# Research Papers — 2026-05 · bundle page 299 of 326 > Full-text markdown of papers #14901–14950 of 16266 published in 2026-05, newest first. AI-readable mirrors of the full 16266-paper 2026-05 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-05/page/299 [← Previous bundle](https://aipod.dev/papers/archive/2026-05/page/298.md) [Next bundle →](https://aipod.dev/papers/archive/2026-05/page/300.md) ### Local Truncation Error-Guided Neural ODEs for Large Scale Traffic Forecasting Authors: Xiao Zhang, Yafei Li, Ruixiang Wang, Wei Wei, Shuo He, Mingliang Xu Published: 2026-05-05 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.03386v1 Spatiotemporal forecasting in physical systems, such as large-scale traffic networks, requires modeling a dual dynamic: continuous macroscopic rhythms and discrete, unpredictable microscopic shocks. While Neural Ordinary Differential Equations (ODEs) excel at capturing smooth evolution, their inherent Lipschitz continuity constraints inevitably cause severe over-smoothing when confronting abrupt anomalies. Recent physics-informed methods attempt to bypass this by penalizing numerical integration errors to enforce manifold smoothness. However, we mathematically reveal that such rigid regularization inherently triggers gradient conflicts and ``attention collapse,'' stripping the model of its sensitivity to anomalies. To resolve this continuity-shock dilemma, we propose Local Truncation Error-Guided Neural ODEs (LTE-ODE). Rather than treating numerical error as a nuisance to be eliminated, we innovatively repurpose the Local Truncation Error (LTE) as an unsupervised forward inductive bias. By mapping the LTE into a dynamic spatial attention mask, our architecture gracefully preserves high-precision continuous ODE evolution in stable regions, while adaptively triggering a discrete compensation branch exclusively at shock points. Trained purely end-to-end without manifold penalties, LTE-ODE achieves state-of-the-art performance on multiple large-scale benchmarks, exhibiting exceptional robustness against highly non-linear fluctuations. Furthermore, our ablation on integration steps demonstrates high deployment flexibility, allowing the model to seamlessly adapt to varying hardware memory constraints in real-world applications. --- ### GeoDecider: An Evidence-Grounded Agent for Geological Interpretation via Deliberative Reasoning Authors: Xiaoyu Tao, Mingyue Cheng, Jiahao Wang, Yitong Zhou, Qingyang Mao, Yimin Dou, Qi Liu, Shijin Wang, Enhong Chen Published: 2026-05-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.03383v2 Geological interpretation infers subsurface properties and structures from indirect geophysical observations. Well-log classification provides a measurable setting by assigning geological classes to depth-indexed petrophysical records. The task is difficult because different subsurface units may exhibit similar logging responses, whereas accurate interpretation often depends on local measurements, depth-wise context, domain knowledge, and reasonable transitions between neighboring layers. Existing automated methods mainly follow fixed prediction pipelines, leaving little room to gather additional evidence for difficult samples. In this work, we propose GeoDecider, an evidence-grounded agent for deliberative geological interpretation. GeoDecider retains efficient numerical prediction as the first stage, then selectively invokes tool-assisted reasoning for difficult intervals. A lightweight classifier produces point-wise predictions and estimates sample difficulty from its prediction scores. High-difficulty points act as routing anchors, triggering interval-level analysis so that nearby observations can be examined together. For each activated interval, specialized tools build an Evidence Profile that summarizes geological knowledge, depth-wise trends, previous predictions from the same well, and similar cases retrieved from training wells. Three complementary scientific views generate candidate interpretations. GeoDecider compares their supporting evidence, resolves disagreements, then applies geology-informed checks on continuity, boundary cues, and petrophysical consistency. Experiments on four public well-log benchmarks show that GeoDecider consistently outperforms representative baselines, demonstrating the value of selective evidence gathering and deliberative reasoning for geological interpretation.~\footnote{Our code is available at https://github.com/Xiaoyu-Tao/GeoDecider} --- ### Two Calls, Two Moments, and the Vote-Accuracy Curve of Repeated LLM Inference Authors: Yi Liu Published: 2026-05-05 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2605.03379v2 Repeated sampling is a standard way to spend test-time compute, but its benefit is controlled by the latent distribution of correctness across examples, not by one-call accuracy alone. We study the binary correctness layer of repeated LLM inference under conditional-i.i.d. calls. One labeled call identifies the mean latent success probability; two labeled calls identify its second moment and hence the same-example correctness correlation that separates stable errors from recoverable call-level randomness. From these two moments, every fixed majority-vote budget has a sharp distribution-free two-call interval. The key technical reduction is that the infinite-dimensional moment problem has three-atom extremizers and quadratic dual certificates for every finite budget, so the bounds are exact rather than discretized or parametric. The first useful budget, three votes, has a closed form, width at most $1/8$, and a certified-improvement criterion. The infinite-vote endpoint is the limit of majority voting as the number of calls tends to infinity; it is also sharply bounded, but remains threshold-sensitive because it depends on latent mass around $q=1/2$. We add maximum-entropy and Latent-difficulty Gaussian-probit point completions, and experiments on LLM calls over QNLI and QQP show that empirical three- and five-vote accuracies are contained in the projected two-call regions while temperature changes and randomized model mixtures can create voting gains not ordered by one-call accuracy. --- ### GRAFT: Auditing Graph Neural Networks via Global Feature Attribution Authors: Rishi Raj Sahoo, Subhankar Mishra Published: 2026-05-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.03377v1 Graph Neural Networks (GNNs) achieve strong performance on node classification tasks but remain difficult to interpret, particularly with respect to which input features drive their predictions. Existing global GNN explainers operate at the structural level identifying recurring subgraph motifs, but none explain model behaviour globally at the level of input node attributes. We propose GRAFT, a posthoc global explanation framework that identifies class-level feature importance profiles for GNNs. The method combines diversity-guided exemplar selection, Integrated Gradients-based attribution, and aggregation to construct a global view of feature influence for each class, which can be further expressed as concise natural language rules using a large language model with self-refinement. We evaluate GRAFT across multiple datasets, architectures, and experimental settings, demonstrating its effectiveness in capturing model-relevant features, supporting bias analysis, and enabling feature-efficient transfer learning. In addition, we introduce a structured human evaluation protocol to assess the interpretability of generated rules along dimensions such as accuracy and usefulness. Our results suggest that GRAFT provides a practical and interpretable approach for analysing feature-level behaviour in GNNs, bridging quantitative attribution with human-understandable explanations. --- ### Learning Dynamics of Zeroth-Order Optimization: A Kernel Perspective Authors: Zhe Li, Bicheng Ying, Zidong Liu, Haibo Yang Published: 2026-05-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.03373v1 Classical optimization theory establishes that zeroth-order (ZO) algorithms suffer from a dimension-dependent slowdown, with convergence rates typically scaling with the model dimension compared to first-order methods. However, in contrast to these theoretical expectations, a growing body of recent work demonstrates the successful application of ZO methods to fine-tuning Large Language Models (LLMs) with billions of parameters. To explain this paradox, we derive the one-step learning dynamics of ZO SGD, where the empirical Neural Tangent Kernel (eNTK) naturally emerges as the key term governing the learning behavior. Inspection of the eNTK produced by ZO SGD reveals that each element corresponds to the inner product of neural tangent vectors projected onto a random low-dimensional subspace. Thus, by invoking the Johnson-Lindenstrauss Lemma, our analysis shows that the fidelity of the ZO eNTK is governed primarily by the number of perturbations. Crucially, the approximation error depends on the model output size rather than the massive parameter dimension. This dimension-free property provides a theoretical justification for the scalability of ZO methods to LLMs finetuning tasks. We believe that this kernel-based framework offers a novel perspective for understanding ZO methods within the context of learning dynamics. --- ### Fully Automatic Trace Gas Plume Detection Authors: Vít Růžička, David R. Thompson, Jay E. Fahlen, Amanda M. Lopez, Steven Lu, Chuchu Xiang, Holly Bender, Daniel Jensen, Philip G. Brodrick, Jake Lee, Brian Bue, Daniel H. Cusworth et al. Published: 2026-05-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.03372v2 Future imaging spectrometers will expand contemporary data volumes by orders of magnitude, requiring automated methods to upscale labor-intensive detection of trace gas point sources. Here we present a fully-automated approach that achieves operational performance for plume detection and labelling without human participation. Our method combines machine learning (ML)-based morphological analysis with physics-based spectroscopic model fitting. We deploy it on data from the EMIT imaging spectrometer, operating in two modes. First, we present a "daily digest" that runs automatically on all downlinked data, flagging the largest events for immediate response. The daily digest demonstrates that a significant fraction of the largest plumes can be detected automatically with negligible false positives. This represents a significant new high-water mark in plume detection accuracy. Second, we use it for retrospective analysis to find plumes that were missed by the existing human review process. We observe that at least 25% of large plumes may have been passed over in the existing workflow due to confirmation bias and ambiguity in the visual cues used by human reviewers. Finally, we extend detection to three understudied trace gases: NH3, NO2 and the first observations of carbon monoxide (CO) plume in EMIT imagery. --- ### SoDa2: Single-Stage Open-Set Domain Adaptation via Decoupled Alignment for Cross-Scene Hyperspectral Image Classification Authors: Yiwen Liu, Minghua Wang, Jing Yao, Xin Zhao, Gemine Vivone Published: 2026-05-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.03371v1 Cross-scene hyperspectral image (HSI) classification stands as a fundamental research topic in remote sensing, with extensive applications spanning various fields. Owing to the inclusion of unknown categories in the target domain and the existence of domain shift across different scenes, open-set domain adaptation techniques are commonly employed to address cross-scene HSI classification. However, existing open-set cross-scene HSI classification methods still face two critical challenges: (1) domain shift issues arising from the direct alignment of mixed spectral-spatial features; (2) high computational costs caused by two-stage training strategies. To address these issues, this paper proposes a single-stage open-set domain adaptation method with decoupled alignment (SoDa$^2$) for cross-scene HSI classification. A contribution-aware dual-modality feature extraction is customized to disentangle the characteristics from spectral sequence signals and spatial details, selectively and adaptively enhancing discriminative features. The decoupled alignment module minimizes the Maximum Mean Discrepancy to independently reduce the spectral discrepancy and the spatial discrepancy between the source and target domains, extracting more fine-grained domain-invariant features. A cost-effective single-stage dual-branch framework is designed to learn MMD-constrainted aligned features and constraint-free intrinsic features for adaptive distinction between known and unknown classes. This framework employs a Gaussian Mixture Model to model the squared cosine similarity distribution between the two feature types, enabling open-set recognition without prior knowledge of unknown classes. Extensive experiments on three groups of HSI datasets demonstrate that SoDa$^2$ outperforms state-of-the-art methods, achieving superior classification accuracy and model transferability for open-set cross-scene tasks. --- ### Dual-Foundation Models for Unsupervised Domain Adaptation Authors: Yerin Cheon, Aruna Balasubramanian, Francois Rameau Published: 2026-05-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.03365v1 Semantic segmentation provides pixel-level scene understanding essential for autonomous driving and fine-grained perception tasks. However, training segmentation models requires costly, labor-intensive annotations on real-world datasets. Unsupervised Domain Adaptation (UDA) addresses this by training models on labeled synthetic data and adapting them to unlabeled real images. While conceptually simple, adaptation is challenging due to the domain gap, i.e., differences in visual appearance and scene structure between synthetic and real data. Prior approaches bridge this gap through pixel-level mixing or feature-level contrastive learning. Yet, these techniques suffer from two major limitations: (1) reliance on high-confidence pseudo-labels restricts learning to a subset of the target domain, and (2) prototype-based contrastive methods initialize class prototypes from source-trained models, yielding biased and unstable anchors during adaptation. To address these issues, we propose a dual-foundation UDA framework that leverages two complementary foundation models. First, we employ the Segment Anything Model (SAM) with superpixel-guided prompting to enable learning from a broader range of target pixels beyond high-confidence predictions. Second, we incorporate DINOv3 to construct stable, domain-invariant class prototypes through its robust representation learning. Our method achieves consistent improvements of +1.3% and +1.4% mIoU over strong UDA baselines on GTA-to-Cityscapes and SYNTHIA-to-Cityscapes, respectively. --- ### Dynamic Distillation and Gradient Consistency for Robust Long-Tailed Incremental Learning Authors: Taigo Sakai, Kazuhiro Hotta Published: 2026-05-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.03364v1 The task of Long-tailed Class Incremental Learning (LT-CIL) addresses the sequential learning of new classes from datasets with imbalanced class distributions. This scenario intensifies the fundamental problem of catastrophic forgetting, inherent to continual learning, with the dual challenges of under-learning minority classes and overfitting majority classes. To tackle these combined issues, this paper proposes two main techniques. First, we introduce gradient consistency regularization, which leverages the moving average of gradients to suppress abrupt fluctuations and stabilize the training process. Second, we dynamically adjust the weight of the distillation loss by measuring the degree of class imbalance with normalized entropy. This adaptive weighting establishes an optimal balance between retaining old knowledge and acquiring new information. Experiments on the CIFAR-100-LT, ImageNetSubset-LT, and Food101-LT benchmarks show that our method achieves consistent accuracy improvements of up to 5.0\%. Furthermore, we demonstrate dramatic gains in the challenging 'In-ordered' setting, where tasks progress from majority to minority classes, highlighting our method's robustness in mitigating forgetting under unfavorable learning dynamics. This enhanced performance is achieved without a significant increase in computational overhead, demonstrating the practicality of our framework. --- ### ReasonAudio: A Benchmark for Evaluating Reasoning Beyond Matching in Text-Audio Retrieval Authors: Honglei Zhang, Yuting Chen, Chenpeng Hu, Pengfei Zhou, Siyue Zhang, Yilei Shi Published: 2026-05-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.03361v3 Existing audio retrieval benchmarks primarily assess semantic matching, lacking comprehensive evaluation of the logical reasoning capabilities required by complex queries. We introduce ReasonAudio, a benchmark for reasoning-intensive Text-Audio Retrieval that evaluates four abilities: negation, temporal order, sound co-occurrence, and sound duration. It comprises five synthetic subtasks with 1,000 queries over 10,000 composite audio clips and a natural subtask with 100 queries over 1,000 real-world clips. Evaluation of 11 state-of-the-art retrieval systems reveals substantial limitations: the best-performing model, OmniEmbed-7B, achieves an overall score of 20.7. In a controlled experiment designed to reduce the influence of sound-event matching, OmniEmbed-7B attains 53.8% average accuracy, compared with 70.6% for its generative backbone, Qwen2.5-Omni-7B-Thinker, and 95.6% for humans. Our results highlight the challenges of reasoning-intensive audio retrieval and reveal a performance gap between OmniEmbed-7B and its generative backbone. --- ### ReasonAudio: A Benchmark for Evaluating Reasoning Beyond Matching in Text-Audio Retrieval Authors: Honglei Zhang, Yuting Chen, Chenpeng Hu, Siyue Zhang, Yilei Shi Published: 2026-05-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.03361v2 As multimodal content continues to expand at a rapid pace, audio retrieval has emerged as a key enabling technology for media search, content organization, and intelligent assistants. However, most existing benchmarks concentrate on semantic matching and fail to capture the fact that real-world queries often demand advanced reasoning abilities, including negation understanding, temporal ordering, concurrent event recognition, and duration discrimination. To address this gap, we introduce ReasonAudio, the first reasoning-intensive benchmark for Text-Audio Retrieval, comprising 1,000 queries and 10,000 composite audio clips across five fundamental reasoning tasks: Negation, Order, Overlap, Duration, and Mix. Despite their intuitive nature for humans and straightforward construction, these tasks pose significant challenges to current models. Our evaluation of ten state-of-the-art models reveals the following findings: All models struggle with reasoning-intensive audio retrieval, performing particularly poorly on Negation and Duration while showing relatively better results on Overlap and Order. Moreover, Multimodal Large Language Model-based embedding models fail to inherit the reasoning capabilities of their backbones through contrastive fine-tuning, suggesting that current training paradigms are insufficient to preserve reasoning capacity in retrieval settings --- ### A-CODE: Fully Atomic Protein Co-Design with Unified Multimodal Diffusion Authors: Chaoran Cheng, Jiaqi Guan, Milong Ren, Chengyue Gong, Cong Liu, Xinshi Chen, Ge Liu, Wenzhi Xiao Published: 2026-05-05 Categories: q-bio.QM, cs.LG arXiv: https://arxiv.org/abs/2605.03360v1 We present A-CODE, a fully atomic unified one-stage protein co-design model that simultaneously refines discrete atom types and continuous atom coordinates. Unlike predominant two-stage methods that cascade structure design with amino acid-level sequence design, our approach is fully atomic within a unified multimodal diffusion framework, in which residue identities are inferred solely from atom-level predictions. Built upon the powerful all-atom architecture, A-CODE achieves superior designability for unconditional protein generation, outperforming all existing one-stage and two-stage design models. For binder design, A-CODE rivals and even outperforms existing state-of-the-art two-stage design models and, compared with the existing one-stage co-design model, achieves a drastic tenfold improvement in success rate on hard tasks. The inherent flexibility of our atomic formulation enables, for the first time, seamless adaptation to non-canonical amino acid (ncAA) modeling. Our fully atomic framework establishes a new, versatile foundation for all-atom generative modeling that can be naturally extended to complex biomolecular systems. --- ### Mix3R: Mixing Feed-forward Reconstruction and Generative 3D Priors for Joint Multi-view Aligned 3D Reconstruction and Pose Estimation Authors: Siyou Lin, Zhou Xue, Hongwen Zhang, Liang An, Dongping Li, Shaohui Jiao, Yebin Liu Published: 2026-05-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.03359v1 Recent trends in sparse-view 3D reconstruction have taken two different paths: feed-forward reconstruction that predicts pixel-aligned point maps without a complete geometry, and generative 3D reconstruction that generates complete geometry but often with poor input-alignment. We present Mix3R, a novel generative 3D reconstruction method which mixes feed-forward reconstruction and 3D generation into a single framework in an aligned manner. Mix3R generates a 3D shape in two stages: a sparse voxel generation stage and a textured geometry generation stage. Unlike pure generative methods, our first-stage generation jointly produces a coarse 3D structure (sparse voxels), per-view point maps and camera parameters aligned to that 3D structure. This is made possible by introducing a Mixture-of-Transformers architecture that inserts global self-attentions to a feed-forward reconstruction model and a 3D generative model, both pretrained on large-scale data. This design effectively retains the pretrained priors but enables better 2D-3D alignment. Based on the initial aligned generations of sparse 3D voxels and point maps, we compute an overlap-based attention bias that is directly added to another pretrained textured geometry generation model, enabling it to correctly place input textures onto generated shapes in a training-free manner. Our design brings mutual benefits to both feed-forward reconstruction and 3D generation: The feed-forward branch learns to ground its predictions to a generative 3D prior, and conversely, the 3D generation branch is conditioned on geometrically informative features from the feed-forward branch. As a result, our method produces 3D shapes with better input alignment compared with pure 3D generative methods, together with camera pose estimations more accurate than previous feed-forward reconstruction methods. Our project page is at https://jsnln.github.io/mix3r/ --- ### Tracing Like a Clinician: Anatomy-Guided Spatial Priors for Cephalometric Landmark Detection Authors: Sidhartha Mohapatra, Pallavi Mohanty Published: 2026-05-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.03358v3 Clinicians trace cephalometric radiographs following a structured anatomical workflow, yet no prior system encodes this into computation. We present a five-phase anatomy-guided pipeline producing confidence-weighted spatial priors that shape HRNet-W32 training, achieving 1.04 mm mean radial error on 25 landmarks across 1,502 radiographs from 7+ imaging devices. A training x inference prior matrix isolates the mechanism: anatomical priors maintain a 1% validation-to-test gap versus 88% without priors (1.94 mm), despite identical validation convergence. The matrix establishes that all trained models are inference-independent, the expanded architecture alone provides no benefit, random priors yield partial but unstable improvement (1.72 mm), and only image-specific anatomically correct priors produce the 1.04 mm result -- functioning as a training-time regularizer requiring no automated prior generation at deployment. Five-fold cross-validation (p=0.0015), patient-level permutation testing (p<0.0001, n=151), quantified Grad-CAM analysis (88% vs. 74% in-zone activation, p<0.001), and clinical measurement validation (skeletal classification kappa=0.79-0.84, zero Class II<->III reversals, ICC>0.95) provide converging evidence. Cross-domain experiments on echocardiography, cervical spine, and hand radiography support the hypothesis that prior effectiveness scales with the spatial entropy of the landmark distribution. --- ### Population-Aware Imitation Learning in Mean-field Games with Common Noise Authors: Grégoire Lambrecht, Mathieu Laurière Published: 2026-05-05 Categories: cs.LG, math.OC arXiv: https://arxiv.org/abs/2605.03357v2 Mean Field Games (MFGs) provide a powerful framework for modeling the collective behavior of large populations of interacting agents. In this paper, we address the problem of Imitation Learning (IL) in MFGs subject to common noise, where the population distribution evolves stochastically. This stochasticity compels agents to adopt population-aware policies to respond to aggregate shocks. We formulate two distinct learning objectives: recovering a Nash equilibrium and maximizing performance against an expert population. We investigate two imitation proxies: Behavioral Cloning (BC) and Adversarial (ADV) divergence. We then establish finite-sample error bounds showing that minimizing these proxies effectively controls both the policy's exploitability and its performance gap relative to the expert. Furthermore, we propose a numerical framework using generalized Fictitious Play and Deep Learning to compute expert population-aware policies. Through experiments on three environments we demonstrate that standard population-unaware policies fail to capture the equilibrium dynamics. Our results highlight that learning population-aware policies is crucial to avoid being misled by the randomness inherent in common noise. --- ### What Happens Inside Agent Memory? Circuit Analysis from Emergence to Diagnosis Authors: Xutao Mao, Jinman Zhao, Gerald Penn, Cong Wang Published: 2026-05-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.03354v3 Agent memory failures are silent: an LLM-based agent can produce a fluent response even when it fails to extract, retain, or retrieve the information needed across sessions. The write-manage-read loop describes the external pipeline of these systems but leaves open which internal computations implement each stage. We trace feature circuits across the Qwen-3 family (0.6B--14B), two memory frameworks (mem0, A-MEM), and a matched Gemma-3 replication, reporting two mechanistic findings plus one deliverable. First, under mem0, control is detectable before content. Routing circuitry is causally active at 0.6B, while content circuitry produces no detectable signal until 4B. Gemma-3 reproduces the same order at 1B and 4B. Small models therefore route memory decisions before they can reliably extract or ground the underlying facts. Second, the shared hub is recruited, not created. Write and Read converge on a late-layer hub that already exists in the base model as a context-grounding substrate, on which memory framing recruits a memory-specific functional direction. The hub and the early routing circuit carry over to A-MEM, indicating that these computations belong to the base model rather than to a particular interface. Building on this circuit structure, we develop an unsupervised stage-level diagnostic that localizes silent failures to the responsible operation at 76.2% accuracy. It outperforms a strong LLM trace judge by 3.1 points as well as a supervised hidden-state probe by 8.8 points. Together, these results establish circuit-level signatures as a practical handle for monitoring agent memory and guiding its structural design. --- ### What Happens Inside Agent Memory? Circuit Analysis from Emergence to Diagnosis Authors: Xutao Mao, Jinman Zhao, Gerald Penn, Cong Wang Published: 2026-05-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.03354v2 Agent memory failures are silent: an LLM-based agent can produce a fluent response even when it fails to extract, retain, or retrieve the information needed across sessions. The write-manage-read loop describes the external pipeline of these systems but leaves open which internal computations implement each stage. Tracing feature circuits across the Qwen-3 family (0.6B--14B) and two memory frameworks (mem0 and A-MEM), we report two mechanistic findings and one deliverable. First, control is detectable before content: routing circuitry is causally active at 0.6B, while content circuitry produces no detectable signal until 4B, exposing a deployment regime where small models route memory decisions before they can reliably extract or ground the underlying facts. Second, the shared hub is recruited, not created: Write and Read converge on a late-layer hub that already exists in the base model as a context-grounding substrate, and memory framing recruits a memory-specific functional direction on this substrate rather than building one of its own. Both findings transfer across mem0 and A-MEM, indicating that the underlying computations are properties of the base model rather than of any particular interface. Building on this circuit structure, we develop an unsupervised stage-level diagnostic that localizes silent failures to the responsible operation up to 76.2% accuracy, outperforming the strongest supervised baseline by 13 points. Together, these results point to circuit-level signatures as a practical handle for monitoring and structurally-guided design of agent memory. --- ### SkCC: Portable and Secure Skill Compilation for Cross-Framework LLM Agents Authors: Yipeng Ouyang, Yi Xiao, Yuhao Gu, Xianwei Zhang Published: 2026-05-05 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2605.03353v4 LLM agents increasingly rely on reusable skills (e.g., SKILL markdown files) to execute complex tasks, yet these artifacts lack portability: agent frameworks are highly sensitive to prompt formatting, leading to a large performance variation for the same skill. Nevertheless, most skills are authored once as format-agnostic Markdown, necessitating costly per-framework rewrites and also leaving security largely unaddressed, with widespread vulnerabilities in practice. To address this, we present SkCC, a compiler for LLM agents that introduces classical compilation design into agent skill development. SkCC centers on SkIR, a strongly-typed intermediate representation that decouples skill semantics from framework-specific formatting, thus enabling portable deployment across agent frameworks. Atop of this IR, a static Optimizer enforces security constraints, blocking vulnerabilities before deployment. Implemented as a four-phase pipeline, SkCC effectively reduces adaptation complexity from $O(m \times n)$ to $O(m + n)$ across $m$ skills and $n$ frameworks. Experiments on SkillsBench demonstrate that SkCC delivers consistent and substantial gains over original counterparts, with pass rate increases from 21.1% to 33.3% on Claude Code and from 35.1% to 48.7% on Kimi CLI. Further, the design achieves sub-10ms compilation latency, 94.8% proactive security trigger rate, and 10-46% runtime token savings across frameworks. --- ### Can Multimodal Large Language Models Understand Pathologic Movements? A Pilot Study on Seizure Semiology Authors: Lina Zhang, Tonmoy Monsoor, Mehmet Efe Lorasdagi, Prateik Sinha, Chong Han, Peizheng Li, Yuan Wang, Jessica Pasqua, Colin McCrimmon, Rajarshi Mazumder, Vwani Roychowdhury Published: 2026-05-05 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.03352v1 Multimodal Large Language Models (MLLMs) have demonstrated robust capabilities in recognizing everyday human activities, yet their potential for analyzing clinically significant involuntary movements in neurological disorders remains largely unexplored. This pilot study evaluates the capability of MLLMs for automated recognition of pathological movements in seizure videos. We assessed the zero-shot performance of state-of-the-art MLLMs on 20 ILAE-defined semiological features across 90 clinical seizure recordings. MLLMs outperformed fine-tuned Convolutional Neural Network (CNN) and Vision Transformer (ViT) baseline models on 13 of 18 features without task-specific training, demonstrating particular strength in recognizing salient postural and contextual features while struggling with subtle, high-frequency movements. Feature-targeted signal enhancement (facial cropping, pose estimation, audio denoising) improved performance on 10 of 20 features. Expert evaluation showed that 94.3 percent of MLLM-generated explanations for correctly predicted cases achieved at least 60 percent faithfulness scores, aligning with epileptologist reasoning. These findings demonstrate the potential of adapting general-purpose MLLMs for specialized clinical video analysis through targeted preprocessing strategies, offering a path toward interpretable, efficient diagnostic assistance. Our code is publicly available at https://github.com/LinaZhangUCLA/PathMotionMLLM. --- ### VLMaxxing through FrameMogging Training-Free Anti-Recomputation for Video Vision-Language Models Authors: JF Bastien, Sam D'Amico Published: 2026-05-05 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.03351v1 Video vision-language models (VLMs) keep paying for visual state the stream already told us was stable. The factory wall did not move, but most VLM pipelines still hand the model dense RGB frames or a fresh prefix again. We study that waste as training-free anti-recomputation: reuse state when validation says it survives, and buy fresh evidence when the scene, query, or cache topology requires it. The largest measured win is after ingest. On frozen Qwen2.5-VL-7B-Instruct-4bit, adaptive same-video follow-up reuse preserves paired choices and correctness on a 93-query VideoMME breadth setting while reducing follow-up latency by 14.90-35.92x. The first query is still cold; the win starts when later questions reuse the same video state. Stress tests bound the result: repeated-question schedules hold through 50 turns, while dense-answer-anchored prompt variation separates conservative fixed K=1 repair from faster aggressive policies that drift. Fresh-video pruning is smaller but real. C-VISION skips timed vision-tower work before the first answer is generated. On Gemma 4-E4B-4bit, the clean 32f short cell reaches 1.316x first-query speedup with no paired drift or parse failures on 20 items; Qwen shows the fidelity/speed boundary. Stage-share ceiling (C-CEILING) is the accounting guardrail: a component speedup becomes an end-to-end speedup only in proportion to the wall-clock share it accelerates, so C-VISION and after-ingest follow-up reuse do not multiply. Candidate C-STREAM remains a native-rate target, not a headline result here. The broader direction is VLM-native media that expose change, motion, uncertainty, object state, sensor time, and active tiles directly, so models do not have to rediscover the world from dense RGB every frame. --- ### Toward Structural Multimodal Representations: Specialization, Selection, and Sparsification via Mixture-of-Experts Authors: Hahyeon Choi, Nojun Kwak Published: 2026-05-05 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.03348v2 We propose S3 (Specialization, Selection, Sparsification), a framework that rethinks multimodal learning through a structural perspective. Instead of encoding all signals into a fixed embedding, S3 decomposes multimodal inputs into semantic experts and selectively routes them for each task. Specialization forms concept-level experts in a shared latent space, Selection adapts routing for task-specific needs, and Sparsification prunes low-utility paths to yield compact, information-minimal representations. Across four MultiBench benchmarks, S3 improves accuracy and shows a consistent reverse U-shaped sparsity-performance trend, with peak performance at intermediate sparsity. These results suggest that structuring multimodal representations as selectable semantic components provides a practical and principled alternative to contrastive learning or InfoMax-driven approaches. --- ### Provable Accuracy Collapse in Embedding-Based Representations under Dimensionality Mismatch Authors: Dionysis Arvanitakis, Vaggos Chatziafratis, Yiyuan Luo Published: 2026-05-05 Categories: cs.DS, cs.LG arXiv: https://arxiv.org/abs/2605.03346v1 Embedding-based representations in Euclidean space $\mathbb{R}^d$ are a cornerstone of modern machine learning, where a major goal is to use the \emph{smallest dimension} that faithfully captures data relations. In this work, we prove sharp dimension--accuracy tradeoffs and identify a fundamental information-theoretic limitation: unless the embedding dimension $d$ is chosen close to the ground-truth dimension $D$, accuracy undergoes a sudden collapse. Our main result shows that this phenomenon arises even in standard contrastive learning settings, where supervision is limited to a set of $m$ anchor--positive--negative triplets $(i,j,k)$ encoding distance comparisons $\mathrm{dist}(i,j) < \mathrm{dist}(i,k)$. Specifically, given triplets realizable by an unknown ground-truth embedding in $D$ dimensions, we prove that there exists constant $c < 1$, such that \emph{every embedding of dimension at most $cD$ violates half of the triplets}, yielding accuracy as low as a trivial one-dimensional solution that ignores the input. We complement our information-theoretic bounds with strong computational hardness results: under the Unique Games Conjecture, even if the given triplets are nearly realizable in $D=1$ dimension, no polynomial-time algorithm -- \textit{regardless of its dimension} -- can achieve accuracy above the trivial $50\%$ baseline. --- ### RAG over Thinking Traces Can Improve Reasoning Tasks Authors: Negar Arabzadeh, Wenjie Ma, Sewon Min, Matei Zaharia Published: 2026-05-05 Categories: cs.IR, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2605.03344v3 Retrieval-augmented generation (RAG) has proven effective for knowledge-intensive tasks, but is widely believed to offer limited benefit for reasoning-intensive problems such as math and code generation. We challenge this assumption by showing that the limitation lies not in RAG itself, but in the choice of corpus. Instead of retrieving documents, we propose retrieving thinking traces, i.e., intermediate thinking trajectories generated during problem solving attempts. We show that thinking traces are already a strong retrieval source, and further introduce T3, an offline method that transforms them into structured, retrieval-friendly representations, to improve usability. Using these traces as a corpus, a simple retrieve-then-generate pipeline consistently improves reasoning performance across strong models and benchmarks such as AIME 2025--2026, LiveCodeBench, and GPQA-Diamond, outperforming both non-RAG baselines and retrieval over standard web corpora. For instance, on AIME 2025-2026, RAG with traces generated by Gemini-2-thinking achieves relative gains of +56.3%, +8.6%, and +7.6% for Gemini-2.5-Flash, GPT-OSS-120B, and GPT-5, respectively, even though these are more recent models. Overall, our results suggest that thinking traces are an effective retrieval corpus for reasoning tasks, and transforming them into structured, compact, or diagnostic representations unlocks even stronger gains. Code available at https://github.com/Narabzad/t3. --- ### RAG over Thinking Traces Can Improve Reasoning Tasks Authors: Negar Arabzadeh, Wenjie Ma, Sewon Min, Matei Zaharia Published: 2026-05-05 Categories: cs.IR, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2605.03344v2 Retrieval-augmented generation (RAG) has proven effective for knowledge-intensive tasks, but is widely believed to offer limited benefit for reasoning-intensive problems such as math and code generation. We challenge this assumption by showing that the limitation lies not in RAG itself, but in the choice of corpus. Instead of retrieving documents, we propose retrieving thinking traces, i.e., intermediate thinking trajectories generated during problem solving attempts. We show that thinking traces are already a strong retrieval source, and further introduce T3, an offline method that transforms them into structured, retrieval-friendly representations, to improve usability. Using these traces as a corpus, a simple retrieve-then-generate pipeline consistently improves reasoning performance across strong models and benchmarks such as AIME 2025--2026, LiveCodeBench, and GPQA-Diamond, outperforming both non-RAG baselines and retrieval over standard web corpora. For instance, on AIME 2025-2026, RAG with traces generated by Gemini-2-thinking achieves relative gains of +56.3%, +8.6%, and +7.6% for Gemini-2.5-Flash, GPT-OSS-120B, and GPT-5, respectively, even though these are more recent models. Overall, our results suggest that thinking traces are an effective retrieval corpus for reasoning tasks, and transforming them into structured, compact, or diagnostic representations unlocks even stronger gains. Code available at https://github.com/Narabzad/t3. --- ### MedSR-Vision: Deep Learning Framework for Multi-Domain Medical Image Super-Resolution Authors: Subhash Gurappa, Trivikram Satharasi, Yashas Hariprasad, Sundararaj Sitharama Iyengar Published: 2026-05-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.03343v1 Medical image super-resolution (MedSR) is essential for improving diagnostic precision across diverse imaging modalities such as MRI, CT, X-ray, Ultrasound, and Fundus imaging. Despite rapid advances in deep learning, challenges remain in preserving anatomical accuracy, maintaining perceptual quality, and generalizing across medical domains. This paper presents MedSR-Vision, a novel unified deep learning framework for evaluating and comparing super-resolution models across five modalities: Brain MRI, Chest X-ray, Renal Ultrasound, Nephrolithiasis CT, and Spine MRI, at magnification scales of $\times2$, $\times3$, and $\times4$. Three representative models namely SRCNN, SwinIR, and Real-ESRGAN are benchmarked using multiple quantitative metrics encompassing fidelity, perceptual realism, and sharpness. Experimental analysis demonstrates that Real-ESRGAN achieves superior perceptual quality and edge recovery at higher scales, SwinIR excels in preserving structural and diagnostic features, and SRCNN provides efficient and stable performance at lower magnifications. The results establish domain-specific insights and practical guidelines for model selection in clinical imaging workflows, offering a standardized evaluation framework for future medical image super-resolution research and deployment. --- ### Automated Large-scale CVRP Solver Design via LLM-assisted Flexible MCTS Authors: Tong Guo, Caishun Chen, Yew Soon Ong Published: 2026-05-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.03339v1 Solving large-scale CVRP (LSCVRP) with hundreds to thousands of nodes remains difficult for even state-of-the-art solvers. Divide-and-conquer can scale by decomposing the instance into size-reduced subproblems, but designing decomposition logic and configuring sub-solvers is highly expertise- and labor-intensive. Large Language Models (LLMs) have emerged as promising tools for automated algorithm design. However, existing LLM-driven approaches struggle with LSCVRP primarily due to the difficulty in generating sophisticated search strategies within a limited context window. To bridge this gap, we propose the LLM-assisted Flexible Monte Carlo Tree Search (LaF-MCTS), a novel framework that automates the design of high-performance LSCVRP solvers. We develop a three-tier decision hierarchy to enable incremental design of decomposition policies and sub-solvers for LSCVRP. To enable efficient search within the algorithmic hypothesis space, we introduce semantic pruning to eliminate semantically and structurally redundant codes, and branch regrowth to regenerate codes and preserve diversity. Extensive experiments on CVRPLib demonstrate that LaF-MCTS autonomously composes and optimizes decomposition-enhanced solvers that surpasses various state-of-the-art CVRP solvers. --- ### Symmetry, Defects, and Diffusion in Continuous-memory Recurrent Networks Authors: Hanson Hanxuan Mo Published: 2026-05-05 Categories: cs.NE, math.DS arXiv: https://arxiv.org/abs/2605.03338v2 Continuous-memory recurrent networks must preserve phase, position, or orientation despite model imperfections and state noise. We develop a geometric framework that separates three questions: how many memory coordinates are neutrally transported, how deterministic perturbations alter their finite-horizon stability, and how ambient noise is decoded along them. Exact per-input equivariance transports analytical group tangents pathwise and, on a compact nondegenerate orbit stratum, yields at least $q=\dim(G/H)$ zero group-tangent Lyapunov exponents under stationary ergodic driving. For imperfect dynamics, a four-block tangent/normal decomposition gives local and finite-horizon bounds on tangent growth and subspace rotation, distinguishing first-order direct damage from second-order leakage through contracting normal directions. For noisy dynamics, a specified decoder maps ambient covariance $Q$ to coordinate covariance $ZQZ^\top$; under isotropic noise and fixed tangent energy, least-squares decoding and scaled-isometric action geometry minimize local diffusion. Local and finite-horizon evaluations include cases both within and outside the sufficient conditions. In a fresh twenty-seed $T^2$ replication, a decoder-covariance objective improves noisy horizon-256 memory in every pair while meeting a prespecified clean-error equivalence margin. Direct noise training also improves noisy memory but incurs a clean-error tradeoff. In coupled $T^4/T^8$ integrators, the advantage of anisotropic-covariance over isotropic regularization reverses when evaluation noise becomes isotropic. These results connect continuous symmetry to measurable limits and design choices for recurrent memory under explicit dynamical, decoder, and noise assumptions. --- ### Symmetry-Protected Lyapunov Neutral Modes in Equivariant Recurrent Networks Authors: Hanson Hanxuan Mo Published: 2026-05-05 Categories: cs.NE, math.DS arXiv: https://arxiv.org/abs/2605.03338v1 Recurrent networks that store position, phase, or other continuous variables need state-space directions that remain neutral over long horizons. We give a symmetry-based account of when such neutral directions are guaranteed rather than merely tuned. For a finite-dimensional autonomous \(C^1\) vector field equivariant under a Lie group \(G\), we prove that any compact invariant set carrying a uniformly nondegenerate group-orbit bundle with stabilizer type \(H\) has, at points where the Lyapunov spectrum is defined, at least \(\dim(G/H)\) zero Lyapunov exponents tangent to the group orbit. These symmetry-protected modes have zero group-tangent growth because of exact equivariance and orbit geometry. When this protection is explicitly broken, the formerly protected direction can acquire a pseudo-gap; in our controlled breaking experiments this pseudo-gap predicts finite memory lifetime. We verify the finite-dimensional consequences with normalized equivariance error, direct group-tangent exponents, principal-angle alignment, autonomous-flow-zero controls, and orbit-dimension scaling across \(S^1\), \(T^q\), \(SO(n)\), \(U(m)\), product-group, and coupled equivariant RNN-style systems. We also train an exactly equivariant recurrent cell on velocity-input \(S^1\) path integration across six seeds and compare it with matched GRU, LSTM, and orthogonal-RNN baselines. The learned equivariant cell preserves step equivariance to \(3.2\times10^{-8}\), has a near-zero group-tangent exponent under the zero-input autonomous restriction, and improves horizon, speed, and restricted-phase generalization in this matched protocol. The learned task results are consequence evidence; the theorem-level evidence remains exact equivariance, group-tangent exponents, orbit-dimension scaling, and tangent-subspace alignment. --- ### FreeTimeGS++: Secrets of Dynamic Gaussian Splatting and Their Principles Authors: Lucas Yunkyu Lee, Soonho Kim, Youngwook Kim, Sangmin Kim, Jaesik Park Published: 2026-05-05 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.03337v3 Recent progress in 4D Gaussian Splatting (4DGS) has achieved impressive dynamic scene reconstruction results. While these methods demonstrate remarkable performance, the specific factors behind their gains remain underexplored, making a systematic understanding of the underlying principles challenging. In this paper, we perform a comprehensive analysis of these hidden factors to provide a clearer perspective on the 4DGS framework. We first establish a controlled baseline, FreeTimeGS_ours, by formalizing and reproducing the heuristics of the state-of-the-art FreeTimeGS. Using this framework, we examine 4DGS along its fundamental axes and identify practical secrets, including the emergent temporal partitioning driven by Gaussian durations and the decoupling between photometric fidelity and motion behavior. Based on these insights, we propose FreeTimeGS++, a principled method that employs gated marginalization, UFM-guided initialization, and color correction to improve stability and reproducibility. Our approach yields reproducible results with reduced run-to-run variance. --- ### LLM-ADAM: A Generalizable LLM Agent Framework for Pre-Print Anomaly Detection in Additive Manufacturing Authors: Ahmadreza Eslaminia, Chuhan Cai, Cameron Smith, Ruo-Syuan Mei, Shichen Li, Rajiv Malhotra, Klara Nahrstedt, Chenhui Shao Published: 2026-05-05 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.03328v1 Additive manufacturing (AM) continues to transform modern manufacturing by enabling flexible, on-demand production of complex geometries across diverse industries. Fused filament fabrication (FFF) has extended AM to laboratories, classrooms, and small production environments, but this accessibility shifts process-planning responsibility to users who may lack manufacturing expertise. A syntactically valid slicer profile can still encode thermally or geometrically harmful settings, and subtle G-code edits can alter extrusion, cooling, or adhesion before a print begins. Pre-print G-code screening catches accidental or adversarial machine-program errors before material or machine time is wasted. This paper proposes LLM-ADAM as a generalizable LLM framework for pre-print anomaly detection in AM. The framework decomposes the task into three roles: Extractor-LLM maps a G-code file to a structured process-parameter schema; Reference-LLM converts printer and material documentation into aligned operating ranges; and Judge-LLM interprets a deterministic deviation table and G-code evidence to decide whether a part is non-defective or belongs to an anomaly class. Printers, materials, and LLM backbones are interchangeable test conditions, not fixed assumptions. We evaluate the framework on an N=200 FFF G-code corpus spanning two desktop printer families, two materials, and five classes including non-defective, under-extrusion, over-extrusion, warping, and stringing. The best framework configuration reaches 87.5% accuracy, compared with 59.5% for the strongest engineered single-LLM baseline. The results show that structured decomposition, rather than backbone strength alone, is the dominant source of improvement, with defect classes identified at or near ceiling for leading configurations while residual errors concentrate on conservative false alarms for non-defective samples. --- ### DGPO: Distribution Guided Policy Optimization for Fine Grained Credit Assignment Authors: Hongbo Jin, Rongpeng Zhu, Zhongjing Du, Xu Jiang, Jingqi Tian, Qiaoman Zhang, Jiayu Ding Published: 2026-05-05 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.03327v2 Reinforcement learning is crucial for aligning large language models to perform complex reasoning tasks. However, current algorithms such as Group Relative Policy Optimization suffer from coarse grained, sequence level credit assignment, which severely struggles to isolate pivotal reasoning steps within long Chain of Thought generations. Furthermore, the standard unbounded Kullback Leibler divergence penalty induces severe gradient instability and mode seeking conservatism, ultimately stifling the discovery of novel reasoning trajectories. To overcome these limitations, we introduce Distribution Guided Policy Optimization, a novel critic free reinforcement learning framework that reinterprets distribution deviation as a guiding signal rather than a rigid penalty. DGPO replaces the volatile KL divergence with the bounded Hellinger distance to safely quantify token level exploration without the risk of gradient explosion. To effectively distinguish genuine reasoning breakthroughs from hallucinatory noise, we propose an entropy gating mechanism that scales this deviation by the policy`s epistemic uncertainty. By dynamically redistributing the coarse sequence-level advantage to individual tokens based on these gated scores, DGPO heavily incentivizes critical exploratory steps while suppressing unwarranted, low-entropy deviations. Consequently, DGPO completely eliminates the traditional token-level KL penalty and achieves fine-grained credit reallocation without the computational overhead of an additional value network. Extensive empirical evaluations demonstrate that DGPO sets a new state-of-the-art for critic free alignment. Notably, on the Qwen2.5-32B architecture, DGPO achieves 60.0% Avg@32 accuracy and 46.0% Avg@32 accuracy on the challenging AIME2024 and AIME2025 benchmarks respectively, substantially outperforming competitive baselines like DAPO. --- ### AHPA: Adaptive Hierarchical Prior Alignment for Diffusion Transformers Authors: Ruibin Min, Yexin Liu, Aimin Pan, Changsheng Lu, Jiafei Wu, Kelu Yao, Xiaogang Xu, Harry Yang Published: 2026-05-05 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.03317v1 Representation alignment has recently emerged as an effective paradigm for accelerating Diffusion Transformer training. Despite their success, existing alignment methods typically impose a fixed supervision target or a fixed alignment granularity throughout the entire denoising trajectory, whether the guidance is provided by external vision encoders, internal self-representations, or VAE-derived features. We argue that such timestep-agnostic alignment is suboptimal because the useful granularity of representation supervision changes systematically with the signal-to-noise ratio. In high-noise regimes, diffusion models benefit more from coarse semantic and layout-level anchoring, whereas in low-noise regimes, the training signal should emphasize spatially detailed and structurally faithful refinement. This non-stationary alignment behavior creates a representational mismatch for static single-level supervisors. To address this issue, we propose Adaptive Hierarchical Prior Alignment (AHPA), a lightweight alignment framework that exploits the hierarchical representations naturally embedded in the frozen VAE encoder. Instead of using only a single compressed latent as the alignment target, AHPA extracts multi-level VAE features that provide complementary priors ranging from local geometry and spatial topology to coarse semantic layout. A timestep-conditioned Dynamic Router adaptively selects and weights these hierarchical priors along the denoising trajectory, thereby synchronizing the alignment granularity with the model's evolving training needs. Extensive experiments show that AHPA improves convergence and generation quality over baselines and incurs no additional inference cost while avoiding external encoder supervision during training. --- ### TACO: Trajectory Aligning Cross-view Optimisation Authors: Tavis Shore, Oscar Mendez, Simon Hadfield Published: 2026-05-05 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2605.03315v1 Cross-View Geo-localisation (CVGL) matches ground imagery against satellite tiles to give absolute position fixes, an alternative to GNSS where signals are occluded, jammed, or spoofed. Recent fine-grained CVGL methods regress sub-tile metric pose, but have only been evaluated as one-shot localisers, never as the primary fix in a live pipeline. Inertial sensing provides high-rate relative motion, but accumulates unbounded drift without an absolute anchor. We propose TACO, a tightly-coupled IMU + fine-grained CVGL pipeline that consumes a single GNSS reading at start-up and thereafter operates on onboard sensing alone. A closed-form cross-track error model triggers CVGL before IMU drift exceeds the matcher's capture radius, and a forward-biased five-point multi-crop search keeps inference cost fixed at five forward passes per fix. A yaw-residual gate rejects fixes that disagree with the onboard compass, and an anisotropic body-frame noise model scales each Unscented Kalman Filter update by per-fix confidence. A factor graph with vetted loop closures provides an offline smoothed trajectory. On the KITTI raw dataset, TACO reduces median Absolute Trajectory Error (ATE) from 97.0m (IMU-only) to 16.3m, a 5.9 times reduction, at <0.1 ms per-frame fusion cost and a 5-10% camera duty cycle. Code is available: github.com/tavisshore/TACO. --- ### When to Think, When to Speak: Learning Disclosure Policies for LLM Reasoning Authors: Jiaqi Wei, Xuehang Guo, Pengfei Yu, Xiang Zhang, Wanli Ouyang, Siqi Sun, Qingyun Wang, Chenyu You Published: 2026-05-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.03314v2 In single-stream autoregressive interfaces, the same tokens both update the model state and constitute an irreversible public commitment. This coupling creates a silence tax: additional deliberation postpones the first task-relevant content, while naive early streaming risks premature commitments that bias subsequent generations. We introduce Side-by-Side (SxS) Interleaved Reasoning, which makes disclosure timing a controllable decision within standard autoregressive generation. SxS interleaves partial disclosures with continued private reasoning in the same context, but releases content only when it is supported by the reasoning so far. To learn such pacing without incentivizing filler, we construct entailment-aligned interleaved trajectories by matching answer prefixes to supporting reasoning prefixes, then train with SFT to acquire the dual-action semantics and RL to recover reasoning performance under the new format. Across two Qwen3 architectures/scales (MoE Qwen3-30B-A3B, dense Qwen3-4B) and both in-domain (AIME25) and out-of-domain (GPQA-Diamond) benchmarks, SxS improves accuracy--content-latency Pareto trade-offs under token-level proxies such as inter-update waiting. --- ### Distributed Learning with Adversarial Gradient Perturbations Authors: Nawapon Sangsiri, Yufei Tao Published: 2026-05-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.03313v1 Privacy concerns in distributed learning often lead clients to return intentionally altered gradient information. We consider the problem of learning convex and $L$-smooth functions under adversarial gradient perturbation, where a client's gradient reply to a server query can deviate arbitrarily from the true gradient subject to a distance bound. Our study focuses on two fundamental questions: (i) what is the smallest achievable sub-optimality gap (i.e., excess error in optimization) under such responses, and (ii) how many queries are sufficient to guarantee a given sub-optimality gap? We establish tight feasibility thresholds on the sub-optimality gap and provide algorithms that achieve these thresholds with provable query complexity guarantees. --- ### Coordination as an Architectural Layer for LLM-Based Multi-Agent Systems Authors: Maksym Nechepurenko, Pavel Shuvalov Published: 2026-05-05 Categories: cs.MA, cs.LG, q-fin.TR arXiv: https://arxiv.org/abs/2605.03310v1 Multi-agent LLM systems fail in production at rates between 41% and 87%, mostly due to coordination defects rather than base-model capability. Existing responses split between cataloguing failure modes empirically and shipping declarative orchestration frameworks as engineering tools; neither delivers a principled mapping from coordination configuration to predictable failure-mode signature. We argue that coordination should be treated as a configurable architectural layer, separable from agent logic and from information access, enabling architectural reasoning rather than only engineering productivity. We instantiate this with an information-controlled design on prediction markets: a single LLM, fixed tools, fixed per-call output cap, and fixed prompt template across five reference coordination configurations, with total compute per question treated as an endogenous architectural output. The Murphy decomposition of the Brier score separates calibration from discriminative power, so configurations leave distinguishable signatures even when aggregate scores coincide. On 100 Polymarket binary markets resolved after the model's training cutoff (claude-opus-4-6) we report Murphy signatures, a cost-quality Pareto frontier, category-conditioned analysis, and a bootstrap power-projection. Three of five pre-specified predictions are upheld in direction; two configurations dominate the Pareto frontier within this regime; exploratory bootstrap intervals separate consensus alignment from others, though pairwise tests do not survive Bonferroni correction at n=100. We also deploy the same configurations as live agents on Foresight Arena under web-search-enabled conditions, as an on-chain replication channel accumulating in parallel. Harness, trace dataset, and production agents are released. We position this as a methodology-validating first instantiation, not a general cross-model claim. --- ### Cryptographic Registry Provenance: Structural Defense Against Dependency Confusion in AI Package Ecosystems Authors: Alan L. McCann Published: 2026-05-05 Categories: cs.CR, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2605.03309v2 Dependency confusion attacks exploit a structural gap in software distribution: once a package is installed, there is no cryptographic proof of which registry distributed it. Every existing defense is configuration-based and fails silently when misconfigured. We present a cryptographic distribution provenance system comprising three components: (1) cryptographic registry identity, where every registry holds an Ed25519 keypair and signs every artifact it distributes; (2) a dual-signature model, where the publisher signs at packaging time and the registry countersigns at publication time; and (3) authoritative namespace binding, where consumers pin registry fingerprints and the resolver cryptographically rejects artifacts from unauthorized registries. These create three defense layers requiring simultaneous compromise for a successful attack. A comparison across eight ecosystems (npm, Cargo, Hex.pm, PyPI, Go modules, Docker/OCI, NuGet, Maven) shows no existing ecosystem combines mandatory publisher signing, cryptographic registry identity, mandatory registry countersigning, and consumer-side cryptographic enforcement. The system extends to AI-generation provenance as a signed attribute and governance-enforced dependency resolution. A case study integrates distribution provenance with a three-layer runtime governance architecture, creating a four-phase lifecycle chain with no cryptographic gaps. --- ### Revisiting the Travel Planning Capabilities of Large Language Models Authors: Bo-Wen Zhang, Jin Ye, Peng-Yu Hua, Jia-Wei Cao, Jie-Jing Shao, Yu-Feng Li, Lan-Zhe Guo Published: 2026-05-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.03308v1 Travel planning serves as a critical task for long-horizon reasoning, exposing significant deficits in LLMs. However, existing benchmarks and evaluations primarily assess final plans in an end-to-end manner, which lacks interpretability and makes it difficult to analyze the root causes of failures. To bridge this gap, we decompose travel planning into five constituent atomic sub-capabilities, including \emph{Constraint Extraction}, \emph{Tool Use}, \emph{Plan Generation}, \emph{Error Identification}, and \emph{Error Correction}. We implement a decoupled evaluation protocol leveraging oracle intermediate contexts to rigorously isolate these components, thereby measuring the atomic performance boundary without the noise of cascading errors. Our results highlight a clear contrast in performance: while LLMs are proficient in extracting explicit constraints, they struggle to infer implicit, open-world requirements. Furthermore, they exhibit structural biases in plan generation and suffer from ineffective self-correction, characterized by excessive sensitivity and erroneous persistence. These findings offer precise directions for improving LLM reasoning and planning abilities. --- ### Will the Carbon Border Adjustment Mechanism Impact European Electricity Prices? A GNN-Based Network Analysis Authors: Jiachen Shen, Jian Shi, Dan Wang, Han Zhu Published: 2026-05-05 Categories: cs.LG, cs.CY, eess.SY arXiv: https://arxiv.org/abs/2605.03304v1 The European Union's Carbon Border Adjustment Mechanism (CBAM) creates a complex challenge for the interconnected European electricity market. Traditional static analyses often miss the cross-border spillover effects that are vital for understanding this policy. This paper addresses this gap by developing a spatio-temporal Graph Neural Network (GNN) framework. It quantifies how CBAM affects electricity prices and carbon intensity (CI) at the same time. We modeled a subgraph of eight European countries. Our results suggest that CBAM is not just a uniform tax. Instead, it acts as a tool that transforms the market and creates structural differences. In our simulated scenarios, we observe that low-carbon countries like France and Switzerland can gain a competitive advantage. This suggests a potential decrease in their domestic electricity prices. Meanwhile, high-carbon countries like Poland face a double burden of rising costs. We identify the primary driver as a fundamental shift in the market's merit order. --- ### Stable Multimodal Graph Unlearning via Feature-Dimension Aware Quantile Selection Authors: Jingjing Zhou, Yongshuai Yang, Qing Qing, Ziqi Xu, Xikun Zhang, Renqiang Luo, Ivan Lee, Feng Xia Published: 2026-05-05 Categories: cs.LG, cs.MM arXiv: https://arxiv.org/abs/2605.03303v1 Graph unlearning remains a critical technique for supporting privacy-preserving and sustainable multimodal graph learning. However, we observe that existing unlearning strategies tend to apply uniform parameter selection and editing across all graph neural network (GNN) layers, which is especially harmful for multimodal graphs where high-dimensional input projections encode dominant cross-modal knowledge. As a result, over-editing these sensitive layers often leads to catastrophic utility degradation after forgetting, undermining both stable learning and effective privacy protection. To address this gap, we propose FDQ, a Feature-Dimension Aware Quantile framework for multimodal graph unlearning. FDQ adaptively identifies high-dimensional input projection layers and applies more conservative, FDQ-guided quantile thresholds when constructing suppression sets, while keeping the underlying importance estimation mechanism unchanged. FDQ is seamlessly integrated with diagonal sensitivity-based parameter importance analysis to enable efficient node and edge unlearning under general forget requests. Through extensive experiments on Ele-Fashion and Goodreads-NC, we demonstrate that FDQ consistently achieves strong utility preservation while maintaining effective forgetting against membership inference attacks. Overall, FDQ offers a principled and robust solution for privacy-aware unlearning in high-dimensional multimodal graph systems. --- ### SHIELD: A Diverse Clinical Note Dataset and Distilled Small Language Models for Enterprise-Scale De-identification Authors: Jose D. Posada, David Love, Somalee Datta, Priya Desai Published: 2026-05-05 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.03301v2 De-identification of clinical text is a prerequisite for the secondary use of electronic health records. Existing public benchmarks such as the i2b2 2006 and 2014 corpora are over a decade old and lack the semantic and demographic diversity of modern clinical narratives. Large Language Models (LLMs) reach state-of-the-art zero-shot extraction, but their use at enterprise scale is limited by computational cost and by hospital data governance that restricts sending Protected Health Information (PHI) to cloud APIs. We introduce SHIELD (Synthetic Human-annotated Identifier-replaced Entries for Learning and De-identification), a diverse clinical note dataset of 1,381 notes with 10,229 gold-standard PHI spans across 9 categories, built with set-cover diversity sampling across demographic and document-type strata and human-in-the-loop adjudication. We evaluate four LLMs (two proprietary, two open-weight) to establish a performance ceiling on SHIELD, then show that a teacher-student distillation framework transfers these capabilities into locally deployable Small Language Models. Our best distilled model reaches micro-averaged span-level precision of 0.89 and recall of 0.88 while running on standard workstation hardware. It trails its cloud teacher on per-category recall (0.90 vs. 0.81 macro-averaged) but remains competitive given its lower cost and on-premise deployability. Cross-dataset evaluation shows that diversity-trained models generalize well on universal structured PHI categories, while institution-specific entities remain hard to transfer in both directions, which suggests pairing broad-coverage models with specialized models for high-volume, semi-structured note types. We publicly release the SHIELD dataset and the distilled DeBERTa v3 model to provide an accurate, cost-effective de-identification pipeline deployable entirely behind institutional firewalls. --- ### LLM-XTM: Enhancing Cross-Lingual Topic Models with Large Language Models Authors: Minh Chu Xuan, Tien-Phat Nguyen, Linh Ngo Van, Dinh Viet Sang, Nguyen Thi Ngoc Diep, Trung Le Published: 2026-05-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.03299v2 Cross-lingual topic modeling aims to discover shared semantic structures across languages, yet existing models depend on sparse bilingual resources and often yield incoherent or weakly aligned topics. Recent LLM-based refinements improve interpretability but are costly, document-level, and prone to hallucination, with prior white-box approaches requiring inaccessible token probabilities. We propose LLM-XTM, a framework that integrates LLM-guided topic refinement with self-consistency uncertainty quantification, enabling black-box, stable, and scalable enhancement of cross-lingual topic models. Experiments on multilingual corpora show that LLM-XTM achieves superior topic coherence and alignment while reducing reliance on bilingual dictionaries and expensive LLM calls. --- ### Contrastive Regularization for Accent-Robust ASR Authors: Van-Phat Thai, Aradhya Dhruv, Duc-Thinh Pham, Sameer Alam Published: 2026-05-05 Categories: cs.SD, cs.LG arXiv: https://arxiv.org/abs/2605.03297v2 ASR systems based on self-supervised acoustic pretraining and CTC fine-tuning achieve strong performance on native speech but remain sensitive to accent variability. We investigate supervised contrastive learning (SupCon) as a lightweight, accent-invariant auxiliary objective for CTC fine-tuning. An utterance-level contrastive loss regularizes encoder representations without architectural modification or explicit accent supervision. Experiments on the L2-ARCTIC benchmark show consistent WER reductions across multiple pretrained encoders, with up to 25 -- 29\% relative reduction under unseen-accent evaluation. Analysis using within-transcript cosine dispersion indicates that SupCon promotes more compact and stable representation geometry under accent variability. Overall, SupCon provides an effective and model-agnostic regularization strategy for improving accent robustness. --- ### FACTOR: Counterfactual Training-Free Test-Time Adaptation for Open-Vocabulary Object Detection Authors: Kaixiang Zhao, Mao Ye, Lihua Zhou, Hu Wang, Luping Ji, Song Tang, Xiatian Zhu Published: 2026-05-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.03294v1 Open-vocabulary object detection often fails under distribution shifts, as it can be misled by spurious correlations between non-causal visual attributes (e.g., brightness, texture) and object categories. Existing test-time adaptation (TTA) methods either depend on costly online optimization or perform global calibration, overlooking the attribute-specific nature of these failures. To address this, we propose FACTOR (counterFACtual training-free Test-time adaptation for Open-vocabulaRy object detection), a lightweight framework grounded in counterfactual reasoning. By perturbing test images along non-causal attributes and comparing region-level predictions between original and counterfactual views, FACTOR quantifies attribute sensitivity, semantic relevance, and prediction variation to selectively suppress attribute-dependent predictions-without parameter updates. Experiments on PASCAL-C, COCO-C, and FoggyCityscapes show that FACTOR consistently outperforms prior TTA methods, demonstrating that explicit counterfactual reasoning effectively improves robustness under distribution shifts. --- ### Imbalanced Classification under Capacity Constraints Authors: Daniel Fraiman, Ricardo Fraiman Published: 2026-05-05 Categories: stat.ML, cs.LG, math.ST arXiv: https://arxiv.org/abs/2605.03289v2 Detecting observations from a minority class under severe class imbalance is a central challenge in applications such as fraud detection, medical screening, and industrial quality control. In these settings, each positive prediction triggers a costly follow-up action, an MRI scan, a transaction audit, whose execution is subject to real operational constraints. This paper proposes a formal classification framework under capacity constraints: given a user-defined bound limit $b$ on the proportion of observations that can be labeled as belonging to the minority class, the goal is to find the classifier that maximizes sensitivity on that class. We characterize the optimal classifier under this constraint and establish its equivalence with the classical Bayes classifier under a reweighting of the prior probabilities. We also introduce a capacity-adjusted performance metric $M$ that accounts for the effective detection rate when the capacity constraint is binding. The framework is implemented on top of standard learning methods, k-NN, SVM, random forests, and neural networks, and statistical consistency is established for each. We further show that these methods reduce to post-hoc thresholding when no hyperparameters are oriented toward the capacity-constrained objective, and introduce a capacity-aware support vector machine that exploits the constraint during training and achieves the strongest empirical performance. Experiments on the Taiwanese credit card default dataset confirm that capacity-constrained classifiers substantially outperform both classical approaches and SMOTE under high imbalance regimes. The framework extends naturally to multiclass settings and online environments. --- ### On the Spectral Structure and Objective Equivalence of Orthogonal Multilabel Fisher Discriminants Authors: Brian Keith-Norambuena, Juan Bekios-Calfa Published: 2026-05-05 Categories: stat.ML, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.03283v2 We provide a unified theoretical analysis of Linear Discriminant Analysis with simultaneous multilabel scatter matrix formulations and Stiefel orthogonality constraints. Our contributions span both algebraic structure and statistical guarantees. On the algebraic side, we characterize the rank of the multilabel between-class scatter matrix, showing that the effective discriminant dimensionality can strictly exceed the classical single-label bound of $C-1$; we establish a multilabel partition of variance and prove that all four Fisher objectives are equivalent under the $W^\top S_t^{ML} W = I_r$ constraint while characterizing their divergence under the Stiefel constraint; and we prove a two-sided label-distance preservation bound relating projected distances to Hamming distances in label space. On the statistical side, we establish a finite-sample $O(k_{\max}\sqrt{d\log d/n}/gap_r)$ bound on the subspace estimation error under sub-Gaussian noise with a matching $Ω(σ^2 d/(n\,gap_r))$ minimax lower bound, establishing a near-minimax-optimal rate (matching up to logarithmic and $k_{\max}$ factors) for multilabel discriminant subspace estimation. We further provide high-probability distance concentration, robustness guarantees under label interactions, and a regularization analysis preserving the spectral structure when $d \gg n$. All results are verified numerically on synthetic data generated from the linear label-effect model, covering both the algebraic identities and the multilabel-specific quantities ($k_{\max}$, $κ(S_t^{ML})$, $\|Γ/n\|_2$, $Δ_r$) that govern the statistical bounds. The numerical experiments are designed as a sanity check for the theorems rather than as an empirical benchmark; evaluation on real multilabel datasets is left to future work targeting application-oriented venues. --- ### Donor-Aware scRNA-seq Benchmarks for IBD Classification Authors: Jonathan Muhire Published: 2026-05-05 Categories: q-bio.QM, cs.LG, stat.ML arXiv: https://arxiv.org/abs/2605.03281v1 Donor-level disease classification from single-cell RNA sequencing (scRNA-seq) requires strict donor-aware cross-validation: naive pipelines that split cells randomly conflate training and test donors, inflating reported performance through pseudoreplication. We present a donor-aware benchmark evaluating three feature representations across two independent IBD cohorts: centered log-ratio (CLR) transformed cell-type composition, GatedStructuralCFN dependency embeddings, and scVI variational autoencoder latent embeddings. The cohorts are the SCP259 ulcerative colitis atlas (UC vs. Healthy, n=30 donors, 51 cell types) and the Kong 2023 Crohn's disease atlas (CD vs. Healthy, n=71 donors, 55-68 cell types across three intestinal regions). Compartment-stratified CLR composition achieves AUROC 0.956 +/- 0.061 on SCP259; GatedStructuralCFN on the same features achieves 0.978 +/- 0.050. In the Kong cohort, CFN achieves its best performance in the colon region (0.960 +/- 0.055 after feature filtering), exceeding linear CLR (0.900 +/- 0.100), while terminal ileum classification is dominated by linear models (CatBoost CLR 0.967 +/- 0.075 vs. CFN 0.811 +/- 0.164). Cross-dataset transfer (CD->UC, four shared cell types) achieves AUC 0.833 with XGBoost CLR; the reverse direction performs at chance. CFN edge stability analysis shows that compartment-wise composition eliminates spurious unit-sum-induced instability present in global composition (Jaccard 0.026 vs. top-20 recurrence 1.0). CFN shows a consistent numerical advantage over linear models in the colon region of CD (AUROC 0.960 vs. 0.900), though no inter-method comparison reached statistical significance at n<=34 donors per region. Compartment-aware feature construction is critical for both classification performance and structural interpretability. Code: https://github.com/Jonathan-321/sfn-scrna-study --- ### RFPrompt: Prompt-Based Expert Adaptation of the Large Wireless Model for Modulation Classification Authors: Md Raihan Uddin, Tolunay Seyfi, Fatemeh Afghah Published: 2026-05-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.03279v1 Automatic modulation classification (AMC) in real-world deployments demands robustness to distribution shifts arising from hardware impairments, unseen propagation environments, and recording conditions never encountered during training. Although wireless foundation models offer a promising starting point for robust RF representation learning, an important open question is how to adapt them efficiently to out-of-distribution (OOD) downstream tasks without overwriting the structure learned during large-scale pre-training. In this paper, we investigate prompt-based adaptation as a general mechanism for OOD transfer in wireless foundation models. We propose RFPrompt, a parameter-efficient framework that introduces learnable deep prompt tokens while keeping the pretrained backbone frozen, enabling task-specific adaptation with minimal trainable parameters. We instantiate and evaluate this approach on the Large Wireless Model (LWM), a mixture-of-experts wireless foundation model, and study its behavior under both standard and OOD modulation-classification settings. Results show that prompt-based adaptation consistently improves robustness under distribution shift and limited supervision, particularly on real-world over-the-air IQ data, while preserving strong parameter efficiency. These findings suggest that prompt learning is a practical and effective strategy for adapting wireless foundation models to challenging downstream RF environments. --- ### Copula-Based Endogeneity Correction for Doubly Robust Estimation of Treatment Effect Authors: Sahil Shikalgar, Md. Noor-E-Alam Published: 2026-05-05 Categories: stat.ME, cs.AI arXiv: https://arxiv.org/abs/2605.03278v2 Doubly Robust (DR) estimation of treatment effect relies on an untestable assumption that is the absence of unobserved confounding. This assumption is par- ticularly problematic in the context of healthcare research, where variables like pre- scription refill rates serve as proxies for unobserved behaviors such as medication adherence. These proxy variables are often endogenous, exhibiting correlation with the regression error term due to unmeasured confounding or measurement error. We propose a copula-corrected doubly robust estimator that addresses endogeneity in both the treatment and outcome models without requiring instrumental variables. Gaussian copulas model the joint distribution of endogenous covariates and the error term, enabling consistent estimation while preserving the doubly robust property that requires correct specification of either the treatment or outcome model, not both. Monte Carlo simulations demonstrate that naive DR estimation exhibits substantial bias under endogeneity, whereas our corrected estimator recovers unbiased treatment effects across different data-generating processes. We apply our method to examine the effect of nutritional counseling on blood pressure using the National Health and Nutrition Examination Survey (NHANES) data. Naive DR estimation suggests counseling is associated with increased blood pressure. After copula correction, this effect becomes statistically insignificant, consistent with literature showing modest effects of nutri- Counseling in reducing blood pressure. Our methodology provides researchers with a practical tool for obtaining treatment effects in the presence of endogeneity. --- ### VEBench:Benchmarking Large Multimodal Models for Real-World Video Editing Authors: Andong Deng, Dawei Du, Zhenfang Chen, Wen Zhong, Fan Chen, Guang Chen, Chia-Wen Kuo, Longyin Wen, Chen Chen, Sijie Zhu Published: 2026-05-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.03276v2 Real-world video editing demands not only expert knowledge of cinematic techniques but also multimodal reasoning to select, align, and combine footage into coherent narratives. While recent Large Multimodal Models (LMMs) have shown remarkable progress in general video understanding, their abilities in multi-video reasoning and operational editing workflows remain largely unexplored. We introduce VEBENCH, the first comprehensive benchmark designed to evaluate both editing knowledge understanding and operational reasoning in realistic video editing scenarios. VEBENCH contains 3.9K high-quality edited videos (over 257 hours) and 3,080 human-verified QA pairs, built through a three-round human-AI collaborative annotation pipeline that ensures precise temporal labeling and semantic consistency. It features two complementary QA tasks: 1) Video Editing Technique Recognition, assessing models' ability to identify 7 editing techniques using multimodal cues; and 2) Video Editing Operation Simulation, modeling real-world editing workflows by requiring the selection and temporal localization of relevant clips from multiple candidates. Extensive experiments across proprietary (e.g., Gemini-2.5-Pro) and open-source LMMs reveal a large gap between current model performance and human-level editing cognition. These results highlight the urgent need for bridging video understanding with creative operational reasoning. We envision VEBENCH as a foundation for advancing intelligent video editing systems and driving future research on complex reasoning. ---