# Research Papers — 2026-03 · bundle page 95 of 243 > Full-text markdown of papers #4701–4750 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/95 [← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/94.md) [Next bundle →](https://aipod.dev/papers/archive/2026-03/page/96.md) ### Language Model Maps for Prompt-Response Distributions via Log-Likelihood Vectors Authors: Yusuke Takase, Momose Oyama, Hidetoshi Shimodaira Published: 2026-03-19 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.18593v1 We propose a method that represents language models by log-likelihood vectors over prompt-response pairs and constructs model maps for comparing their conditional distributions. In this space, distances between models approximate the KL divergence between the corresponding conditional distributions. Experiments on a large collection of publicly available language models show that the maps capture meaningful global structure, including relationships to model attributes and task performance. The method also captures systematic shifts induced by prompt modifications and their approximate additive compositionality, suggesting a way to analyze and predict the effects of composite prompt operations. We further introduce pointwise mutual information (PMI) vectors to reduce the influence of unconditional distributions; in some cases, PMI-based model maps better reflect training-data-related differences. Overall, the framework supports the analysis of input-dependent model behavior. --- ### A Novel FACS-Aligned Anatomical Text Description Paradigm for Fine-Grained Facial Behavior Synthesis Authors: Jiahe Wang, Cong Liang, Xuandong Huang, Yuxin Wang, Xin Yun, Yi Wu, Yanan Chang, Shangfei Wang Published: 2026-03-19 Categories: cs.CV, cs.MM arXiv: https://arxiv.org/abs/2603.18588v2 Facial behavior constitutes the primary medium of human nonverbal communication. Existing synthesis methods predominantly follow two paradigms: coarse emotion category labels or one-hot Action Unit (AU) vectors from the Facial Action Coding System (FACS). Neither paradigm reliably renders fine-grained facial behaviors nor resolves anatomically implausible artifacts caused by conflicting AUs. Therefore, we propose a novel task paradigm: anatomically grounded facial behavior synthesis from FACS-based AU descriptions. This paradigm explicitly encodes FACS-defined muscle movement rules, inter-AU interactions, and conflict resolution mechanisms into natural language control signals. To enable systematic research, we develop a dynamic AU text processor, a FACS rule-based module that converts raw AU annotations into anatomically consistent natural language descriptions. Using this processor, we construct BP4D-AUText, the first large-scale text-image paired dataset for fine-grained facial behavior synthesis, comprising over 302K high-quality samples. Given that existing general semantic consistency metrics cannot capture the alignment between anatomical facial descriptions and synthesized muscle movements, we propose the Alignment Accuracy of AU Probability Distributions (AAAD), a task-specific metric that quantifies semantic consistency. Finally, we design VQ-AUFace, a robust baseline framework incorporating anatomical priors and progressive cross-modal alignment, to validate the paradigm. Extensive quantitative experiments and user studies demonstrate the paradigm significantly outperforms state-of-the-art methods, particularly in challenging conflicting AU scenarios, achieving superior anatomical fidelity, semantic consistency, and visual quality. --- ### Color image restoration based on nonlocal saturation-value similarity Authors: Wei Wang, Yakun Li Published: 2026-03-19 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.18586v1 In this paper, we propose and develop a novel nonlocal variational technique based on saturation-value similarity for color image restoration. In traditional nonlocal methods, image patches are extracted from red, green and blue channels of a color image directly, and the color information can not be described finely because the patch similarity is mainly based on the grayscale value of independent channel. The main aim of this paper is to propose and develop a novel nonlocal regularization method by considering the similarity of image patches in saturation-value channel of a color image. In particular, we first establish saturation-value similarity based nonlocal total variation by incorporating saturation-value similarity of color image patches into the proposed nonlocal gradients, which can describe the saturation and value similarity of two adjacent color image patches. The proposed nonlocal variational models are then formulated based on saturation-value similarity based nonlocal total variation. Moreover, we design an effective and efficient algorithm to solve the proposed optimization problem numerically by employing bregmanized operator splitting method, and we also study the convergence of the proposed algorithms. Numerical examples are presented to demonstrate that the performance of the proposed models is better than that of other testing methods in terms of visual quality and some quantitative metrics including peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), quaternion structural similarity index (QSSIM) and S-CIELAB color error. --- ### HAViT: Historical Attention Vision Transformer Authors: Swarnendu Banik, Manish Das, Shiv Ram Dubey, Satish Kumar Singh Published: 2026-03-19 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.18585v1 Vision Transformers have excelled in computer vision but their attention mechanisms operate independently across layers, limiting information flow and feature learning. We propose an effective cross-layer attention propagation method that preserves and integrates historical attention matrices across encoder layers, offering a principled refinement of inter-layer information flow in Vision Transformers. This approach enables progressive refinement of attention patterns throughout the transformer hierarchy, enhancing feature acquisition and optimization dynamics. The method requires minimal architectural changes, adding only attention matrix storage and blending operations. Comprehensive experiments on CIFAR-100 and TinyImageNet demonstrate consistent accuracy improvements, with ViT performance increasing from 75.74% to 77.07% on CIFAR-100 (+1.33%) and from 57.82% to 59.07% on TinyImageNet (+1.25%). Cross-architecture validation shows similar gains across transformer variants, with CaiT showing 1.01% enhancement. Systematic analysis identifies the blending hyperparameter of historical attention (alpha = 0.45) as optimal across all configurations, providing the ideal balance between current and historical attention information. Random initialization consistently outperforms zero initialization, indicating that diverse initial attention patterns accelerate convergence and improve final performance. Our code is publicly available at https://github.com/banik-s/HAViT. --- ### Breaking Hard Isomorphism Benchmarks with DRESS Authors: Eduar Castrillo Velilla Published: 2026-03-19 Categories: cs.DS, cs.DM, cs.LG arXiv: https://arxiv.org/abs/2603.18582v2 DRESS is a deterministic, parameter-free framework for structural graph refinement that iteratively refines the structural similarity of edges in a graph to produce a canonical fingerprint: a real-valued edge vector, obtained by converging a nonlinear dynamical system to its unique fixed point. $Δ$-DRESS is a member of the DRESS family of graph fingerprints that applies a single level of vertex deletion. We test it on a benchmark of 51,813 distinct graphs across 34 hard families, including the complete Spence collection of strongly regular graphs (43,703 SRGs, 12 families), four additional SRG families (8,015 graphs), and 18 classical hard constructions (102 family entries corresponding to 99 distinct graphs). $Δ$-DRESS produces unique fingerprints in 33 of 34 benchmark families at $k=1$, resolving all but one within-family collision among over 576 million non-isomorphic pairs. One genuine collision exists at deletion depth $k=1$, between two vertex-transitive SRGs in SRG(40,12,2,4), which is resolved by a single-step fallback to $Δ^2$-DRESS. For every family with pairwise-comparable full sorted-multiset fingerprints, the minimum observed separation margin remains at least $137 \times ε$, confirming that the reported separations are numerically robust and not artifacts of the convergence threshold. We also show that $Δ$-DRESS separates the Rook $L_2(4)$/Shrikhande pair, proving it escapes the theoretical boundary of 3-WL. The method runs in $\mathcal{O}(n \cdot I \cdot m \cdot d_{\max})$ time per graph. --- ### WarPGNN: A Parametric Thermal Warpage Analysis Framework with Physics-aware Graph Neural Network Authors: Haotian Lu, Jincong Lu, Sachin Sachdeva, Sheldon X. -D. Tan Published: 2026-03-19 Categories: cs.AR, cs.LG, eess.SY arXiv: https://arxiv.org/abs/2603.18581v2 With the advent of system-in-package (SiP) chiplet-based design and heterogeneous 2.5D/3D integration, thermal-induced warpage has become a critical reliability concern. While conventional numerical approaches can deliver highly accurate results, they often incur prohibitively high computational costs, limiting their scalability for complex chiplet-package systems. In this paper, we present WarPGNN, an efficient and accurate parametric thermal warpage analysis framework powered by Graph Neural Networks (GNNs). By operating directly on graphs constructed from the floorplans, WarPGNN enables fast warpage-aware floorplan exploration and exhibits strong transferability across diverse package configurations. Our method first encodes multi-die floorplans into reduced Transitive Closure Graphs (rTCGs), then a Graph Convolution Network (GCN)-based encoder extracts hierarchical structural features, followed by a U-Net inspired decoder that reconstructs warpage maps from graph feature embeddings. Furthermore, to address the long-tailed pattern of warpage data distribution, we developed a physics-informed loss and revised a message-passing encoder based on Graph Isomorphic Network (GIN) that further enhance learning performance for extreme cases and expressiveness of graph embeddings. Numerical results show that WarPGNN achieves more than 205.91x speedup compared with the 2-D efficient FEM-based method and over 119766.64x acceleration with 3-D FEM method COMSOL, respectively, while maintaining comparable accuracy at only 1.26% full-scale normalized RMSE and 2.21% warpage value error. Compared with recent DeepONet-based model, our method achieved comparable prediction accuracy and inference speedup with 3.4x lower training time. In addition, WarPGNN demonstrates remarkable transferability on unseen datasets with up to 3.69% normalized RMSE and similar runtime. --- ### ICE: Intervention-Consistent Explanation Evaluation with Statistical Grounding for LLMs Authors: Abhinaba Basu, Pavan Chakraborty Published: 2026-03-19 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.18579v2 Evaluating whether explanations faithfully reflect a model's reasoning remains an open problem. Existing benchmarks use single interventions without statistical testing, making it impossible to distinguish genuine faithfulness from chance-level performance. We show that faithfulness is not a fixed property but an operator-dependent quantity that changes with the intervention method used to measure it. We introduce ICE (Intervention-Consistent Explanation), a framework that evaluates explanations against random baselines of equal size under multiple operators. Evaluating 7 LLMs across 4 tasks with deletion and retrieval infill operators, we find that switching operators crosses the positive-evidence threshold in 18% of configurations (5 of 28 attention comparisons), with gaps reaching 44 percentage points. Randomized baselines detect anti-faithfulness (explanations worse than random) in nearly one-third of English deletion configurations, invisible without random comparisons. These patterns persist across 6 non-English languages and 2 attribution methods. The methodology generalizes to step-level chain-of-thought evaluation, where preliminary results on 3 frontier models suggest that high accuracy does not imply faithful reasoning. --- ### ICE: Intervention-Consistent Explanation Evaluation with Statistical Grounding for LLMs Authors: Abhinaba Basu, Pavan Chakraborty Published: 2026-03-19 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.18579v1 Evaluating whether explanations faithfully reflect a model's reasoning remains an open problem. Existing benchmarks use single interventions without statistical testing, making it impossible to distinguish genuine faithfulness from chance-level performance. We introduce ICE (Intervention-Consistent Explanation), a framework that compares explanations against matched random baselines via randomization tests under multiple intervention operators, yielding win rates with confidence intervals. Evaluating 7 LLMs across 4 English tasks, 6 non-English languages, and 2 attribution methods, we find that faithfulness is operator-dependent: operator gaps reach up to 44 percentage points, with deletion typically inflating estimates on short text but the pattern reversing on long text, suggesting that faithfulness should be interpreted comparatively across intervention operators rather than as a single score. Randomized baselines reveal anti-faithfulness in one-third of configurations, and faithfulness shows zero correlation with human plausibility (|r| < 0.04). Multilingual evaluation reveals dramatic model-language interactions not explained by tokenization alone. We release the ICE framework and ICEBench benchmark. --- ### MedForge: Interpretable Medical Deepfake Detection via Forgery-aware Reasoning Authors: Zhihui Chen, Kai He, Qingyuan Lei, Bin Pu, Jian Zhang, Yuling Xu, Mengling Feng Published: 2026-03-19 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.18577v2 Text-guided image editors can now manipulate authentic medical scans with high fidelity, enabling lesion implantation/removal that threatens clinical trust and safety. Existing defenses are inadequate for healthcare. Medical detectors are largely black-box, while MLLM-based explainers are typically post-hoc, lack medical expertise, and may hallucinate evidence on ambiguous cases. We present MedForge, a data-and-method solution for pre-hoc, evidence-grounded medical forgery detection. We introduce MedForge-90K, a large-scale benchmark of realistic lesion edits across 19 pathologies with expert-guided reasoning supervision via doctor inspection guidelines and gold edit locations. Building on it, MedForge-Reasoner performs localize-then-analyze reasoning, predicting suspicious regions before producing a verdict, and is further aligned with Forgery-aware GSPO to strengthen grounding and reduce hallucinations. Experiments demonstrate state-of-the-art detection accuracy and trustworthy, expert-aligned explanations. --- ### Interplay: Training Independent Simulators for Reference-Free Conversational Recommendation Authors: Jerome Ramos, Feng Xia, Xi Wang, Shubham Chatterjee, Xiao Fu, Hossein A. Rahmani, Aldo Lipani Published: 2026-03-19 Categories: cs.AI, cs.IR arXiv: https://arxiv.org/abs/2603.18573v1 Training conversational recommender systems (CRS) requires extensive dialogue data, which is challenging to collect at scale. To address this, researchers have used simulated user-recommender conversations. Traditional simulation approaches often utilize a single large language model (LLM) that generates entire conversations with prior knowledge of the target items, leading to scripted and artificial dialogues. We propose a reference-free simulation framework that trains two independent LLMs, one as the user and one as the conversational recommender. These models interact in real-time without access to predetermined target items, but preference summaries and target attributes, enabling the recommender to genuinely infer user preferences through dialogue. This approach produces more realistic and diverse conversations that closely mirror authentic human-AI interactions. Our reference-free simulators match or exceed existing methods in quality, while offering a scalable solution for generating high-quality conversational recommendation data without constraining conversations to pre-defined target items. We conduct both quantitative and human evaluations to confirm the effectiveness of our reference-free approach. --- ### UEPS: Robust and Efficient MRI Reconstruction Authors: Xiang Zhou, Hong Shang, Zijian Zhan, Tianyu He, Jintao Meng, Dong Liang Published: 2026-03-19 Categories: eess.IV, cs.CV arXiv: https://arxiv.org/abs/2603.18572v1 Deep unrolled models (DUMs) have become the state of the art for accelerated MRI reconstruction, yet their robustness under domain shift remains a critical barrier to clinical adoption. In this work, we identify coil sensitivity map (CSM) estimation as the primary bottleneck limiting generalization. To address this, we propose UEPS, a novel DUM architecture featuring three key innovations: (i) an Unrolled Expanded (UE) design that eliminates CSM dependency by reconstructing each coil independently; (ii) progressive resolution, which leverages k-space-to-image mapping for efficient coarse-to-fine refinement; and (iii) sparse attention tailored to MRI's 1D undersampling nature. These physics-grounded designs enable simultaneous gains in robustness and computational efficiency. We construct a large-scale zero-shot transfer benchmark comprising 10 out-of-distribution test sets spanning diverse clinical shifts -- anatomy, view, contrast, vendor, field strength, and coil configurations. Extensive experiments demonstrate that UEPS consistently and substantially outperforms existing DUM, end-to-end, diffusion, and untrained methods across all OOD tests, achieving state-of-the-art robustness with low-latency inference suitable for real-time deployment. --- ### CAPSUL: A Comprehensive Human Protein Benchmark for Subcellular Localization Authors: Yicheng Hu, Xinyu Lin, Shulin Li, Wenjie Wang, Fengbin Zhu, Fuli Feng Published: 2026-03-19 Categories: cs.AI, cs.CE, q-bio.QM arXiv: https://arxiv.org/abs/2603.18571v1 Subcellular localization is a crucial biological task for drug target identification and function annotation. Although it has been biologically realized that subcellular localization is closely associated with protein structure, no existing dataset offers comprehensive 3D structural information with detailed subcellular localization annotations, thus severely hindering the application of promising structure-based models on this task. To address this gap, we introduce a new benchmark called $\mathbf{CAPSUL}$, a $\mathbf{C}$omprehensive hum$\mathbf{A}$n $\mathbf{P}$rotein benchmark for $\mathbf{SU}$bcellular $\mathbf{L}$ocalization. It features a dataset that integrates diverse 3D structural representations with fine-grained subcellular localization annotations carefully curated by domain experts. We evaluate this benchmark using a variety of state-of-the-art sequence-based and structure-based models, showcasing the importance of involving structural features in this task. Furthermore, we explore reweighting and single-label classification strategies to facilitate future investigation on structure-based methods for this task. Lastly, we showcase the powerful interpretability of structure-based methods through a case study on the Golgi apparatus, where we discover a decisive localization pattern $α$-helix from attention mechanisms, demonstrating the potential for bridging the gap with intuitive biological interpretability and paving the way for data-driven discoveries in cell biology. --- ### Attack by Unlearning: Unlearning-Induced Adversarial Attacks on Graph Neural Networks Authors: Jiahao Zhang, Yilong Wang, Suhang Wang Published: 2026-03-19 Categories: cs.LG, cs.CR arXiv: https://arxiv.org/abs/2603.18570v1 Graph neural networks (GNNs) are widely used for learning from graph-structured data in domains such as social networks, recommender systems, and financial platforms. To comply with privacy regulations like the GDPR, CCPA, and PIPEDA, approximate graph unlearning, which aims to remove the influence of specific data points from trained models without full retraining, has become an increasingly important component of trustworthy graph learning. However, approximate unlearning often incurs subtle performance degradation, which may incur negative and unintended side effects. In this work, we show that such degradations can be amplified into adversarial attacks. We introduce the notion of \textbf{unlearning corruption attacks}, where an adversary injects carefully chosen nodes into the training graph and later requests their deletion. Because deletion requests are legally mandated and cannot be denied, this attack surface is both unavoidable and stealthy: the model performs normally during training, but accuracy collapses only after unlearning is applied. Technically, we formulate this attack as a bi-level optimization problem: to overcome the challenges of black-box unlearning and label scarcity, we approximate the unlearning process via gradient-based updates and employ a surrogate model to generate pseudo-labels for the optimization. Extensive experiments across benchmarks and unlearning algorithms demonstrate that small, carefully designed unlearning requests can induce significant accuracy degradation, raising urgent concerns about the robustness of GNN unlearning under real-world regulatory demands. The source code will be released upon paper acceptance. --- ### SpecForge: A Flexible and Efficient Open-Source Training Framework for Speculative Decoding Authors: Shenggui Li, Chao Wang, Yikai Zhu, Yubo Wang, Fan Yin, Shuai Shi, Yefei Chen, Xiaomin Dong, Qiaoling Chen, Jin Pan, Ji Li, Laixin Xie et al. Published: 2026-03-19 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2603.18567v1 Large language models incur high inference latency due to sequential autoregressive decoding. Speculative decoding alleviates this bottleneck by using a lightweight draft model to propose multiple tokens for batched verification. However, its adoption has been limited by the lack of high-quality draft models and scalable training infrastructure. We introduce SpecForge, an open-source, production-oriented framework for training speculative decoding models with full support for EAGLE-3. SpecForge incorporates target-draft decoupling, hybrid parallelism, optimized training kernels, and integration with production-grade inference engines, enabling up to 9.9x faster EAGLE-3 training for Qwen3-235B-A22B. In addition, we release SpecBundle, a suite of production-grade EAGLE-3 draft models trained with SpecForge for mainstream open-source LLMs. Through a systematic study of speculative decoding training recipes, SpecBundle addresses the scarcity of high-quality drafts in the community, and our draft models achieve up to 4.48x end-to-end inference speedup on SGLang, establishing SpecForge as a practical foundation for real-world speculative decoding deployment. --- ### Transformers Learn Robust In-Context Regression under Distributional Uncertainty Authors: Hoang T. H. Cao, Hai D. V. Trinh, Tho Quan, Lan V. Truong Published: 2026-03-19 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.18564v1 Recent work has shown that Transformers can perform in-context learning for linear regression under restrictive assumptions, including i.i.d. data, Gaussian noise, and Gaussian regression coefficients. However, real-world data often violate these assumptions: the distributions of inputs, noise, and coefficients are typically unknown, non-Gaussian, and may exhibit dependency across the prompt. This raises a fundamental question: can Transformers learn effectively in-context under realistic distributional uncertainty? We study in-context learning for noisy linear regression under a broad range of distributional shifts, including non-Gaussian coefficients, heavy-tailed noise, and non-i.i.d. prompts. We compare Transformers against classical baselines that are optimal or suboptimal under the corresponding maximum-likelihood criteria. Across all settings, Transformers consistently match or outperform these baselines, demonstrating robust in-context adaptation beyond classical estimators. --- ### Reasonably reasoning AI agents can avoid game-theoretic failures in zero-shot, provably Authors: Enoch Hyunwook Kang Published: 2026-03-19 Categories: cs.AI, cs.MA, econ.TH arXiv: https://arxiv.org/abs/2603.18563v2 As autonomous AI agents increasingly mediate online platform markets, a fundamental question emerges: do these markets generate stable strategic outcomes? In repeated strategic environments, the Nash equilibrium provides a natural benchmark for this stability. However, empirical evidence on off-the-shelf LLM agents is mixed, leaving it unclear whether independently deployed agents can converge to equilibrium behavior without explicit strategic post-training. In this paper, we provide an affirmative answer. Extending the Bayesian learning literature in theoretical economics, we prove that AI agents, acting as Bayesian posterior samplers rather than expected utility maximizers, are guaranteed to eventually become weakly close to a Nash equilibrium in infinitely repeated games. We further extend this analysis to settings in which stage payoffs are unknown ex ante, and agents observe only their privately realized stochastic payoffs, and obtain the same convergence guarantees. Finally, we empirically evaluate these theoretical implications across five repeated-game environments, ranging from the Prisoner's Dilemma to marketing promotion games. Taken together, our findings suggest that strategic stability in AI-mediated markets can emerge from the intrinsic reasoning and learning properties of modern AI agents, without the need for unrealistic universal fine-tuning. --- ### CausalVAD: De-confounding End-to-End Autonomous Driving via Causal Intervention Authors: Jiacheng Tang, Zhiyuan Zhou, Zhuolin He, Jia Zhang, Kai Zhang, Jian Pu Published: 2026-03-19 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2603.18561v2 Planning-oriented end-to-end driving models show great promise, yet they fundamentally learn statistical correlations instead of true causal relationships. This vulnerability leads to causal confusion, where models exploit dataset biases as shortcuts, critically harming their reliability and safety in complex scenarios. To address this, we introduce CausalVAD, a de-confounding training framework that leverages causal intervention. At its core, we design the sparse causal intervention scheme (SCIS), a lightweight, plug-and-play module to instantiate the backdoor adjustment theory in neural networks. SCIS constructs a dictionary of prototypes representing latent driving contexts. It then uses this dictionary to intervene on the model's sparse vectorized queries. This step actively eliminates spurious associations induced by confounders, thereby eliminating spurious factors from the representations for downstream tasks. Extensive experiments on benchmarks like nuScenes show CausalVAD achieves state-of-the-art planning accuracy and safety. Furthermore, our method demonstrates superior robustness against both data bias and noisy scenarios configured to induce causal confusion. --- ### HiMu: Hierarchical Multimodal Frame Selection for Long Video Question Answering Authors: Dan Ben-Ami, Gabriele Serussi, Kobi Cohen, Chaim Baskin Published: 2026-03-19 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.18558v2 Long-form video question answering requires reasoning over extended temporal contexts, making frame selection a critical bottleneck for multi-modal large language models (MLLMs) bound by finite context windows. Within the controlled frame-budget regime that governs practical deployment, prior selectors score frames against a single global query embedding; as a result, compositional multimodal questions that involve temporal ordering or cross-modal cues such as ``what happens on screen right after the narrator mentions the reaction?'' are flattened into a representation that loses sub-event ordering and modality bindings. We introduce \textbf{HiMu}, a training-free framework for compositional multimodal frame selection. A single text-only LLM call decomposes the query into a hierarchical logic tree whose leaves are atomic predicates, each routed to a lightweight expert spanning vision (CLIP, open-vocabulary detection, OCR) and audio (speech recognition and non-speech sound matching). Expert signals are normalized, smoothed to align across modalities, and composed bottom-up through fuzzy-logic operators that enforce temporal sequencing and adjacency, yielding a continuous per-frame satisfaction curve. Under the standard 16-frame budget on Video-MME, LongVideoBench, and HERBench-Lite, HiMu achieves state-of-the-art accuracy among frame selection methods and improves over uniform sampling across seven diverse MLLMs as a drop-in module, matching the accuracy of uniform sampling at $4\times$ its frame budget, without retraining and without multiple iterative MLLM calls during selection. --- ### Cross-Lingual LLM-Judge Transfer via Evaluation Decomposition Authors: Ivaxi Sheth, Zeno Jonke, Amin Mantrach, Saab Mansour Published: 2026-03-19 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.18557v1 As large language models are increasingly deployed across diverse real-world applications, extending automated evaluation beyond English has become a critical challenge. Existing evaluation approaches are predominantly English-focused, and adapting them to other languages is hindered by the scarcity and cost of human-annotated judgments in most languages. We introduce a decomposition-based evaluation framework built around a Universal Criteria Set (UCS). UCS consists of a shared, language-agnostic set of evaluation dimensions, producing an interpretable intermediate representation that supports cross-lingual transfer with minimal supervision. Experiments on multiple faithfulness tasks across languages and model backbones demonstrate consistent improvements over strong baselines without requiring target-language annotations. --- ### End-to-End QGAN-Based Image Synthesis via Neural Noise Encoding and Intensity Calibration Authors: Xue Yang, Rigui Zhou, Shizheng Jia, Dax Enshan Koh, Siong Thye Goh, Yaochong Li, Hongyu Chen, Fuhui Xiong Published: 2026-03-19 Categories: quant-ph, cs.CV arXiv: https://arxiv.org/abs/2603.18554v1 Quantum Generative Adversarial Networks (QGANs) offer a promising path for learning data distributions on near-term quantum devices. However, existing QGANs for image synthesis avoid direct full-image generation, relying on classical post-processing or patch-based methods. These approaches dilute the quantum generator's role and struggle to capture global image semantics. To address this, we propose ReQGAN, an end-to-end framework that synthesizes an entire N=2^D-pixel image using a single D-qubit quantum circuit. ReQGAN overcomes two fundamental bottlenecks hindering direct pixel generation: (1) the rigid classical-to-quantum noise interface and (2) the output mismatch between normalized quantum statistics and the desired pixel-intensity space. We introduce a learnable Neural Noise Encoder for adaptive state preparation and a differentiable Intensity Calibration module to map measurements to a stable, visually meaningful pixel domain. Experiments on MNIST and Fashion-MNIST demonstrate that ReQGAN achieves stable training and effective image synthesis under stringent qubit budgets, with ablation studies verifying the contribution of each component. --- ### Learning Decision-Sufficient Representations for Linear Optimization Authors: Yuhan Ye, Saurabh Amin, Asuman Ozdaglar Published: 2026-03-19 Categories: math.OC, cs.CC, cs.LG arXiv: https://arxiv.org/abs/2603.18551v2 We study how to construct compressed datasets that suffice to recover optimal decisions in linear programs with an unknown cost vector $c$ lying in a prior set $\mathcal{C}$. Recent work by Bennouna et al. provides an exact geometric characterization of sufficient decision datasets (SDDs) via an intrinsic decision-relevant dimension $d^\star$. However, their algorithm for constructing minimum-size SDDs requires solving mixed-integer programs. In this paper, we establish hardness results showing that computing $d^\star$ is NP-hard and deciding whether a dataset is globally sufficient is coNP-hard, thereby resolving a recent open problem posed by Bennouna et al. To address this worst-case intractability, we introduce pointwise sufficiency, a relaxation that requires sufficiency for an individual cost vector. Under nondegeneracy, we provide a polynomial-time cutting-plane algorithm for constructing pointwise-sufficient decision datasets. In a data-driven regime with i.i.d.\ costs, we further propose a cumulative algorithm that aggregates decision-relevant directions across samples, yielding a stable compression scheme of size at most $d^\star$. This leads to a distribution-free PAC guarantee: with high probability over the training sample, the pointwise sufficiency failure probability on a fresh draw is at most $\tilde{O}(d^\star/n)$, and this rate is tight up to logarithmic factors. Finally, we apply decision-sufficient representations to contextual linear optimization, obtaining compressed predictors with generalization bounds scaling as $\tilde{O}(\sqrt{d^\star/n})$ rather than $\tilde{O}(\sqrt{d/n})$, where $d$ is the ambient cost dimension. --- ### SINDy-KANs: Sparse identification of non-linear dynamics through Kolmogorov-Arnold networks Authors: Amanda A. Howard, Nicholas Zolman, Bruno Jacob, Steven L. Brunton, Panos Stinis Published: 2026-03-19 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.18548v1 Kolmogorov-Arnold networks (KANs) have arisen as a potential way to enhance the interpretability of machine learning. However, solutions learned by KANs are not necessarily interpretable, in the sense of being sparse or parsimonious. Sparse identification of nonlinear dynamics (SINDy) is a complementary approach that allows for learning sparse equations for dynamical systems from data; however, learned equations are limited by the library. In this work, we present SINDy-KANs, which simultaneously train a KAN and a SINDy-like representation to increase interpretability of KAN representations with SINDy applied at the level of each activation function, while maintaining the function compositions possible through deep KANs. We apply our method to a number of symbolic regression tasks, including dynamical systems, to show accurate equation discovery across a range of systems. --- ### HEP Statistical Inference for UAV Fault Detection: CLs, LRT, and SBI Applied to Blade Damage Authors: Khushiyant Published: 2026-03-19 Categories: cs.LG, cs.RO, eess.SY arXiv: https://arxiv.org/abs/2603.18546v1 This paper transfers three statistical methods from particle physics to multirotor propeller fault detection: the likelihood ratio test (LRT) for binary detection, the CLs modified frequentist method for false alarm rate control, and sequential neural posterior estimation (SNPE) for quantitative fault characterization. Operating on spectral features tied to rotor harmonic physics, the system returns three outputs: binary detection, controlled false alarm rates, and calibrated posteriors over fault severity and motor location. On UAV-FD, a hexarotor dataset of 18 real flights with 5% and 10% blade damage, leave-one-flight-out cross-validation gives AUC 0.862 +/- 0.007 (95% CI: 0.849--0.876), outperforming CUSUM (0.708 +/- 0.010), autoencoder (0.753 +/- 0.009), and LSTM autoencoder (0.551). At 5% false alarm rate the system detects 93% of significant and 81% of subtle blade damage. On PADRE, a quadrotor platform, AUC reaches 0.986 after refitting only the generative models. SNPE gives a full posterior over fault severity (90% credible interval coverage 92--100%, MAE 0.012), so the output includes uncertainty rather than just a point estimate or fault flag. Per-flight sequential detection achieves 100% fault detection with 94% overall accuracy. --- ### CoDA: Exploring Chain-of-Distribution Attacks and Post-Hoc Token-Space Repair for Medical Vision-Language Models Authors: Xiang Chen, Fangfang Yang, Chunlei Meng, Yuxian Dong, Ang Li, Yiwei Wei, Jiahuan Long, Jiujiang Guo, Chengyin Hu Published: 2026-03-19 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.18545v2 Medical vision--language models (MVLMs) are increasingly used as perceptual backbones in radiology pipelines and as the visual front end of multimodal assistants, yet their reliability under real clinical workflows remains underexplored. Prior robustness evaluations often assume clean, curated inputs or study isolated corruptions, overlooking routine acquisition, reconstruction, display, and delivery operations that preserve clinical readability while shifting image statistics. To address this gap, we propose CoDA, a chain-of-distribution framework that constructs clinically plausible pipeline shifts by composing acquisition-like shading, reconstruction and display remapping, and delivery and export degradations. Under masked structural-similarity constraints, CoDA jointly optimizes stage compositions and parameters to induce failures while preserving visual plausibility. Across brain MRI, chest X-ray, and abdominal CT, CoDA substantially degrades the zero-shot performance of CLIP-style MVLMs, with chained compositions consistently more damaging than any single stage. We also evaluate multimodal large language models (MLLMs) as technical-authenticity auditors of imaging realism and quality rather than pathology. Proprietary multimodal models show degraded auditing reliability and persistent high-confidence errors on CoDA-shifted samples, while the medical-specific MLLMs we test exhibit clear deficiencies in medical image quality auditing. Finally, we introduce a post-hoc repair strategy based on teacher-guided token-space adaptation with patch-level alignment, which improves accuracy on archived CoDA outputs. Overall, our findings characterize a clinically grounded threat surface for MVLM deployment and show that lightweight alignment improves robustness in deployment. --- ### SCISSR: Scribble-Conditioned Interactive Surgical Segmentation and Refinement Authors: Haonan Ping, Jian Jiang, Cheng Yuan, Qizhen Sun, Lv Wu, Yutong Ban Published: 2026-03-19 Categories: eess.IV, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2603.18544v1 Accurate segmentation of tissues and instruments in surgical scenes is annotation-intensive due to irregular shapes, thin structures, specularities, and frequent occlusions. While SAM models support point, box, and mask prompts, points are often too sparse and boxes too coarse to localize such challenging targets. We present SCISSR, a scribble-promptable framework for interactive surgical scene segmentation. It introduces a lightweight Scribble Encoder that converts freehand scribbles into dense prompt embeddings compatible with the mask decoder, enabling iterative refinement for a target object by drawing corrective strokes on error regions. Because all added modules (the Scribble Encoder, Spatial Gated Fusion, and LoRA adapters) interact with the backbone only through its standard embedding interfaces, the framework is not tied to a single model: we build on SAM 2 in this work, yet the same components transfer to other prompt-driven segmentation architectures such as SAM 3 without structural modification. To preserve pre-trained capabilities, we train only these lightweight additions while keeping the remaining backbone frozen. Experiments on EndoVis 2018 demonstrate strong in-domain performance, while evaluation on the out-of-distribution CholecSeg8k further confirms robustness across surgical domains. SCISSR achieves 95.41% Dice on EndoVis 2018 with five interaction rounds and 96.30% Dice on CholecSeg8k with three interaction rounds, outperforming iterative point prompting on both benchmarks. --- ### Remedying Target-Domain Astigmatism for Cross-Domain Few-Shot Object Detection Authors: Yongwei Jiang, Yixiong Zou, Yuhua Li, Ruixuan Li Published: 2026-03-19 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.18541v1 Cross-domain few-shot object detection (CD-FSOD) aims to adapt pretrained detectors from a source domain to target domains with limited annotations, suffering from severe domain shifts and data scarcity problems. In this work, we find a previously overlooked phenomenon: models exhibit dispersed and unfocused attention in target domains, leading to imprecise localization and redundant predictions, just like a human cannot focus on visual objects. Therefore, we call it the target-domain Astigmatism problem. Analysis on attention distances across transformer layers reveals that regular fine-tuning inherently shows a trend to remedy this problem, but results are still far from satisfactory, which we aim to enhance in this paper. Biologically inspired by the human fovea-style visual system, we enhance the fine-tuning's inherent trend through a center-periphery attention refinement framework, which contains (1) a Positive Pattern Refinement module to reshape attention toward semantic objects using class-specific prototypes, simulating the visual center region; (2) a Negative Context Modulation module to enhance boundary discrimination by modeling background context, simulating the visual periphery region; and (3) a Textual Semantic Alignment module to strengthen center-periphery distinction through cross-modal cues. Our bio-inspired approach transforms astigmatic attention into focused patterns, substantially improving adaptation to target domains. Experiments on six challenging CD-FSOD benchmarks consistently demonstrate improved detection accuracy and establish new state-of-the-art results. --- ### GAPSL: A Gradient-Aligned Parallel Split Learning over Data-Heterogeneous Edge Computing Systems Authors: Zheng Lin, Ons Aouedi, Zihan Fang, Wei Ni, Yue Gao, Symeon Chatzinotas, Xianhao Chen Published: 2026-03-19 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.18540v2 The increasing complexity of neural networks poses significant challenges for democratizing federated learning (FL) on resource-constrained edge devices. Parallel split learning (PSL) has emerged as a promising solution by offloading substantial computing workload to a server via model partitioning, shrinking client-side computing load, and eliminating the client-side model aggregation for reduced communication and deployment costs. However, the highly heterogeneous nature of client data in edge computing systems causes aggregation-free PSL to suffer from severe training divergence, stemming from gradient directional inconsistency across clients. To address this challenge, we propose GAPSL, a gradient-aligned PSL framework tailored for data-heterogeneous edge systems, which comprises two key components: leader gradient identification (LGI) and gradient direction alignment (GDA). LGI dynamically selects a set of directionally consistent device gradients to construct a leader gradient as a robust proxy for the global convergence trend. GDA employs a direction-aware regularization to align each client's gradient with the leader gradient, thereby mitigating inter-device gradient directional inconsistency and enhancing model convergence. We evaluate GAPSL on a prototype computing testbed. Extensive experiments demonstrate that GAPSL consistently outperforms state-of-the-art benchmarks in training accuracy, convergence latency, and system robustness under severe data heterogeneity. --- ### iSatCR: Graph-Empowered Joint Onboard Computing and Routing for LEO Data Delivery Authors: Jiangtao Luo, Bingbing Xu, Shaohua Xia, Yongyi Ran Published: 2026-03-19 Categories: cs.NI, cs.LG arXiv: https://arxiv.org/abs/2603.18539v1 Sending massive Earth observation data produced by low Earth orbit (LEO) satellites back to the ground for processing consumes a large amount of on-orbit bandwidth and exacerbates the space-to-ground link bottleneck. Most prior work has concentrated on optimizing the routing of raw data within the constellation, yet cannot cope with the surge in data volume. Recently, advances in onboard computing have made it possible to process data in situ, thus significantly reducing the data volume to be transmitted. In this paper, we present iSatCR, a distributed graph-based approach that jointly optimizes onboard computing and routing to boost transmission efficiency. Within iSatCR, we design a novel graph embedding utilizing shifted feature aggregation and distributed message passing to capture satellite states, and then propose a distributed graph-based deep reinforcement learning algorithm that derives joint computing-routing strategies under constrained on-board storage to handle the complexity and dynamics of LEO networks. Extensive experiments show iSatCR outperforms baselines, particularly under high load. --- ### Beyond Passive Aggregation: Active Auditing and Topology-Aware Defense in Decentralized Federated Learning Authors: Sheng Pan, Niansheng Tang Published: 2026-03-19 Categories: cs.LG, stat.ME arXiv: https://arxiv.org/abs/2603.18538v1 Decentralized Federated Learning (DFL) remains highly vulnerable to adaptive backdoor attacks designed to bypass traditional passive defense metrics. To address this limitation, we shift the defensive paradigm toward a novel active, interventional auditing framework. First, we establish a dynamical model to characterize the spatiotemporal diffusion of adversarial updates across complex graph topologies. Second, we introduce a suite of proactive auditing metrics, stochastic entropy anomaly, randomized smoothing Kullback-Leibler divergence, and activation kurtosis. These metrics utilize private probes to stress-test local models, effectively exposing latent backdoors that remain invisible to conventional static detection. Furthermore, we implement a topology-aware defense placement strategy to maximize global aggregation resilience. We provide theoretical property for the system's convergence under co-evolving attack and defense dynamics. Numeric empirical evaluations across diverse architectures demonstrate that our active framework is highly competitive with state-of-the-art defenses in mitigating stealthy, adaptive backdoors while preserving primary task utility. --- ### Data-efficient pre-training by scaling synthetic megadocs Authors: Konwoo Kim, Suhas Kotha, Yejin Choi, Tatsunori Hashimoto, Nick Haber, Percy Liang Published: 2026-03-19 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.18534v1 Synthetic data augmentation has emerged as a promising solution when pre-training is constrained by data rather than compute. We study how to design synthetic data algorithms that achieve better loss scaling: not only lowering loss at finite compute but especially as compute approaches infinity. We first show that pre-training on web data mixed with synthetically generated rephrases improves i.i.d. validation loss on the web data, despite the synthetic data coming from an entirely different distribution. With optimal mixing and epoching, loss and benchmark accuracy improve without overfitting as the number of synthetic generations grows, plateauing near $1.48\times$ data efficiency at 32 rephrases per document. We find even better loss scaling under a new perspective: synthetic generations from the same document can form a single substantially longer megadocument instead of many short documents. We show two ways to construct megadocs: stitching synthetic rephrases from the same web document or stretching a document by inserting rationales. Both methods improve i.i.d. loss, downstream benchmarks, and especially long-context loss relative to simple rephrasing, increasing data efficiency from $1.48\times$ to $1.80\times$ at $32$ generations per document. Importantly, the improvement of megadocs over simple rephrasing widens as more synthetic data is generated. Our results show how to design synthetic data algorithms that benefit more from increasing compute when data-constrained. --- ### Balancing the Reasoning Load: Difficulty-Differentiated Policy Optimization with Length Redistribution for Efficient and Robust Reinforcement Learning Authors: Yinan Xia, Haotian Zhang, Huiming Wang Published: 2026-03-19 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2603.18533v2 Large Reasoning Models (LRMs) have shown exceptional reasoning capabilities, but they also suffer from the issue of overthinking, often generating excessively long and redundant answers. For problems that exceed the model's capabilities, LRMs tend to exhibit the overconfidence phenomenon, generating overly short but incorrect answers, which may contribute to suboptimal performance. To address these issues, we propose Difficulty-Differentiated Policy Optimization (DDPO), an efficient reinforcement learning algorithm that optimizes simple and complex tasks separately based on the overconfidence phenomenon. Specifically, it reduces the output length for simple tasks without compromising accuracy, while for complex tasks, it expands the exploration space to improve performance. We further derive the theoretical conditions for maximizing expected accuracy, which require the length distribution to closely approximate the optimal length and be as concentrated as possible. Based on these conditions, we propose using the difficulty-level average as a well-founded reference for length optimization. Extensive experiments on both in-domain and out-of-domain benchmarks validate the superiority and effectiveness of DDPO. Compared to GRPO, DDPO reduces the average answer length by 12% while improving accuracy by 1.85% across multiple benchmarks, achieving a better trade-off between accuracy and length. The code is available at https://github.com/Yinan-Xia/DDPO. --- ### Scaling Sim-to-Real VLA Reinforcement Learning with Generative 3D Worlds Authors: Andrew Choi, Xinjie Wang, Zhizhong Su, Wei Xu Published: 2026-03-19 Categories: cs.RO, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.18532v3 The strong performance of large vision-language models (VLMs) trained with reinforcement learning (RL) has motivated similar approaches for fine-tuning vision-language-action (VLA) models in robotics. Many recent works fine-tune VLAs directly in the real world to avoid addressing the sim-to-real gap. While real-world RL circumvents sim-to-real issues, it inherently limits the generality of the resulting VLA, as scaling scene and object diversity in the physical world is prohibitively difficult. This leads to the paradoxical outcome of transforming a broadly pretrained model into an overfitted, scene-specific policy. Training in simulation can instead provide access to diverse scenes, but designing those scenes is also costly. In this work, we show that VLAs can be RL fine-tuned across broad scene and object distributions and with reduced labor by leveraging 3D world generative models. Using these models together with a language-driven scene designer, we generate 100 diverse interactive scenes containing unique objects and backgrounds, enabling scalable and highly parallel policy learning. Starting from a pretrained imitation baseline, our approach increases simulation success from 9.7% up to 79.8% while achieving a 1.25$\times$ speedup in task completion time. We further demonstrate successful sim-to-real transfer enabled by the quality of the generated scenes together with domain randomization, improving real-world success from 21.7% to 75% and achieving a 1.13$\times$ speedup. Finally, we further highlight the benefits of leveraging the effectively unlimited data from 3D world generative models through an ablation study showing that increasing scene diversity directly improves zero-shot generalization. --- ### Scaling Sim-to-Real Reinforcement Learning for Robot VLAs with Generative 3D Worlds Authors: Andrew Choi, Xinjie Wang, Zhizhong Su, Wei Xu Published: 2026-03-19 Categories: cs.RO, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.18532v2 The strong performance of large vision-language models (VLMs) trained with reinforcement learning (RL) has motivated similar approaches for fine-tuning vision-language-action (VLA) models in robotics. Many recent works fine-tune VLAs directly in the real world to avoid addressing the sim-to-real gap. While real-world RL circumvents sim-to-real issues, it inherently limits the generality of the resulting VLA, as scaling scene and object diversity in the physical world is prohibitively difficult. This leads to the paradoxical outcome of transforming a broadly pretrained model into an overfitted, scene-specific policy. Training in simulation can instead provide access to diverse scenes, but designing those scenes is also costly. In this work, we show that VLAs can be RL fine-tuned without sacrificing generality and with reduced labor by leveraging 3D world generative models. Using these models together with a language-driven scene designer, we generate hundreds of diverse interactive scenes containing unique objects and backgrounds, enabling scalable and highly parallel policy learning. Starting from a pretrained imitation baseline, our approach increases simulation success from 9.7% to 79.8% while achieving a 1.25$\times$ speedup in task completion time. We further demonstrate successful sim-to-real transfer enabled by the quality of the generated digital twins together with domain randomization, improving real-world success from 21.7% to 75% and achieving a 1.13$\times$ speedup. Finally, we further highlight the benefits of leveraging the effectively unlimited data from 3D world generative models through an ablation study showing that increasing scene diversity directly improves zero-shot generalization. --- ### When Names Change Verdicts: Intervention Consistency Reveals Systematic Bias in LLM Decision-Making Authors: Abhinaba Basu, Pavan Chakraborty Published: 2026-03-19 Categories: cs.CL, cs.AI, cs.CY, cs.LG arXiv: https://arxiv.org/abs/2603.18530v1 Large language models (LLMs) are increasingly used for high-stakes decisions, yet their susceptibility to spurious features remains poorly characterized. We introduce ICE-Guard, a framework applying intervention consistency testing to detect three types of spurious feature reliance: demographic (name/race swaps), authority (credential/prestige swaps), and framing (positive/negative restatements). Across 3,000 vignettes spanning 10 high-stakes domains, we evaluate 11 LLMs from 8 families and find that (1) authority bias (mean 5.8%) and framing bias (5.0%) substantially exceed demographic bias (2.2%), challenging the field's narrow focus on demographics; (2) bias concentrates in specific domains -- finance shows 22.6% authority bias while criminal justice shows only 2.8%; (3) structured decomposition, where the LLM extracts features and a deterministic rubric decides, reduces flip rates by up to 100% (median 49% across 9 models). We demonstrate an ICE-guided detect-diagnose-mitigate-verify loop achieving cumulative 78% bias reduction via iterative prompt patching. Validation against real COMPAS recidivism data shows COMPAS-derived flip rates exceed pooled synthetic rates, suggesting our benchmark provides a conservative estimate of real-world bias. Code and data are publicly available. --- ### Correlation-Weighted Multi-Reward Optimization for Compositional Generation Authors: Jungmyung Wi, Hyunsoo Kim, Donghyun Kim Published: 2026-03-19 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.18528v2 Text-to-image models produce images that align well with natural language prompts, but compositional generation has long been a central challenge. Models often struggle to satisfy multiple concepts within a single prompt, frequently omitting some concepts and resulting in partial success. Such failures highlight the difficulty of jointly optimizing multiple concepts during reward optimization, where competing concepts can interfere with one another. To address this limitation, we propose Correlation-Weighted Multi-Reward Optimization (CMO), a framework that leverages the correlation structure among concept rewards to adaptively weight each attribute concept in optimization. By accounting for interactions among concepts, CMO balances competing reward signals and emphasizes concepts that are partially satisfied yet inconsistently generated across samples, improving compositional generation. Specifically, we decompose multi-concept prompts into pre-defined concept groups (e.g., objects, attributes, and relations) and obtain reward signals from dedicated reward models for each concept. We then adaptively reweight these rewards, assigning higher weights to conflicting or hard-to-satisfy concepts using correlation-based difficulty estimation. By focusing optimization on the most challenging concepts within each group, CMO encourages the model to consistently satisfy all requested attributes simultaneously. We apply our approach to train state-of-the-art diffusion models, SD3.5 and FLUX.1-dev, and demonstrate consistent improvements on challenging multi-concept benchmarks, including ConceptMix, GenEval 2, and T2I-CompBench. The code is available at https://github.com/TheDarkKnight-21th/CMO. --- ### 3DreamBooth: High-Fidelity 3D Subject-Driven Video Generation Model Authors: Hyun-kyu Ko, Jihyeon Park, Younghyun Kim, Dongheok Park, Eunbyung Park Published: 2026-03-19 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.18524v1 Creating dynamic, view-consistent videos of customized subjects is highly sought after for a wide range of emerging applications, including immersive VR/AR, virtual production, and next-generation e-commerce. However, despite rapid progress in subject-driven video generation, existing methods predominantly treat subjects as 2D entities, focusing on transferring identity through single-view visual features or textual prompts. Because real-world subjects are inherently 3D, applying these 2D-centric approaches to 3D object customization reveals a fundamental limitation: they lack the comprehensive spatial priors necessary to reconstruct the 3D geometry. Consequently, when synthesizing novel views, they must rely on generating plausible but arbitrary details for unseen regions, rather than preserving the true 3D identity. Achieving genuine 3D-aware customization remains challenging due to the scarcity of multi-view video datasets. While one might attempt to fine-tune models on limited video sequences, this often leads to temporal overfitting. To resolve these issues, we introduce a novel framework for 3D-aware video customization, comprising 3DreamBooth and 3Dapter. 3DreamBooth decouples spatial geometry from temporal motion through a 1-frame optimization paradigm. By restricting updates to spatial representations, it effectively bakes a robust 3D prior into the model without the need for exhaustive video-based training. To enhance fine-grained textures and accelerate convergence, we incorporate 3Dapter, a visual conditioning module. Following single-view pre-training, 3Dapter undergoes multi-view joint optimization with the main generation branch via an asymmetrical conditioning strategy. This design allows the module to act as a dynamic selective router, querying view-specific geometric hints from a minimal reference set. Project page: https://ko-lani.github.io/3DreamBooth/ --- ### Counting Circuits: Mechanistic Interpretability of Visual Reasoning in Large Vision-Language Models Authors: Liwei Che, Zhiyu Xue, Yihao Quan, Benlin Liu, Zeru Shi, Michelle Hurst, Jacob Feldman, Ruixiang Tang, Ranjay Krishna, Vladimir Pavlovic Published: 2026-03-19 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.18523v1 Counting serves as a simple but powerful test of a Large Vision-Language Model's (LVLM's) reasoning; it forces the model to identify each individual object and then add them all up. In this study, we investigate how LVLMs implement counting using controlled synthetic and real-world benchmarks, combined with mechanistic analyses. Our results show that LVLMs display a human-like counting behavior, with precise performance on small numerosities and noisy estimation for larger quantities. We introduce two novel interpretability methods, Visual Activation Patching and HeadLens, and use them to uncover a structured "counting circuit" that is largely shared across a variety of visual reasoning tasks. Building on these insights, we propose a lightweight intervention strategy that exploits simple and abundantly available synthetic images to fine-tune arbitrary pretrained LVLMs exclusively on counting. Despite the narrow scope of this fine-tuning, the intervention not only enhances counting accuracy on in-distribution synthetic data, but also yields an average improvement of +8.36% on out-of-distribution counting benchmarks and an average gain of +1.54% on complex, general visual reasoning tasks for Qwen2.5-VL. These findings highlight the central, influential role of counting in visual reasoning and suggest a potential pathway for improving overall visual reasoning capabilities through targeted enhancement of counting mechanisms. --- ### On the Peril of (Even a Little) Nonstationarity in Satisficing Regret Minimization Authors: Yixuan Zhang, Ruihao Zhu, Qiaomin Xie Published: 2026-03-19 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2603.18514v2 Motivated by the principle of satisficing in decision-making, we study satisficing regret guarantees for nonstationary $K$-armed bandits. We show that in the general realizable, piecewise-stationary setting with $L$ stationary segments, the optimal regret is $Θ(L\log T)$ as long as $L\geq 2$. This stands in sharp contrast to the case of $L=1$ (i.e., the stationary setting), where a $T$-independent $Θ(1)$ satisficing regret is achievable under realizability. In other words, the optimal regret has to scale with $T$ even if just a little nonstationarity presents. A key ingredient in our analysis is a novel Fano-based framework tailored to nonstationary bandits via a \emph{post-interaction reference} construction. This framework strictly extends the classical Fano method for passive estimation as well as recent interactive Fano techniques for stationary bandits. As a complement, we also discuss a special regime in which constant satisficing regret is again possible. --- ### CAFlow: Adaptive-Depth Single-Step Flow Matching for Efficient Histopathology Super-Resolution Authors: Elad Yoshai, Ariel D. Yoshai, Natan T. Shaked Published: 2026-03-19 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.18513v1 In digital pathology, whole-slide images routinely exceed gigapixel resolution, making computationally intensive generative super-resolution (SR) impractical for routine deployment. We introduce CAFlow, an adaptive-depth single-step flow-matching framework that routes each image tile to the shallowest network exit that preserves reconstruction quality. CAFlow performs flow matching in pixel-unshuffled rearranged space, reducing spatial computation by 16x while enabling direct inference. We show that dedicating half of training to exact t=0 samples is essential for single-step quality (-1.5 dB without it). The backbone, FlowResNet (1.90M parameters), mixes convolution and window self-attention blocks across four early exits spanning 3.1 to 13.3 GFLOPs. A lightweight exit classifier (~6K parameters) achieves 33% compute savings at only 0.12 dB cost. On multi-organ histopathology x4 SR, adaptive routing achieves 31.72 dB PSNR versus 31.84 dB at full depth, while the shallowest exit exceeds bicubic by +1.9 dB at 2.8x less compute than SwinIR-light. The method generalizes to held-out colon tissue with minimal quality loss (-0.02 dB), and at x8 upscaling it outperforms all comparable-compute baselines while remaining competitive with the much larger SwinIR-Medium model. Downstream nuclei segmentation confirms preservation of clinically relevant structure. The model trains in under 5 hours on a single GPU, and adaptive routing can reduce whole-slide inference from minutes to seconds. --- ### OnlinePG: Online Open-Vocabulary Panoptic Mapping with 3D Gaussian Splatting Authors: Hongjia Zhai, Qi Zhang, Xiaokun Pan, Xiyu Zhang, Yitong Dong, Huaqi Zhang, Dan Xu, Guofeng Zhang Published: 2026-03-19 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.18510v1 Open-vocabulary scene understanding with online panoptic mapping is essential for embodied applications to perceive and interact with environments. However, existing methods are predominantly offline or lack instance-level understanding, limiting their applicability to real-world robotic tasks. In this paper, we propose OnlinePG, a novel and effective system that integrates geometric reconstruction and open-vocabulary perception using 3D Gaussian Splatting in an online setting. Technically, to achieve online panoptic mapping, we employ an efficient local-to-global paradigm with a sliding window. To build local consistency map, we construct a 3D segment clustering graph that jointly leverages geometric and semantic cues, fusing inconsistent segments within sliding window into complete instances. Subsequently, to update the global map, we construct explicit grids with spatial attributes for the local 3D Gaussian map and fuse them into the global map via robust bidirectional bipartite 3D Gaussian instance matching. Finally, we utilize the fused VLM features inside the 3D spatial attribute grids to achieve open-vocabulary scene understanding. Extensive experiments on widely used datasets demonstrate that our method achieves better performance among online approaches, while maintaining real-time efficiency. --- ### Foundations and Architectures of Artificial Intelligence for Motor Insurance Authors: Teerapong Panboonyuen Published: 2026-03-19 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.18508v1 This handbook presents a systematic treatment of the foundations and architectures of artificial intelligence for motor insurance, grounded in large-scale real-world deployment. It formalizes a vertically integrated AI paradigm that unifies perception, multimodal reasoning, and production infrastructure into a cohesive intelligence stack for automotive risk assessment and claims processing. At its core, the handbook develops domain-adapted transformer architectures for structured visual understanding, relational vehicle representation learning, and multimodal document intelligence, enabling end-to-end automation of vehicle damage analysis, claims evaluation, and underwriting workflows. These components are composed into a scalable pipeline operating under practical constraints observed in nationwide motor insurance systems in Thailand. Beyond model design, the handbook emphasizes the co-evolution of learning algorithms and MLOps practices, establishing a principled framework for translating modern artificial intelligence into reliable, production-grade systems in high-stakes industrial environments. --- ### Expert Personas Improve LLM Alignment but Damage Accuracy: Bootstrapping Intent-Based Persona Routing with PRISM Authors: Zizhao Hu, Mohammad Rostami, Jesse Thomason Published: 2026-03-19 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.18507v1 Persona prompting can steer LLM generation towards a domain-specific tone and pattern. This behavior enables use cases in multi-agent systems where diverse interactions are crucial and human-centered tasks require high-level human alignment. Prior works provide mixed opinions on their utility: some report performance gains when using expert personas for certain domains and their contribution to data diversity in synthetic data creation, while others find near-zero or negative impact on general utility. To fully leverage the benefits of the LLM persona and avoid its harmfulness, a more comprehensive investigation of the mechanism is crucial. In this work, we study how model optimization, task type, prompt length, and placement can impact expert persona effectiveness across instruction-tuned and reasoning LLMs, and provide insight into conditions under which expert personas fail and succeed. Based on our findings, we developed a pipeline to fully leverage the benefits of an expert persona, named PRISM (Persona Routing via Intent-based Self-Modeling), which self-distills an intent-conditioned expert persona into a gated LoRA adapter through a bootstrapping process that requires no external data, models, or knowledge. PRISM enhances human preference and safety alignment on generative tasks while maintaining accuracy on discriminative tasks across all models, with minimal memory and computing overhead. --- ### From Snapshots to Symphonies: The Evolution of Protein Prediction from Static Structures to Generative Dynamics and Multimodal Interactions Authors: Jingzhi Chen, Lijian Xu Published: 2026-03-19 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.18505v1 The protein folding problem has been fundamentally transformed by artificial intelligence, evolving from static structure prediction toward the modeling of dynamic conformational ensembles and complex biomolecular interactions. This review systematically examines the paradigm shift in AI driven protein science across five interconnected dimensions: unified multimodal representations that integrate sequences, geometries, and textual knowledge; refinement of static prediction through MSA free architectures and all atom complex modeling; generative frameworks, including diffusion models and flow matching, that capture conformational distributions consistent with thermodynamic ensembles; prediction of heterogeneous interactions spanning protein ligand, protein nucleic acid, and protein protein complexes; and functional inference of fitness landscapes, mutational effects, and text guided property prediction. We critically analyze current bottlenecks, including data distribution biases, limited mechanistic interpretability, and the disconnect between geometric metrics and biophysical reality, while identifying future directions toward physically consistent generative models, multimodal foundation architectures, and experimental closed loop systems. This methodological transformation marks artificial intelligence's transition from a structural analysis tool into a universal simulator capable of understanding and ultimately rewriting the dynamic language of life. --- ### HOMEY: Heuristic Object Masking with Enhanced YOLO for Property Insurance Risk Detection Authors: Teerapong Panboonyuen Published: 2026-03-19 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.18502v1 Automated property risk detection is a high-impact yet underexplored frontier in computer vision with direct implications for real estate, underwriting, and insurance operations. We introduce HOMEY (Heuristic Object Masking with Enhanced YOLO), a novel detection framework that combines YOLO with a domain-specific masking mechanism and a custom-designed loss function. HOMEY is trained to detect 17 risk-related property classes, including structural damages (e.g., cracked foundations, roof issues), maintenance neglect (e.g., dead yards, overgrown bushes), and liability hazards (e.g., falling gutters, garbage, hazard signs). Our approach introduces heuristic object masking to amplify weak signals in cluttered backgrounds and risk-aware loss calibration to balance class skew and severity weighting. Experiments on real-world property imagery demonstrate that HOMEY achieves superior detection accuracy and reliability compared to baseline YOLO models, while retaining fast inference. Beyond detection, HOMEY enables interpretable and cost-efficient risk analysis, laying the foundation for scalable AI-driven property insurance workflows. --- ### Efficient Video Diffusion with Sparse Information Transmission for Video Compression Authors: Mingde Zhou, Zheng Chen, Yulun Zhang Published: 2026-03-19 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.18501v1 Video compression aims to maximize reconstruction quality with minimal bitrates. Beyond standard distortion metrics, perceptual quality and temporal consistency are also critical. However, at ultra-low bitrates, traditional end-to-end compression models tend to produce blurry images of poor perceptual quality. Besides, existing generative compression methods often treat video frames independently and show limitations in time coherence and efficiency. To address these challenges, we propose the Efficient Video Diffusion with Sparse Information Transmission (Diff-SIT), which comprises the Sparse Temporal Encoding Module (STEM) and the One-Step Video Diffusion with Frame Type Embedder (ODFTE). The STEM sparsely encodes the original frame sequence into an information-rich intermediate sequence, achieving significant bitrate savings. Subsequently, the ODFTE processes this intermediate sequence as a whole, which exploits the temporal correlation. During this process, our proposed Frame Type Embedder (FTE) guides the diffusion model to perform adaptive reconstruction according to different frame types to optimize the overall quality. Extensive experiments on multiple datasets demonstrate that Diff-SIT establishes a new state-of-the-art in perceptual quality and temporal consistency, particularly in the challenging ultra-low-bitrate regime. Code is released at https://github.com/MingdeZhou/Diff-SIT. --- ### Recovering Sparse Neural Connectivity from Partial Measurements: A Covariance-Based Approach with Granger-Causality Refinement Authors: Quilee Simeon Published: 2026-03-19 Categories: q-bio.QM, cs.NE arXiv: https://arxiv.org/abs/2603.18497v1 Inferring the connectivity of neural circuits from incomplete observations is a fundamental challenge in neuroscience. We present a covariance-based method for estimating the weight matrix of a recurrent neural network from sparse, partial measurements across multiple recording sessions. By accumulating pairwise covariance estimates across sessions where different subsets of neurons are observed, we reconstruct the full connectivity matrix without requiring simultaneous recording of all neurons. A Granger-causality refinement step enforces biological constraints via projected gradient descent. Through systematic experiments on synthetic networks modeling small brain circuits, we characterize a fundamental control-estimation tradeoff: stimulation aids identifiability but disrupts intrinsic dynamics, with the optimal level depending on measurement density. We discover that the ``incorrect'' linear approximation acts as implicit regularization -- outperforming the oracle estimator with known nonlinearity at all operating regimes -- and provide an exact characterization via the Stein--Price identity. --- ### NymeriaPlus: Enriching Nymeria Dataset with Additional Annotations and Data Authors: Daniel DeTone, Federica Bogo, Eric-Tuan Le, Duncan Frost, Julian Straub, Yawar Siddiqui, Yuting Ye, Jakob Engel, Richard Newcombe, Lingni Ma Published: 2026-03-19 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.18496v1 The Nymeria Dataset, released in 2024, is a large-scale collection of in-the-wild human activities captured with multiple egocentric wearable devices that are spatially localized and temporally synchronized. It provides body-motion ground truth recorded with a motion-capture suit, device trajectories, semi-dense 3D point clouds, and in-context narrations. In this paper, we upgrade Nymeria and introduce NymeriaPlus. NymeriaPlus features: (1) improved human motion in Momentum Human Rig (MHR) and SMPL formats; (2) dense 3D and 2D bounding box annotations for indoor objects and structural elements; (3) instance-level 3D object reconstructions; and (4) additional modalities e.g., basemap recordings, audio, and wristband videos. By consolidating these complementary modalities and annotations into a single, coherent benchmark, NymeriaPlus strengthens Nymeria into a more powerful in-the-wild egocentric dataset. We expect NymeriaPlus to bridge a key gap in existing egocentric resources and to support a broader range of research, including unique explorations of multimodal learning for embodied AI. --- ### Cross-Domain Demo-to-Code via Neurosymbolic Counterfactual Reasoning Authors: Jooyoung Kim, Wonje Choi, Younguk Song, Honguk Woo Published: 2026-03-19 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.18495v1 Recent advances in Vision-Language Models (VLMs) have enabled video-instructed robotic programming, allowing agents to interpret video demonstrations and generate executable control code. We formulate video-instructed robotic programming as a cross-domain adaptation problem, where perceptual and physical differences between demonstration and deployment induce procedural mismatches. However, current VLMs lack the procedural understanding needed to reformulate causal dependencies and achieve task-compatible behavior under such domain shifts. We introduce NeSyCR, a neurosymbolic counterfactual reasoning framework that enables verifiable adaptation of task procedures, providing a reliable synthesis of code policies. NeSyCR abstracts video demonstrations into symbolic trajectories that capture the underlying task procedure. Given deployment observations, it derives counterfactual states that reveal cross-domain incompatibilities. By exploring the symbolic state space with verifiable checks, NeSyCR proposes procedural revisions that restore compatibility with the demonstrated procedure. NeSyCR achieves a 31.14% improvement in task success over the strongest baseline Statler, showing robust cross-domain adaptation across both simulated and real-world manipulation tasks. --- ### FILT3R: Latent State Adaptive Kalman Filter for Streaming 3D Reconstruction Authors: Seonghyun Jin, Jong Chul Ye Published: 2026-03-19 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.18493v2 Streaming 3D reconstruction maintains a persistent latent state that is updated online from incoming frames, enabling constant-memory inference. A key failure mode is the state update rule: aggressive overwrites forget useful history, while conservative updates fail to track new evidence, and both behaviors become unstable beyond the training horizon. To address this challenge, we propose FILT3R, a training-free latent filtering layer that casts recurrent state updates as stochastic state estimation in token space. FILT3R maintains a per-token variance and computes a Kalman-style gain that adaptively balances memory retention against new observations. Process noise -- governing how much the latent state is expected to change between frames -- is estimated online from EMA-normalized temporal drift of candidate tokens. Using extensive experiments, we demonstrate that FILT3R yields an interpretable, plug-in update rule that generalizes common overwrite and gating policies as special cases. Specifically, we show that gains shrink in stable regimes as uncertainty contracts with accumulated evidence, and rise when genuine scene change increases process uncertainty, improving long-horizon stability for depth, pose, and 3D reconstruction, compared to the existing methods. --- ### FILT3R: Latent State Adaptive Kalman Filter for Streaming 3D Reconstruction Authors: Seonghyun Jin, Jong Chul Ye Published: 2026-03-19 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.18493v1 Streaming 3D reconstruction maintains a persistent latent state that is updated online from incoming frames, enabling constant-memory inference. A key failure mode is the state update rule: aggressive overwrites forget useful history, while conservative updates fail to track new evidence, and both behaviors become unstable beyond the training horizon. To address this challenge, we propose FILT3R, a training-free latent filtering layer that casts recurrent state updates as stochastic state estimation in token space. FILT3R maintains a per-token variance and computes a Kalman-style gain that adaptively balances memory retention against new observations. Process noise -- governing how much the latent state is expected to change between frames -- is estimated online from EMA-normalized temporal drift of candidate tokens. Using extensive experiments, we demonstrate that FILT3R yields an interpretable, plug-in update rule that generalizes common overwrite and gating policies as special cases. Specifically, we show that gains shrink in stable regimes as uncertainty contracts with accumulated evidence, and rise when genuine scene change increases process uncertainty, improving long-horizon stability for depth, pose, and 3D reconstruction, compared to the existing methods. Code will be released at https://github.com/jinotter3/FILT3R. ---