny-DroNeRF's memory footprint compared to Instant-NGP, with only a 5.7 dB drop in reconstruction accuracy. Finally, our federated learning scheme allows Tiny-DroNeRF to train with an amount of data otherwise impossible to keep in a single drone's memory, increasing the overall reconstruction accuracy. Ultimately, our work combines, for the first time, NeRF training on an ULP MCU with federated learning on nano-drones. --- ### GroupEnsemble: Efficient Uncertainty Estimation for DETR-based Object Detection Authors: Yutong Yang, Katarina Popović, Julian Wiederer, Markus Braun, Vasileios Belagiannis, Bin Yang Published: 2026-03-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.01847v1 Detection Transformer (DETR) and its variants show strong performance on object detection, a key task for autonomous systems. However, a critical limitation of these models is that their confidence scores only reflect semantic uncertainty, failing to capture the equally important spatial uncertainty. This results in an incomplete assessment of the detection reliability. On the other hand, Deep Ensembles can tackle this by providing high-quality spatial uncertainty estimates. However, their immense memory consumption makes them impractical for real-world applications. A cheaper alternative, Monte Carlo (MC) Dropout, suffers from high latency due to the need of multiple forward passes during inference to estimate uncertainty. To address these limitations, we introduce GroupEnsemble, an efficient and effective uncertainty estimation method for DETR-like models. GroupEnsemble simultaneously predicts multiple individual detection sets by feeding additional diverse groups of object queries to the transformer decoder during inference. Each query group is transformed by the shared decoder in isolation and predicts a complete detection set for the same input. An attention mask is applied to the decoder to prevent inter-group query interactions, ensuring each group detects independently to achieve reliable ensemble-based uncertainty estimation. By leveraging the decoder's inherent parallelism, GroupEnsemble efficiently estimates uncertainty in a single forward pass without sequential repetition. We validated our method under autonomous driving scenes and common daily scenes using the Cityscapes and COCO datasets, respectively. The results show that a hybrid approach combining MC-Dropout and GroupEnsemble outperforms Deep Ensembles on several metrics at a fraction of the cost. The code is available at https://github.com/yutongy98/GroupEnsemble. ---