CS Paper Radar

Daily listing

2026-11-17 · cs.CV

178 papers first submitted this day (UTC) in cs.CV — Computer Vision and Pattern Recognition. arXiv announces a day's submissions the next evening, so the newest day fills up over ~24 h.

Score = weighted mean ofCitations 20%Upvotes 10%Code stars 10%Venue 15%Code/data 10%Novelty 15%Momentum 15%Substance 5%how it works

AI & Machine Learning 14

MillBill: A Machining Agent with Geometric Feedback

cs.AI cs.CV ·2026-11-17 ·Gennadii Savrasov, Artem Slavin, Soslan Kabisov, Andrey Kuznetsov, Anton Konushin +1 more
code/data
Abstract · 18 pages, 8 figures. Code: https://github.com/kulibinai/millbill

Turning a CAD model into a machining program requires both process knowledge and geometric precision. Language models can generate such workflows by writing geometry and toolpath code for each part, but this makes numerical execution part of the model's task. We introduce MillBill, an agent that separates machining decisions from their geometric execution and verification. Given a boundary-representation part, stock, and material, a locally hosted, mid-sized open model proposes operations and cutting parameters for indexed three-axis machining. A geometric harness generates toolpaths, simulates material removal, and returns feedback for plan refinement. Setup planning accounts for what the available cutters can physically reach, rather than visibility alone. We evaluate the resulting programs by the fraction of the target surface area within the specified tolerance after simulated machining. With Qwen3.8-27B, MillBill outperforms frontier-model approaches that construct their own geometry and toolpath workflows on the evaluated machining tasks. Code: https://github.com/kulibinai/millbill

0.63

FrameRouter: Frame Budget Routing for Long-Video Understanding on Video-MME-v2 and Beyond

cs.AI cs.CVcs.LG ·2026-11-17 ·Bingjun Luo, Jialin Guo, Siqi Li
top venuecode/data
Abstract · Accepted to NeurIPS 2026

Long-video question answering benchmarks like Video-MME-v2 typically evaluate systems under a uniform frame budget for every question, despite stark variation in the evidence each query actually demands. We analyze this mismatch on Video-MME-v2 and find that different question subsets prefer different frozen frame-budget policies, while uniformly increasing the frame count does not reliably improve all metrics. These findings motivate fixed-mean frame-budget routing, a controlled inference setting in which each question may receive a different frame budget while the average budget matches a uniform reference. We introduce FrameRouter, a training-free test-time framework that combines an evidence-demand router, a budget-constrained allocation rule, and a frozen bank of frame-sampling policies. By controlling how much visual evidence each question receives while leaving the underlying samplers unchanged, FrameRouter is orthogonal to query-aware frame selection. Across Video-MME-v2 and several long-video understanding benchmarks, FrameRouter improves long-video question answering by using the same average visual budget more selectively across questions. Code is available at https://github.com/bingjunluo/FrameRouter.

0.51

CADMAGEDDON: High-Fidelity CAD Reconstruction with Local Inference

cs.AI cs.CV ·2026-11-17 ·Oleg Pavlov, Nikita Gavrilov, Soslan Kabisov, Gennadii Savrasov, Andrey Kuznetsov +2 more
code/data
Abstract · 13 pages, 4 figures, 4 tables. Code: https://github.com/kulibinai/cadmageddon

CAD practitioners often work with proprietary parts that they are unwilling to share with external services. For individuals and small companies, bringing this work in-house also requires computational resources they can afford. We introduce CADMAGEDDON, an open-source agent that reconstructs editable CAD from meshes and can run locally on single-GPU systems and current-generation laptops. The method decomposes this complex task into a chain of simple visual decisions about what to construct and what to correct. Geometric tools support precise construction, allowing the agent to focus on visual reasoning. Using the mid-sized Qwen3.8-27B model, CADMAGEDDON outperforms prior approaches, including systems based on frontier models, in reconstruction quality on all the main benchmarks. CADMAGEDDON integrates with FreeCAD, bringing its reconstructions into practitioners' existing CAD workflows. Code: https://github.com/kulibinai/cadmageddon

0.45

The Illusion of Grounding: Do Vision-Language Models Weigh Retrieved Text and Images Like an Ideal Observer?

cs.LG cs.CLcs.CV ·2026-11-17 ·Basavaraj R M, Harshal Chandel
code/data
Abstract · 10 pages, 5 figures, 11 tables. Code and data: https://github.com/Basavarajearlybird/GIB-200K (archived at https://doi.org/10.5281/zenodo.24…

Retrieval-augmented vision-language models (VLMs) answer from two sources of evidence that can each fail: the retrieved text can be wrong or noisy, and the image can be degraded or contradicted. We introduce GIB-200K, 200,000 image-question items built from COCO and Visual Genome, each expanded into controlled stress conditions: hard-negative retrieval noise (1-20 documents), visual ambiguity (Gaussian and motion blur, low-resource crops) and cross-modal conflict with counterfactual flip pairs. Instead of only measuring degradation, we ask how models combine the two modalities. From ideal Bayesian cue combination we derive a testable prediction that yields formal definitions of Knowledge Inertia (text over-weighted relative to the model's own image evidence) and Sensory Capture (the reverse), and we estimate the weights with a measurement-error-corrected regression whose recovery of known weights we verify. A counterfactual flip decomposition separates modality bias from concept priors, and a Grounding Illusion Score measures how much clean accuracy is not grounded in the image. We evaluate 4 open models on one frozen, seeded evaluation split of 2,000 GIB-200K items. All models are accurate on clean inputs (92.5%-94.2%) but adverse accuracy falls to 41.3%-59.9%. The two LLaVA models over-weight the text (Knowledge Inertia, w_T/w_I = 1.229 and 1.322), whereas InternVL3 favours the image (w_T/w_I = 0.680); between 34.8% and 71.4% of the clean answers that are correct are abandoned when the document lies. A training-free consistency check that compares the answer with and without retrieval trades coverage for accuracy under misleading documents, while falling back to the image fails when the image is heavily degraded. All per-instance outputs, code, and an audit that recomputes every reported open-model number are released.

0.39

Distortion-Free Semantic Watermarks for Black-Box Generative Models

cs.LG cs.CRcs.CV ·2026-11-17 ·Aviv Yossef, Or Patashnik, Or Zamir
code/data
Abstract · Project page: https://avivyossef.github.io/distortion-free-semantic-watermark/ Code: https://github.com/avivyossef/distortion-free-semantic-…

We present a provably distortion-free semantic watermarking algorithm that requires only black-box sampling access to a generative model and black-box access to a semantic embedding. A watermark is called distortion-free if, averaged over its secret detection key, it leaves the generator's output distribution unchanged; this provides a formal guarantee that watermarking does not reduce output quality. Conventional such watermarks are often brittle to transformations. This is especially problematic for images, where benign operations such as compression, cropping, or resizing change the entire pixel representation. Semantic watermarks therefore require detection to be robust to content-preserving transformations. Previous methods combining these properties rely heavily on specific white-box properties of the underlying model: either to autoregressive text generation or to diffusion-specific access, such as control of the initial noise or diffusion inversion. In contrast, our construction is fully generic: It only interacts with the generative model and the semantic embedding, a function that determines the notion of "preserving content" under which detection should be robust, via sample and query access. We then instantiate this framework for image generation. We characterize the two competing properties required of the semantic embedding: it should remain stable under transformations that preserve an image's content, but should distinguish independent generations from the same prompt. We train an image encoder directly for this tradeoff. The resulting encoder improves separation while maintaining stability over off-the-shelf alternatives and enables strong aggregate detection while remaining robust to various transformations. We demonstrate the effectiveness of our watermark scheme on both open-source and closed models, which is possible only due to restriction to black-box access.

0.37

Not All Evidence Is Admissible: Evidence Governance for Tool-Augmented Medical Agents

cs.AI cs.CV ·2026-11-17 ·Runxun Zhang, Onofrio Antonio Catalano, Jinqiao Wang, Jingwei Wei
Abstract · 11 pages

Tool-augmented medical agents treat whatever their tools return as evidence. We show that this default is costly: medical tools fail silently, and a frozen reasoner cannot ignore what enters its context, so many cases it would have answered correctly are flipped to wrong once tool outputs are added. We call this evidence hijacking. $\mathcal{V}$-information explains the phenomenon: the reasoner has lost the option to ignore, so a tool output can carry negative usable information, and moving the decision of what to keep outside the reasoner restores that option. The theory further shows that part of any filtering gain comes merely from letting the reasoner see less, and only selection by content lets it see better. Building on this, we recast tool-augmented reasoning as evidence governance, which separates deciding what counts as evidence from the reasoner that answers, and propose context-independent evidence governance (CIEG), the first medical agent framework that rules, after acquisition, on whether each individual tool output may enter as evidence. Every tool output is written as a card that can be checked against the raw input; an independent checker that sees only the input and the cards decides which cards become evidence; and the reasoner answers from the admitted cards alone, with no training anywhere. On medical question answering, fundus glaucoma screening, and chest X-ray pneumonia classification, it achieves the highest accuracy among models of comparable scale, and significantly outperforms state-of-the-art tool-using agents and multi-agent systems built on the same backbone at lower cost.

0.36

Perception into Imagination: Executable Memory for World Models

cs.LG cs.AIcs.CVcs.RO ·2026-11-17 ·Xianyao Li, Ruitong Tian, Rui Min, Fang Xu, Jing Du
hot direction
Abstract · 18 pages, 6 figures, 15 tables

Physical agents must use what they previously saw or touched after visibility or contact is lost. Their actions change the implications of these remembered facts. We introduce executable perceptual memory (EPM), a world-model framework that advances explicit perceptual state alongside neural latent dynamics during imagination. Its evolving read-out informs reward prediction, value estimation and policy learning. We align reward inputs with the transitions that produced them and supervise the reward head on open-loop rollouts of recorded actions. On Memory Maze $13\times13$, EPM reaches 30.37 return at 1M interactions, compared with 23.37 when the same evolving read-out informs only policy and value on a common perceptual front end: a gain of 7.00 [3.80, 10.21]. The complete method also improves control on $9\times9$ and in MiniWorld. Probes of remembered, out-of-view arrivals show stronger reward anticipation in both visual implementations. Concentration, tactile recovery and real tactile frames demonstrate complementary roles of remembered facts in imagined-policy training, motion-dependent decisions and sensory-content retention. EPM connects persistent perception to the predicted consequences and learning signals of a world model for Physical AI.

0.33

What Does Weight-Space Programming Add to Diffusion Editing? Program Construction versus Score-Guided Refinement

cs.LG cs.CV ·2026-11-17 ·Zitian Gao, Jingyi Xu, Ziqi Han, Chengkai Lei, Ye Liu +1 more
hot direction
Abstract · 25 pages, 7 figures, 13 tables

Weight-space programming (WSP) lets a language model construct executable weight updates for another network. We study diffusion concept editing through an interface in which programs define training contexts, preservation constraints and update directions, while a shared numerical fitter selects their coefficients. We compare generated programs with published editors, a shared-panel hand-written baseline and budget-matched search over conventional edits on 24 TIMED test edits for Stable Diffusion 1.4. The preregistered rewrite pipeline shows no advantage over the strong baseline. In an exploratory comparison, one-shot Astra programs improve fresh-scene VLM joint success, which requires subject presence together with whole-image target matching, by +10.0 points over the shared-panel baseline, while CLIP relative match shows no detectable change. These programs organise adjustable directions around protected text-state subspaces. Probing and rewriting the Astra programs then raises CLIP by +6.2 points while joint success changes by -9.7. On a fixed 144-image sample, five additional evaluators give point estimates in the same directions under whole-image target matching; checks of subject-local change give less decisive differences. Together, these results identify useful behaviour in complete generated programs and show that score-guided revision need not preserve it: program construction and revision should be evaluated as distinct stages of a WSP pipeline.

0.29

Seeing What Still Holds: How Waiting Control Changes Cooperative Recovery

cs.AI cs.CVcs.RO ·2026-11-17 ·Kangjie Huang, Yucheng Xin, Xinchun Wang, Dong Liu, Yan Fan +1 more
Abstract · 8 pages, 3figures

Changing the holding reference during a fixed-duration wait can reverse the ranking of visual recovery methods. We introduce AuditEmbodied, a paired execution and trajectory evaluation framework that independently varies holding references and waiting duration while keeping recovery executors frozen. A 960-execution factorial on 20 initialization seeds compares CAD-feedback and landmark-based recovery from identical post-wait physical states within each holding and waiting condition. Protected completion requires successful recovery under the declared execution constraints. At a 1 s wait with registered wrist references, their protected completion counts change from 6/40 versus 19/40 under registered holding to 34/40 versus 22/40 under position-only holding. Trajectory analysis localizes the main separation to grasp formation: small observation-time target residuals precede large tracking errors. Joint diagnostics and controlled interventions characterize persistent tracking constraints. On fresh seeds, gradual position reanchoring improves CAD recovery's long-wait completion by 42.5 percentage points and reduces mean complete recovery time by 6.06 s relative to registered holding. Additional evaluations with pretrained policies, handover, and a mass-matched geometry characterize holding-reference sensitivity across executors and physical regimes. These findings establish holding reference as an explicit evaluation condition for comparing visual recovery performance.

0.29

When Lower Harm Misleads: Auditing Concept Corrections

cs.AI cs.CV ·2026-11-17 ·Tianyou Lai, Wentao Yue, Hongji Li, Chaoyuan Hao, Qilei Li
Abstract

An expert who corrects a concept prediction, such as a BI-RADS descriptor in breast ultrasound, expects the diagnosis to improve. Objectives that reduce the harm of such corrections are judged by harm and accuracy, yet harm also falls when a model responds less to every correction or starts from a worse uncorrected prediction, the reference for harm. We propose a correction audit, a reproducible evaluation protocol that separates these explanations: it splits each correction into harm and useful gain on identical events, separates the starting prediction from the correction response, and compares objectives with amplitude baselines matched in source harm. On three breast-ultrasound sources, worst-case correction training, evaluated from the mean-case model's start, lowers harm but gives up more useful gain in all 21 source comparisons pooled over six seeds, raising the corrected negative log-likelihood (NLL); from the same start, selecting between the objectives by harm picks the less helpful model on all three sources. On CUB, a monotonic harm penalty at its reported weight lowers path-level harm in all 80 concept sets while raising the NLL in all 80, almost entirely before any correction. At a fixed start, rescaling both responses to the same source harm reproduces the penalty's mean harm, useful gain and NLL on single-concept corrections within a prespecified margin of 0.0005 nat, although the two responses point in different directions. Lower harm therefore signals better corrections only when it is read together with useful gain, absolute risk and the start it is measured from.

0.28

RiDEx: Reward-Informed Directional Exploration for Diffusion Reinforcement Learning

cs.LG cs.CV ·2026-11-17 ·Song Yan, Chenfeng Wang, Wei Zhai, Xingchen Wang, Tao Zhang +4 more
Abstract

Reward-based diffusion post-training updates model parameters from task rewards. While existing approaches improve sampling and policy optimization, they still rely largely on isotropic random perturbations in a high-dimensional latent space for exploration. Such variation often spends the perturbation budget on changes with little influence on the target reward, making exploration inefficient. We study where diffusion models should explore and how to learn from that exploration. We introduce Reward-Informed Directional Exploration (RiDEx), which concentrates local exploration in reward-relevant subspaces of the denoising velocity field. Conditional residuals or differentiable reward gradients propose the subspace, and equal-and-opposite perturbations at the same generation state measure its terminal reward response. The reward difference supplies signed local supervision in the explored coordinates, while all generated images contribute to global DiffusionNFT learning. RiDEx thereby couples exploration-space selection with feedback about the changes it produces. On Stable Diffusion 3.5 Medium, RiDEx attains the highest peak rewards among the evaluated methods on four alignment and preference tasks. Its gradient-direction instantiation achieves 15x and 29x target-attainment speedups over DiffusionNFT on HPSv2 and PickScore, measured in H200 GPU hours. Controlled experiments confirm that reward-aligned directions elicit stronger responses at equal perturbation magnitude, and that applying their measured credit improves curve-average reward across all four tasks.

0.23

Scaling Laws for Vision-Language Fusion

cs.LG cs.CV ·2026-11-17 ·Seng Pei Liew, Teppei Suzuki, Seiya Tokui
Abstract

Vision-language models (VLMs) have emerged as central foundation models combining multimodal perception with text comprehension. While VLMs are traditionally built via late fusion by continued pre-training an off-the-shelf language model on multimodal (image-text) data, mixing modalities from the start of pre-training (early fusion) has recently gained traction. However, the fundamental trade-offs between these paradigms remain poorly understood, resulting in expensive and heuristic-driven training recipes. In this paper, we systematically analyze the learning dynamics of these multimodal pre-training processes, showing that validation losses during vision-language fusion can be modeled by separating the base learning trajectory, where different data types contribute at distinct rates, from a distribution shift driven by the competition between modalities. We derive a scaling law that unifiedly describes these dynamics, enabling accurate loss predictions across mixture ratios, training steps, and model sizes. Crucially, our law characterizes the trade-off between early and late fusion: under a fixed data budget, early fusion preserves text performance while late fusion favors vision performance, yielding a trade-off that increasingly tilts toward early fusion as model size scales. Extensive experiments confirm that our scaling law reliably captures multimodal learning dynamics and governs fusion decisions.

0.18

PACT: Plan-Action Consistency Training by Preserving Correct Plans and Repairing Actions in GUI Agents

cs.AI cs.CVcs.LG ·2026-11-17 ·Zhuohan Cai, Zhangxuan Gu, Runze Li, Beitong Zhou, Ming Wen +3 more
benchmark
Abstract · 21 pages, 11 figures, including supplementary material

GUI agents can formulate a correct immediate plan in their reasoning yet emit an action that fails to execute it, creating an execution gap. This recurring mismatch undermines reliable task completion, as a single execution error can derail a multi-step workflow. In outcome-supervised reinforcement learning, broadcasting a negative response-level advantage across all generated tokens risks penalizing valid planning together with erroneous execution. To address this credit-assignment conflict, we introduce PACT (Plan-Action Consistency Training), a training framework that combines reference-aligned plan protection and plan-aligned action repair without altering the original scalar action reward. Using annotated reference actions to identify execution gaps, PACT protects the verified plan by masking negative policy-gradient contributions across its token span. Keeping the generated reasoning unchanged as context, it then favors the reference action over the emitted erroneous action through a preference surrogate with stop-gradient pair weights. Compared with the outcome-only baseline, PACT raises AndroidWorld task success from 60.34% to 68.97% (+8.63 pp), while improving rule-based plan-action consistency on AndroidControl High from 83.71% to 89.74% (+6.03 pp). Adapting PACT to desktop agents also improves task success from 40.92% to 44.17% (+3.25 pp) on the full 369-task OSWorld benchmark. Together, these results establish PACT as an effective credit-assignment framework for bridging execution gaps and improving task completion across mobile and desktop GUI environments.

0.14

LoCaR: Localized Canvas Retrieval With Frozen Vision-Language Model

cs.AI cs.CVcs.LG ·2026-11-17 ·Rahim Hossain, Dipanwita Saha, Ashirbad Mishra, Xinxin Shu, Svetlana Strunjas +1 more
benchmark
Abstract · Zero-shot retrieval , Local Retrieval

Whole image embeddings can under represent small objects and local attributes that distinguish relevant images for localized queries. We present LoCaR, a training free framework for localized text to image retrieval with frozen vision language models. LoCaR constructs and utilizes independent proposals (SAM2) focusing on the object so as to get a more localized representation that independently encodes each region using the frozen visual encoder. We further study paired-support adaptation as an optional residual that uses matched text region exemplars without updating the underlying encoders. We also introduce LOCFG, a Visual Genome derived benchmark designed to evaluate localized retrieval for smaller and more challenging referents. Across several vision language backbones, our results show that improving the region representation itself is the main source of gain.

0.06

Language, Vision & Media 154

ForceBody: Force-Paired Parametric Body Motion with Torque Uncertainty

cs.CV cs.AI ·2026-11-17 ·Joonwoo Kwon, Yufei Zhang, Xiaoming Liu, Zijun Cui
top venuebenchmarkcode/data
Abstract · Accepted at NeurIPS 2026 Evaluations & Datasets Track

Forces and torques drive human motion, central to biomechanics, robotics, and physics-aware animation, yet they are often absent from learning-based motion pipelines for body reconstruction and generation. Some motion datasets provide force annotations, but the paired motion sequences are incompatible with the parametric body models used in modern learning pipelines. In addition, existing datasets typically obtain joint torques through inverse dynamics, which is inherently sensitive to the quality of the input kinematics; however, they do not quantify the reliability of these torque estimates. We close this gap with ForceBody, the first dataset to bring ground truth force supervision into a parametric body representation, shipped with per-sample torque uncertainty. ForceBody pairs the SKEL body model with ground reaction forces and inverse-dynamics joint torques across 10,386 motion trials (9.7M frames, 26.9 hours) from 140 subjects. Each torque label is paired with a per-frame, per-joint uncertainty annotation obtained by Monte Carlo sampling through the inverse dynamics pipeline. We benchmark six architectures (MLP, Conv1D, LSTM, GRU, Mamba2, Transformer) on ground reaction force and joint torque prediction from SKEL motion. As one example use of the released uncertainty, finetuning a Transformer with per-sample uncertainty weighting reduces the MAE of joint torque estimation by 5%. Code and data are at https://github.com/DominoAI-Lab/ForceBody-NeurIPS-ED-2026.

0.66

EfficientVPR: Toward Efficient Visual Place Recognition via Scene-Aware Prompt Tuning and Adaptive Feature Enhancement

cs.CV ·2026-11-17 ·Wenjing Tang, Chuanguang Yang, Zhulin An, Libo Huang, Yongjun Xu
top venuecode/data
Abstract · 18 pages, 11 figures, accepted at CVPR 2026

Visual place recognition (VPR) faces critical challenges in handling extreme environmental variations while meeting the computational constraints of practical applications. Current methods predominantly address these challenges by either scaling up model capacity or employing computationally intensive reranking stages, creating a significant efficiency bottleneck. To overcome this limitation, we propose EfficientVPR, a lightweight one-stage framework that achieves speed-accuracy trade-offs through two key innovations: i) a scene-aware visual prompt tuning method which adapts pretrained features with less parameters while dynamically adjusting to sample-specific characteristics, and ii) an instance-dependent key local feature enhancement module that further reinforces discriminative regions. Comprehensive evaluations on Pitts250k, MSLS, Eynsham, AmsterTime and SVOX demonstrate that our method establishes a new SOTA for DINOv2-small models by outperforming all same-scale competitors, and delivers a $73\times$ speedup with 60\% lower-dimensional features while maintaining competitive (within 2.5% average R@1 gap) against the SOTA DINOv2-large-based two-stage method. Our code will be available at https://github.com/WiniTang/EfficientVPR.

0.65

SWIR-cGAN: Synthesizing SWIR Bands Using VNIR Imagery from CBERS and AMAZONIA-1

cs.CV ·2026-11-17 ·Alisson Cleiton de Oliveira, Rennan de Freitas Bezerra Marujo, Thales Sehn Körting
publishedcode/data
Abstract · Available on: https://ieeexplore.ieee.org/document/11793642. Early Access

Spectral band synthesis is an important research topic in remote sensing. The Brazilian satellites CBERS-4, CBERS-4A, and AMAZONIA-1 provide high temporal resolution (1-3 days) visible and near-infrared (VNIR) imagery through the Wide-Field Imager (WFI), but lack short-wave infrared (SWIR) bands. Recent deep learning methods have shown promising results for spectral synthesis, but challenges remain for cross-satellite generalization and spectrally distant bands. This paper presents a conditional Generative Adversarial Network (cGAN) with an attention-enhanced U-net generator and a multi-loss objective to synthesize Sentinel-2/MSI SWIR bands 11 (SWIR-1) and 12 (SWIR-2) from WFI VNIR inputs. Experiments used 4,125 patches for training and cross-validation and 1,026 unseen patches for testing, with hyperparameters optimized using a subset of the training data. On SWIR-1, the model achieved SSIM of 0.900, PSNR of 31.44 dB, and MAE of 0.019 in TOA reflectance. On SWIR-2, it achieved SSIM of 0.917, PSNR of 33.35 dB, and MAE of 0.015. Benchmarking was performed on the test dataset, with each model optimized separately prior to evaluation. SWIR-cGAN outperformed the baselines, particularly for SWIR-2. The ablation study identified the multi-loss as the main performance contributor, while the attention and residual modules provided additional improvements. Our results demonstrate the feasibility of synthesizing SWIR bands from WFI imagery. The SWIR-cGAN is available at https://github.com/marujore/swir-cgan.

0.63

Retrieval-Augmented Correction for Generalist Medical Image Segmentation

cs.CV cs.AI ·2026-11-17 ·Vincenzo Marcianò, Xiaoming Zhang, Michela Antonelli, Sebastien Ourselin, Maria A Zuluaga
top venuecode/data
Abstract · Accepted in Transactions on Machine Learning Research (10/2026). Reviewed on OpenReview: https://openreview.net/forum?id=PeDqE0SQRe

Automated segmentation is routinely used as the first step in quantitative image-based clinical workflows, where predicted masks are subsequently reviewed and corrected by human operators. Automated correction of the residual structural errors is currently addressed one anatomy at a time, so the number of trained correctors scales with the number of segmentation targets. We present GRACE, a retrieval-augmented refiner that edits an input mask under context drawn from a shared knowledge base of verified prior cases. The refiner is trained once on a multi-organ corpus spanning compact organs, cavities, branching vessels, and tumour lesions. In the held-out dataset-transfer evaluation, three held-out datasets query a fixed knowledge base built from training data only. A separate leave-one-anatomy-out experiment tests parameter-free index extension: verified cases of an unseen solid organ enter the knowledge base, and the refiner stays frozen. A light shape-aware projection head on a frozen multimodal encoder maps embedding similarity to the shape agreement of the masks, so the knowledge base returns cases that the refiner can use as a shape prior. We evaluate GRACE on 18 public datasets spanning CT, MRI and ultrasound, using nine refinement configurations and seven upstream segmentors that span weak, mid, and strong operating regimes. We report each change with the DSC of its input mask. On the weak tier, the input DSC is 0.386 and GRACE adds up to +0.35 DSC. On a strong segmentor, the input DSC is 0.932 and the mean change is small. Transfer to an imaging modality that is absent from training and from the index is partial and depends on the anatomy, and remains an open research direction. Our results suggest retrieval-conditioned refinement is a practical alternative to anatomy-specific correction models. Code: https://github.com/robustml-eurecom/grace

0.54

When Neural Collapse Fails to Route: Routing Drift in Continual Learning

cs.CV ·2026-11-17 ·Nicolas Larue
publishedcode/data
Abstract · Accepted at BMVC 2026. 36 pages, including supplementary material

Progressive neural-collapse classifiers can learn strong task-local decision boundaries while failing the global Class-IL decision. We identify the failure as routing drift: under an expanding equiangular tight frame (ETF) classifier, a held-out sample can remain separable within its true task block, yet receive its largest global score from another block. This produces high Task-IL accuracy but low Class-IL accuracy, and small replay buffers can mask the problem because stored samples are optimized repeatedly and often remain well routed. We formalize route accuracy directly from the global class argmax and use it to separate within-task discrimination from cross-task score comparability. This decomposition suggests a classifier-level correction: preserve the global ETF frame, but add class-local reliability evidence when transported class measures deviate from the neural-collapse ideal. We develop Reliability-Calibrated Routing (RCR), a lightweight post-hoc classifier that composes standardized ETF alignment with radial plausibility around transported, NC-shrunk class representatives. Across three datasets and two fixed replay-buffer budgets, this NC-consistent correction gives the best Class-IL FAA on five of six settings and reduces Class-IL forgetting on all six. The largest gains occur in harder routing regimes: up to $+6.75$ FAA over the reported reference, and up to $+8.03$ FAA and $+8.62$ final route accuracy in paired evaluations, while preserving Task-IL performance. These results position routing drift as a measurable failure mode of ETF-based continual learning and show that reliability-calibrated class evidence can repair part of the lost global routing without retraining the representation. Code and evaluation artifacts are available at https://towzeur.github.io/RCR/.

0.50

Know When to Zoom: How Small Is Too Small for Vision-Language Models?

cs.CV cs.AI ·2026-11-17 ·Siddharth Vohra, Min Xu
published
Abstract · Accepted to the SENSE workshop at WACV 2027. 10 pages, 5 figures, 3 tables. Code, data and a plain-language walkthrough: https://siddvoh.com…

Online vision-language models often miss small details in large images, because the service shrinks the image, or rejects it, before the model ever sees it. A car 20 pixels wide in an 8000 by 6000 drone photo reaches Claude about 6 pixels wide, and OpenAI's default rejects the photo outright. We measure how much detail survives for seven models from four vendors, using an eye exam of printed codes and 601 questions about small objects in real photos. The smallest text a model reads half the time grows from 6 pixels or less on a small image to a predicted 10 to 13 pixels on a 7680 by 4320 photo at default settings. Zooming wins much of the detail back. Splitting each photo into up to nine tiles raises accuracy by 11.1 to 14.5 percentage points for Claude and Gemini, and by avoiding rejections it lifts GPT-6.1 Sol from 50.7% to 83.7%. A two-step zoom, in which the model first names the region it needs, matches or beats tiles on Gemini and comes within 2.5 points of them on Claude, with about 40% of their image tokens. When a detail is lost, Claude and Gemini both say they cannot tell about half the time, but they fail differently in the rest: Gemini gives a confident wrong answer in 43% to 48% of these cases, Claude in only 5% to 9%. A check based on detail size, run before the call, misses most of the gain from tiles on Claude and Gemini, but Claude's own reply is a reliable trigger: tiling only after an unsure reply comes within 1.4 points of always tiling with 62% to 70% of the image tokens. We turn these findings into five rules for when an agent should zoom, and release the eye exam, every model answer and code at https://siddvoh.com/know-when-to-zoom.

0.49

Do Diffusion Models Learn to Generalize Basic Visual Skills?

cs.CV cs.LG ·2026-11-17 ·Amish Sethi, Boya Zeng, Wenhao Chai, Zhuang Liu
publishedcode/data
Abstract · NeurIPS 2026. Project page: https://amishsethi.github.io/visual-generative-lab/

Diffusion models generate high-quality images, but how well they generalize is hard to measure. Current evaluations rely on human preference and prompt-following benchmarks for text-to-image models and on distribution-level metrics such as Fréchet Inception Distance (FID) and Inception Score (IS) for class-conditional models, but these cannot tell whether models learn basic visual skills or largely match training patterns. To address this gap, we introduce the Visual Generative Lab (VGL), a controlled experimental framework for studying how diffusion models learn and generalize basic visual skills, including size, position, rotation, count, color, shape, and their compositions. By generating synthetic training data for each skill from explicit rules, VGL allows precise measurement of generalization with task-specific, rule-based metrics. Training models from scratch on these data reveals three patterns in this controlled setting: (1) with our default design, extrapolation is limited: size and position degrade within a few pixels of the training range, count reaches only about 50% accuracy on out-of-range counts, and rotation extrapolation accuracy varies widely across seeds; (2) for compositional generalization, the number of distinct basic visual skill combinations matters more than dataset size; (3) skills mostly fail independently, except position. Code and data-generation scripts: https://github.com/AmishSethi/visual-generative-lab.

0.49

SandBAR: A Band-Asymmetric Architecture for Underwater Image Restoration with Removable Frequency Scaffolding

cs.CV ·2026-11-17 ·Yanzi Liu, Haoran Hou, Jiachi Tong, Shen Chen, Shuang Chen
publishedbenchmark
Abstract · Accepted to WACV 2027. Camera-ready version. 9 pages, 6 figures

Light scattering, absorption and turbidity degrade underwater imagery through complementary frequency tendencies: large-scale, channel-dependent appearance distortions and local high-frequency detail loss. This band asymmetry motivates different forms of restoration for global colour and local structure. Our primary contribution is SandBAR, a novel unified band-asymmetric architecture for underwater image restoration. Rather than treating frequency processing as an isolated module, SandBAR organises the restoration pipeline around band asymmetry at the block, encoder and objective levels. Within this architecture, we introduce a novel band-asymmetric sandwich block with a reserved post-attention repair slot and the Split-band Parallel Asymmetric Repair Kernel (SPARK), a novel module that applies channel-aware large-support correction to the low band, which predominantly carries large-scale appearance distortions, and lightweight local refinement to the detail-bearing high band, thereby restoring global colour while preserving fine structure. To deliver the global cast to the depth at which the network represents global structure, a learnable low-frequency guidance path injects a low-pass view of the input into the deep encoder, and a three-scale spatial-frequency objective supervises the same axis throughout the decoder. Building on the reserved post-attention slot, we further introduce Post-attention Scaffold Training (PaST), a novel removable nonlinear scaffolding scheme that trains the permanent SandBAR graph under band-specific capacity and runs it at inference without that capacity. SandBAR reaches 29.89 / 28.92 / 22.77 dB on EUVP / LSUI / UIEB and establishes a new three-domain mean state of the art on the benchmark family, exceeding the strongest published result by approximately 1.0 dB. PaST adds no parameters or FLOPs at inference and preserves matched latency.

0.49

FastFrames: Video Generation at Interactive Latency

cs.CV ·2026-11-17 ·Gene Chou, Guandao Yang, Fujun Luan, Lars Mescheder, Wei Dong +2 more
code/data
Abstract · Project page: https://apple-aiml-research.github.io/ml-fastframes/

Autoregressive video diffusion models generate frames based on new observations and inputs, making them a promising medium for interactive visual experiences. However, existing models are orders of magnitude too slow for fluid interaction. This paper introduces FastFrames, a family of fast autoregressive video diffusion models that achieves interactive latency without compromising quality. We profile the inference pipeline of state-of-the-art models and find that operations often treated as fixed implementation details, such as VAE decoding and KV cache updates, in fact account for over half of end-to-end inference latency. We challenge the assumption that these components are indispensable and replace them with lightweight alternatives. Our 1.3B variant runs at 36 ms per frame, or 28 responses per second (RPS), on a single B200 GPU while matching the generation quality of a state-of-the-art model that is nearly 8x slower. Our 14B variant runs at 45 ms per frame, or 22 RPS, on an 8x B200 GPU node and yields significantly better visual quality and realism. We show that FastFrames enables real-time interaction in applications such as camera-conditioned and physics-conditioned video generation.

0.48

RoadAnnoter: An AI-Assisted Desktop Tool for Semantic Segmentation Annotation of Road-Scene Images

cs.CV ·2026-11-17 ·Nazmul Haque, Arnob Protim Roy
code/datahot direction
Abstract · 33 Pages

Accurate semantic segmentation of road-scene images is essential for road safety research; however, polygon-level annotation remains labor-intensive and limits labelled dataset development. This study presents RoadAnnoter v1.0, an open-source desktop application integrating six human-in-the-loop annotation modes: manual polygon delineation, SAM Prompt, Smart Select, Edge Fill, pretrained semantic segmentation using Cityscapes-trained SegFormer models, and task-specific segmentation using TinySegNet, a lightweight MobileNetV3-Small-based network trained with accumulated annotator-generated labels. SAM-based modes use promptable segmentation with SAM and MobileSAM, while Edge Fill combines Canny edge detection with colour-similarity-based flood filling in CIELAB colour space. SegFormer generates suggestions for common road-scene categories, whereas TinySegNet supports adaptation to project-specific features through iterative training. All six modes use a shared mask-to-polygon pipeline comprising morphological processing, Gaussian smoothing, contour extraction, and Douglas-Peucker simplification; annotations are exported in YOLO segmentation format. RoadAnnoter was used to annotate Google Street View images from recorded road accident locations in Bangladesh for a vision-language model-based investigation of free-flow speed estimation. Qualitative assessment across representative scenes demonstrated the complementary applicability of the annotation modes to diverse feature geometries. Visual comparison with Roboflow indicated similar qualitative boundary characteristics for principal road-scene regions, although no quantitative accuracy assessment was conducted. RoadAnnoter provides a locally operated framework for developing road-scene segmentation datasets for transportation engineering and road safety research.

0.48

HandRSI: Recursive Self-Improvement for Hand Pose Estimation

cs.CV ·2026-11-17 ·Zhekai Wang, Luqing Luo, Wenhao Yao, Junjie Lu, Huanan Zhu +2 more
benchmarkcode/data
Abstract · Technical report. 25 pages, 17 figures, 18 tables. Code: https://github.com/GinKenneth/Hand_RSI

Egocentric 3D hand pose estimators still miss hands at the image border, misplace occluded hands and jitter. Such failures are usually fixed by hand-written inference refinements and by fine-tuning, two interdependent steps that are tuned separately. We present HandRSI, a recursive self-improvement framework for a hand pose system that consists of a video estimator and its executable inference harness. In each round, coding agents refine the harness, and a change is accepted only if it improves a composite of all ten benchmark metrics on held-out data without significantly regressing any of them. The accepted harness then guides the estimator through selective distillation, which supervises hand presence and temporal consistency on frames without ground truth, while ground truth supervises pose. Since the updated estimator may change which refinements remain useful, the harness is re-validated on it. Starting from the released ACE-Ego-Hand model, five rounds reduce MPJPE-p by 35% on HOT3D and 11% on ARCTIC and roughly halve jitter. On HOT3D, HandRSI's MPJPE-p is 16% lower than that of ACE-Ego-Hand fine-tuned on the same ground truth. Under a unified protocol, it outperforms nine re-run published methods on all ten metrics of both benchmarks. Hypervolume selection on the accuracy-speed front yields Lite, Medium and Heavy systems running at 20.7 to 41.1 FPS. This report gives the full method, every stage of every round and the complete protocol. Code, evaluator, harnesses and weights are public.

0.47

Joint-ID: Transformer-based Joint Image Enhancement and Depth Estimation for Underwater Environments

cs.CV ·2026-11-17 ·Geonmo Yang, Gilhwan Kang, Juhui Lee, Younggun Cho
published
Abstract · 11 pages. (c) 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any curren…

Underwater imaging is a challenging task due to factors such as scattering, absorption, and turbulence, which degrade image quality and limit visibility. In this paper, we propose a novel approach for enhancing underwater images that leverages the benefits of joint learning for simultaneous image enhancement and depth estimation. We introduce Joint-ID, a transformer-based neural network that can obtain high-perceptual image quality and depth information from raw underwater images. Our approach formulates a multi-modal objective function that addresses invalid depth, lack of sharpness, and image degradation based on color and local texture. We design an end-to-end training pipeline that enables joint restoration and depth estimation in a shared hierarchical feature space. Additionally, we propose a synthetic dataset with various distortions and scene depths for multi-task learning. We evaluate Joint-ID on synthetic and standard datasets, as well as real underwater images with diverse spectra and harsh turbidity, demonstrating its effectiveness for underwater image enhancement and depth estimation. Furthermore, we demonstrate the ability of Joint-ID to perform feature matching and saliency detection for visually guided underwater robots. Our proposed method has the potential to improve the visual perception of underwater environments and benefit applications such as oceanography and underwater robotics. Supplements are available at: https://sites.google.com/view/joint-id/home

0.46

Practice in Text, Act in Pixels: Correcting Where Perception Fails

cs.CV ·2026-11-17 ·Zhongling Xu, Hairong Wang
code/data
Abstract · 25 pages, 8 figures, 19 tables. Code: https://github.com/Carlxuzhl/practice-in-text ; aligned reader: https://huggingface.co/xuzhl/practice-…

GUI agents can learn cheaply from a text view of an interface and act on screenshots through a vision-language screen reader. Yet on a simulated mobile platform with a low-cost reader, policies that complete most tasks in text fail on many of the same tasks on screenshots. We ask where a small budget of real screenshots is best spent to close this gap. We find that the reader's misreads are systematic and recur on the same interface elements. Noise matched to the reader's error rates, a natural sim-to-real remedy, misses this structure, and in a pre-specified comparison it did not beat training the policy directly on screenshots. We therefore correct the misreads at their source with reader alignment, which fine-tunes the reader on screenshots paired with their text views. The reader learns to output each screen's text view, as in text practice, and the policy stays unchanged. Spent on the reader, the screenshot budget gives 80.9% full-task success on new task instances, against 55.6% when spent on training the policy. Given the same labeled screenshots, adapting the reader also beats adapting the policy, and a single aligned reader improves every text-trained policy we tried without retraining any of them.

0.45

Event-Based Person Detection for Maritime UAV Search

cs.CV ·2026-11-17 ·Zijie Ning, Patrick Vandewalle
published
Abstract · Accepted at the 5th Workshop on Maritime Computer Vision (MaCVi), WACV 2027

Maritime search and rescue requires locating people in the water quickly, motivating the use of UAVs to survey the search area. During flight, camera motion and the longer exposures used in low light can blur small targets and hinder person detection. Event cameras record brightness changes as they occur, reducing the motion blur associated with longer exposure times. We investigate their use for maritime person detection by generating synthetic events using the v2e event simulator from public MOBDrone rescue videos and retaining the existing annotations. We compare the recurrent vision transformer model RVT-T with YOLOv8n and RT-DETR-R18 trained on exposure-matched images. On 843 test samples from held-out recordings, RVT-T achieves 15.63% person AP50:95, compared with 6.89% for YOLOv8n and 9.44% for RT-DETR-R18 under 100 ms synthetic exposure. Increasing exposure times from 50 to 200 ms progressively reduces frame-based detection accuracy, widening the gap to the event model. To evaluate the feasibility of on-board processing, we deploy our detector on a Jetson Orin Nano platform. The RVT-T model retains its accuracy and requires approximately 136 ms and 2.02 J per sample, allowing real-time processing on the UAV.

0.44

Cross3DGS: Unified Vulkan-Based 3D Gaussian Splatting Training from Workstations to Mobile Devices

cs.CV cs.GR ·2026-11-17 ·Hiroyuki Sato, Yusuke Takimoto
publishedhot direction
Abstract · Accepted to the WACV 2027 Workshop on Eyes of the Future (EOF). 19 pages including appendix

Scenes captured with handheld and wearable cameras are still shipped to a CUDA workstation for 3D Gaussian Splatting (3DGS) training. Training on the device that captured them, or on its companion phone or laptop, requires more than translating kernels: GPUs and drivers differ in parallel features, memory limits, and command submission, and these differences decide whether a training schedule executes and returns the correct update. We present Cross3DGS, a cross-vendor, cross-platform Vulkan trainer that keeps one optimizer and chooses its execution schedule per device in two steps: rules on device limits, workload size, and observed driver failures exclude the schedules known not to run correctly, and a fixed preference order selects one of the remaining feasible schedules. The same core completes standard-quality training (30k updates, one million Gaussians) on every configuration we measured, spanning NVIDIA, Intel, Arm, Qualcomm, AMD, and Apple GPUs and including Android devices and an iPhone, whereas the existing portable trainers VkSplat and Brush do not run on some of the mobile GPUs. Averaged over 11 scenes on a workstation, Cross3DGS has higher quality than VkSplat with no significant difference in training time, and Brush reports about three times the training time at comparable quality. The final quality of Cross3DGS is nearly identical across devices. A mobile application turns a handheld capture into a viewable model entirely on the device.

0.44

Efficient Self-Supervised Adaptation of 3D Abdominal Vision-Language Model for Institution-Specific HCC Classification via Full Fine-Tuning and PEFT

cs.CV cs.LG ·2026-11-17 ·Febryan Putra Kartika, Cheng-Yu Ma, Ying-Jia Lin, Chi-Tung Cheng, Kuan-Fu Chen
published
Abstract · 25 pages, 12 figures, 7 tables, including appendices. Corrected author revision of the MIDL 2026 paper. Revised preprocessing yields lower H…

Medical vision-language models (VLMs) capture cross-modal relationships between image and text, yet their adaptation to institution-specific clinical tasks remains underexplored. In this study, we fine-tuned a pretrained 3D medical VLM for hepatocellular carcinoma (HCC) classification using paired abdominal CT scans and radiology reports from a different institution with acquisition characteristics that differ from the model's original pretraining corpus. We compared two adaptation strategies: full fine-tuning and parameter-efficient fine-tuning (PEFT), motivated by the common use of PEFT to reduce computational cost and enable adaptation under limited-data constraints. Both approaches achieve strong downstream HCC classification performance despite the cross-institutional domain shift, with PEFT reaching an AUC of 0.79 and F1 of 0.70, and full fine-tuning achieving an AUC of 0.82 and F1 of 0.71. Although these results remain below the performance reported by supervised HCC classification approaches, our method achieves improvements over the pretrained model through CT-report contrastive learning without requiring lesion-level annotations or HCC classification labels as training targets. Full fine-tuning converges rapidly but overfits within a few epochs, whereas PEFT (ConvLoRA for the image encoder and LoRA for the text encoder) attains comparable performance while updating approximately 1% of the model parameters, although requiring more training steps. We also examine the role of contrastive temperature, observing that temperature initialization significantly affects classification performance. This study demonstrates that a 3D medical VLM can be efficiently adapted to institution-specific HCC classification using self-supervised CT-report contrastive learning, while highlighting the practical trade-offs between full fine-tuning and parameter-efficient fine-tuning.

0.42

Multimodal Masked Diffusion Autoencoder for Modeling Impaired Driving

cs.CV cs.LGcs.RO ·2026-11-17 ·John Gideon, Anshul Gupta, Megan Applegate-Kenton, Kayli Battel, Todd J Rowell +5 more
dataset
Abstract · Equal contribution: John Gideon and Anshul Gupta. 8 pages, 6 figures. Submitted to CVPR 2027

Human cognitive state (e.g., distraction, stress), gaze, and behavior interact with the driving environment in subtle ways that AI could use to assist drivers, but ground-truth labels for human cognitive state levels are costly to obtain at scale. We propose a multimodal masked-diffusion autoencoder, ID-MAE, that adds a self-supervised objective to impairment prediction: alongside predicting the impairment label as a masked token, it reconstructs masked gaze and trajectory. To probe the effects of different modalities and types of impairments on model inference, we introduce DIVERT, a driving-simulator dataset containing 103 hours of driving data (n = 109) with ten impairment inductions spanning cognitive, visual, and emotional states, recorded with gaze, vehicle state, and scene video. We demonstrate how ID-MAE affords significant improvements in impairment inference results on both the DIVERT dataset and existing datasets, and enables strong results in additional tasks, such as trajectory and gaze prediction. For location-matched pairs, our model achieves a paired AUC of 0.706 for impairment detection on DIVERT, compared to 0.590 without reconstruction loss, improving over existing baselines. We probe the design space in terms of input modalities, masking scheme, positional encoding, attention mechanism, and decoding. We also investigate the interaction between various impairments in training and inference, demonstrating that visually demanding distractions are separable from cognitive distractions. Code and data will be available upon acceptance.

0.42

ConserveEdit: Source-Anchored Transactions for Drift-Bounded Iterative Image Editing

cs.CV ·2026-11-17 ·Shuyang Jiang, Fucheng Deng, Yuchuan Luo, Zhenyu Wu
Abstract · 28 pages, 15 figures, 16 tables; includes supplementary material

Instruction-driven image editors produce convincing local edits in one pass, but inside an iterative refinement loop every accepted output becomes the next input, so unintended changes outside the requested region accumulate. We present ConserveEdit, an editor-agnostic controller that treats each refinement step as a transaction against a contract compiled once from the source image. The contract separates what may change (the authorized region of each target, with a bounded halo for shadows and reflections) from what must survive (protected content with explicit tolerances). Candidates from a frozen editor are composed through a source-anchored mask and committed only if target progress, drift on protected content, and image quality pass as separate tests; otherwise the step is rolled back and repaired. For any editor, this keeps every pixel outside the authorized region bit-exact and bounds drift independently of the number of iterations, whereas no fixed scalar trade-off reproduces the separate tests. On all 1053 MagicBrush test edits with a frozen Qwen-Image-Edit, masking is necessary but not sufficient. A direct call already alters the rest of the image (12.1 dB PSNR outside the mask); composed through the source-anchored mask, it stays bit-exact there and its DINO similarity to the ground truth rises from 0.67 to 0.85, matching ConserveEdit, so this gain comes from the mask. Next to automatic masks, however, masked direct editing changes a protected face or text on 48% of turns, and masked scalar selection avoids this mostly by abstaining (23% constrained success). Because its drift test names the violated entity, ConserveEdit shrinks the support instead (2% violations, 53% constrained success); in a simulator with known ground truth it raises constrained success from 55.0% to 61.8%. Automatic grounding remains the main bottleneck. Code and logs will be released.

0.42

Let Samples Sing in Chorus: Mitigating Spurious Correlations via Cross-view Consensus

cs.CV ·2026-11-17 ·Weiwei Li, Junzhuo Liu, Tong Chu, Hengfu Yu, Yinjie Lei +1 more
code/data
Abstract · Under Review

Deep models often rely on spurious correlations that fail on underrepresented groups. Existing methods either correct human-specified bias attributes or infer unknown biases from the learned representation space. However, annotated attributes may differ from the shortcuts actually used by a model, while a single representation provides incomplete and potentially noisy evidence. Our key insight is that a bias obscured in one feature space may be clearer in another. In this work, we propose CHORD (Cross-view Harmonic Outlier Reversal for Debiasing), a multi-view framework that allows samples to "sing in chorus" across diverse representation spaces to achieve robust cross-view consensus. CHORD aggregates aggregates within-class partitions in each view, which is a proxy for specifying candidate biases, to debias the target representation and classifier without bias supervision. Experiments on seven benchmarks across vision, language, speech, and medical imaging show consistent improvements in worst-group accuracy. Code and checkpoints are available at https://github.com/davelee-uestc/chord_debiasing_public .

0.42

Diagnosing and Correcting Skin-Tone Color Casts in Ultra-Lightweight Photo Retouching Networks

cs.CV ·2026-11-17 ·Choulhyouc Lee
published
Abstract · Accepted to ACCV 2026. 32 pages including supplementary material

Color casts are expected to come from the scene or camera, and a retouching model removes them. We report a case where the model leaves one: an ultra-lightweight retouching network renders human skin greener than the original because it under-restores the skin chroma that a degradation removes and, even on clean inputs, adds a small hue shift of its own. A common paired-data-free recipe, synthetic self-supervision (degrade high-quality images, learn to restore them), silently produces this cast while whole-frame PSNR/SSIM stay essentially unchanged. In a representative 6,825-parameter histogram-conditioned network, the cast appears in 52 of 70 portraits under a ground-truth-anchored, white-balance-free protocol. Both a histogram-conditioned and a CNN-conditioned network develop the cast under synthetic self-supervision, yet both are near-neutral in-domain when trained on MIT-Adobe FiveK expert pairs. Image-level objectives miss it: a global-mean green loss worsens the cast (52/70 to 57/70). With zero added parameters, our per-pixel, skin-masked, asymmetric green-excess loss reduces the cast to 1/70 across stratified skin tones (68.1% to 2.2% over the 70,000-image FFHQ corpus; 64.4% to 0.8% on held-out CelebA-HQ photos). In-domain FiveK PSNR/SSIM are unchanged (+0.13 dB, +0.004), at the cost of a small whole-frame CIEDE2000 increase (8.77 to 9.69) and a magenta over-correction where skin is nearly neutral (skin Delta a* -1.5 to +4.4 and CIEDE2000 5.6 to 6.4 on clean inputs): the loss targets the signed green-magenta axis, not total skin-color fidelity. The same loss suppresses the cast across three ultra-light architectures. A whole-frame green metric correlates with skin-region CIEDE2000 at the method level (Spearman rho = 0.89). The final network exports to a 35 KB TFLite model.

0.41

MAVE: Reconstructing Editable 3D Blockouts from Film Shots

cs.CV cs.AI ·2026-11-17 ·Yifan Chang, Jian Yang, Jiangnan Shao, Xiaofeng Wang, Zhaoxin Fan +4 more
benchmarkcode/data
Abstract · https://github.com/yfChang-cv/MAVE

Animated 3D blockouts support previsualization and provide structural conditions for video generation, but manually reconstructing them from reference videos is costly. This limits both their use by creators and the production of paired training data. We introduce MAVEBench, a benchmark of 50 film shots and 416 keyframes for evaluating camera motion, spatial layout, and character blocking. We propose MAVE, a video-to-blockout framework combining Metric observation, Agentic modeling, and evidence-based VErification. It first recovers cameras, human motion, and geometric evidence in a shared metric coordinate system. An agent then constructs an editable environment under constraints from these observations, while programmatic checks report inconsistencies to guide refinement. The outputs are editable Blender scenes and frame-aligned blockout videos. On MAVEBench, MAVE improves overall scores by 27.5-41.1 points across three agents over their corresponding vanilla baselines. With DeepSeek-1017, it achieves a score of 73.2 at an average API cost of only USD 0.054 per shot, outperforming the strongest vanilla baseline while reducing total reconstruction time by 53%. We also construct MAVEData-3K, a dataset of 3,000 blockout-video pairs for model training, and will release it publicly. Creators can change viewpoints and character paths, then render the edited scenes to condition video generation. Code is available at https://github.com/yfChang-cv/MAVE.

0.41

Data-Efficient Algal Bloom Segmentation in Freshwater Bodies Using a Three-Stage Transfer Learning Pipeline with Attention-Enhanced Decoding

cs.CV ·2026-11-17 ·Albert Ansah Benson, Mingqiang Han, Rahul Harsha Cheppally, Treman Singh Kaloya, Menard Soni +1 more
Abstract · 10 pages, 5 figures, 6 tables

Harmful algal blooms threaten water quality and public health, yet automated segmentation of bloom regions in freshwater imagery is limited by scarce pixel-level annotations. We present a transfer-learning pipeline that adapts a Swin Transformer U-Net with SCSE decoder attention from ImageNet-1K to water-body segmentation (RIWA, LaRS) and then to algal blooms, together with a model-assisted annotation protocol that expands the labelled algae corpus from 145 to 655 images. In a single-seed ablation on a 41-image validation split (checkpoint S2-A4 and its variants), corpus expansion is the dominant factor (+11 mIoU points); water-body pretraining helps only on the expanded corpus (+1.1 points over ImageNet initialization); SCSE attention adds 0.3 points; and a learnable positional encoding lowers mIoU by 0.6 points and is omitted. Adding 476 images, 188 of them algae-free, reduces false-positive area on non-water scenes from 22.6% to 4.9%. Against nine baselines on a 256-image test split, a separately trained checkpoint of the same architecture (S2-A13) reaches mIoU 0.919 and the highest algae IoU (0.902); a near-duplicate audit indicates train/test overlap in this split, so these scores are likely optimistic.

0.41

FMT: Lossless Image Compression in a Fixed Number of Parallel Steps

cs.CV ·2026-11-17 ·Ammar Ali
position
Abstract

Learned lossless codecs make smaller files than common classical codecs, yet the images we archive and train on still use classical codecs. A main reason is time: a learned codec takes seconds to decode a photo that a classical codec opens in milliseconds, and their faster variants give up most of their gain in size. We argue that much of this time is spent waiting, not computing: the network waits for the values it conditions on, and the arithmetic decoder reads them one at a time. We introduce FMT (Fixed-step Multiscale Tree), a lossless image format whose decoding order and entropy coder are designed together so that neither wait grows with the image. FMT stores an image as a coarse-to-fine tree whose steps hold independent values: a GPU predicts each step at once and decodes it in parallel rANS lanes that carry their states from step to step, so an image of any size decodes in a fixed number of sequential steps. On this tree we build a family of bit-exact codecs, from a rule with no network to large networks, and no codec we ran on the same GPU beats any of them on both size and speed. Without a network, FMT beats NVIDIA's HTJ2K on size and speed; with a large one, it decodes Kodak photos 9 to 61X faster than learned codecs we ran on the same A100, at the same size or smaller.

0.40

Unsupervised Multi-modal Medical Image Registration via Invertible Translation

cs.CV ·2026-11-17 ·Mengjie Guo, Jinming Duan
code/data
Abstract

In medical imaging, the alignment of multi-modal images plays a critical role in providing comprehensive information for image-guided therapies. Despite its importance, multi-modal image registration poses significant challenges due to the complex and often unknown spatial relationships between different image modalities. To address this, we introduce a novel unsupervised translation-based multi-modal registration method, termed Invertible Neural Network-based Registration (INNReg). INNReg consists of an image-to-image translation network that converts multi-modal images into mono-modal counterparts and a registration network that uses the translated mono-modal images to align the multi-modal images. Specifically, to ensure the preservation of geometric consistency after image translation, we introduce an Invertible Neural Network (INN) that leverages a dynamic depthwise convolution-based local attention mechanism. Additionally, we design a novel barrier loss function based on Normalized Mutual Information to impose constraints on the registration network, which enhances the registration accuracy. The superior performance of INNReg is demonstrated through experiments on two public multi-modal medical image datasets, including MRI T1/T2 and MRI/CT pairs. The code is available at https://github.com/MeggieGuo/INNReg

0.40

REVEAL: Reconstructing Video Amodal Content via Language and Flow Guidance

cs.CV ·2026-11-17 ·Minh Tran, Winston Bounsavy, Ngan Le
top venuebenchmarkcode/data
Abstract · Accepted to Transactions on Machine Learning Research (TMLR)

Amodal perception enables humans to perceive entire objects even when parts are occluded, a remarkable cognitive skill that artificial intelligence struggles to replicate. Recent diffusion-based methods have extended amodal completion from images to videos, yet they lack auxiliary priors to guide the reconstruction of heavily occluded content and to maintain temporal consistency over long sequences. We present REVEAL (REconstructing VidEo Amodal content via Language and flow guidance), a unified diffusion-based framework that integrates complementary motion and semantic priors for video amodal completion. A user-provided text query serves a dual role: it drives open-vocabulary video segmentation to obtain visible masks, and it provides semantic guidance for texture reconstruction. For amodal segmentation, we introduce optical flow guidance: by warping visible masks from previous frames to the current frame, the flow-warped mask propagates visible content into occluded regions and approximates the object's current shape, providing a strong shape prior even under simultaneous deformation and occlusion. For texture completion, text guidance constrains the reconstruction of occluded appearance while providing a stable semantic reference that helps maintain visual coherence throughout the reconstruction. We also contribute LAVAT, a new benchmark featuring long video sequences paired with text descriptions, enabling evaluation of text-guided video amodal completion under heavy occlusion. Extensive experiments demonstrate that REVEAL achieves state-of-the-art performance on existing benchmarks while maintaining high-quality temporal consistency over extended sequences. The project page is available at: https://uark-aicv.github.io/REVEAL/.

0.39

Cross-Modal StateKV Injection for Communication in Multi-LLM Systems across Temporal Segments

cs.CV ·2026-11-17 ·Xiaoqian Zhang, Zhihao Zhang
hot direction
Abstract · 9 pages, 4 figures, 5 tables. Equal contribution: Xiaoqian Zhang and Zhihao Zhang

Video understanding with multiple multimodal language models requires a communication interface that can transfer complementary visual observations. Text-based communication is interpretable, but introduces an intermediate generation stage and can omit information relevant to subsequent reasoning. We study Cross-Modal StateKV Injection, a provider--carrier framework that communicates through key--value (KV) states at visual-token positions. Provider models observe complementary video frames; their visual KV states are aligned and projected into the carrier representation space and integrated through gated residual updates. Injection preserves the carrier KV-cache sequence length, while retaining the additional computation required by the providers and the communication module. The reported experiments on MLVU and Video-MME compare carrier-only inference, text-based communication, and StateKV with two provider architectures. Results show modest, task-dependent accuracy gains and lower reported latency than the evaluated text-communication baseline, with some categories remaining unchanged or decreasing. Increasing provider-side frame coverage offers additional gains on selected tasks without extending the carrier cache. These findings motivate further study of visual KV communication under explicitly controlled observation and computation budgets.

0.39

A tiny object detection dataset for high-altitude open-water search and rescue

cs.CV cs.AI ·2026-11-17 ·Hui Xu, Chenfei Hou, Henglai Wei, Hao Wang, Yidong Jin +4 more
benchmark
Abstract

Large aerial platforms often operate above 1.5 km to improve the efficiency of open water search and rescue (OW-SAR). However, long imaging distances, wide fields of view, and limited sensor resolution reduce objects to a few pixels, increasing their susceptibility to environmental interference and making detection harder than for conventional small objects. To our knowledge, existing datasets focus on low-altitude scenarios and do not support automatic detection at 1.5-2.0 km. We therefore present TinyWOD, a tiny object detection dataset for high-altitude OW-SAR using lifebuoys as detection targets. It comprises 11,359 manually annotated images and 19,505 annotated instances spanning diverse environments, lighting conditions, water colors, and surface ripples. Compared with existing datasets, TinyWOD contains much smaller objects and a substantially higher proportion of tiny instances; its average equivalent side length (ESL) is 12.7 pixels, with 76.96% of in stances having ESLs below 15 pixels, reflecting actual high-altitude water-rescue imaging conditions. We define APtiny, a tiny object detection metric, and benchmark mainstream and state-of-the-art networks using this and conventional metrics. All evaluated networks effectively detect tiny waterborne objects, demonstrating TinyWOD's reusability and utility. We expect this dataset to support research on tiny object detection in high-altitude OW-SAR.

0.39

VIVID: Beyond Static Multi-Image Reasoning with Localized Visual Differences and Textual Entity Dynamics in Video MLLMs

cs.CV ·2026-11-17 ·Chengqian Ma, Yuanze Zeng, Hongyu Wang, Mingfeng Lin, YiCheng Xiao +6 more
code/data
Abstract · 14 pages, 2 figures, 5 tables. Code: https://github.com/step-out/VIVID

Video question answering under tight frame budgets requires preserving localized motion and event order, yet current video multimodal large language models (MLLMs) reduce comprehension to static multi-image reasoning by discarding unselected frames and encoding keyframes independently: (1) Spatially, whole-frame pooling dilutes small moving targets into static backgrounds; (2) Temporally, dropping intermediate frames or averaging tokens erases inter-frame motion, confusing direction, speed, and order. We propose VIVID (Visual Inter-frame Variations and Individual Dynamics), which moves beyond static snapshots like human vision: keyframes capture core states while inter-frame Patch Diffs preserve neighboring changes within $K_f$ keyframes. Autonomous Delta-Driven Keyframe Planning (ADKP) selects anchors from a $3\text{ fps}$ pool and feeds surrounding patch differences into Dual-Road Change Memory (DRCM), forming Dynamic Tokens at zero extra anchor tokens alongside a camera-stabilized Active Motion Region of Interest (ROI) and chronological Entity State Chains. Adding $0.22\%$ trainable parameters to a frozen $27\text{B}$ backbone, VIVID-Small ($50$ examples, $\sim 1.5$ GPU hours) beats all eleven baselines in overall accuracy across $8,083$ questions and trails only VIVID-Large ($3,600$ examples), which leads on MotionBench, Video-MME, and TempCompass.

0.39

MedImgAnnotator: batch wise medical image annotation and preprocessing tool

cs.CV eess.IV ·2026-11-17 ·Ewunate Assaye Kassaw, Amit Mehndiratta, Peush Sahni, Anup Singh
code/data
Abstract

The curation of large cohort datasets for developing deep learning models demands annotating or editing over the AI-predicted labels. However, openly available annotation tools are time-consuming, as they involve complex navigational steps and do not support batch processing. This study aimed to develop a simple, fast, semiautomated batch-annotation tool named MedImgAnnotator. The tool was created by integrating editing tools, image mask pair matching logic, preprocessing functions, object detection tool, and a plugin management system. Its performance was evaluated in terms of active annotation time and ease of use by the developer and 12 users using open-source MR and CT scans. Additionally, the tool had been used by the MICCAI 2026 CARE Challenge liver segmentation best performance winner to annotate liver masks from 6412 MRI volumes. The evaluation results showed that MedImgAnnotator reduced active annotation time from 206.5+/-32.87 (average active annotation time of openly available tools) seconds to 59 seconds. The Wilcoxon signed-rank test of 12 users evaluation results recorded P values of 0.032 and 0.008 for the active annotation time and higher ease of use rating metrics, respectively, indicating that the current tool significantly reduces annotation time and enhances ease of use. Moreover, the tool utilization for annotating different regions of interest from different modality scans and body regions shows its usability across modalities and areas of interest. MedImgAnnotator, which significantly reduces active annotation time for large imaging cohorts, is available as an executable file and as a Python script. Access link: https://ewunateassaye.github.io/developed-tools/medimgannotator.html

0.39

PlaceTally: A Test-Time Memory Model for Constructing Place Graphs

cs.CV ·2026-11-17 ·Alibay Osmanli, Shaogang Gong
benchmark
Abstract · 12 pages, 3 figures, 5 tables. Code, data and supplementary material: https://alibayosmanli.github.io/placetally

Counting distinct objects in a video is hard when the camera returns to places it has seen. A vision-language model that reads the whole video answers with a few small numbers, and one that reads it in parts counts an object again in every part that shows it. Object trackers need the category before the video is read, or a fixed vocabulary. We observe that a memory for counting does not need to know what it is looking for, only where it has been: objects rarely leave a room between visits, so parts of a video grouped by place share almost no objects, and a place can be recognised before any question is asked. Building on this, we introduce PlaceTally, a question-blind test-time memory model that constructs a place graph for each video. Each video fragment joins the place of its best earlier match or opens a new place, because a strong match is rarely wrong while a true return often matches weakly. A frozen vision-language model counts each place once from at most 128 of its frames, so every read has a bounded input and the memory grows with the number of places; the graph is built 15 times faster than real time. On RevisitCount, our benchmark of real room re-recordings, it scores 62.5% mean relative accuracy on test rooms and 61.2% on held-out rooms, against 51.2% and 44.3% for the same model reading the whole video. It stays level as rooms return more or less often, while a fixed correction of summed visits moves from 38.5% to 61.4%.

0.39

Sub-Meter Canopy Height Synthesis from Sentinel-2 via Anchored Residual Flow

cs.CV cs.LG ·2026-11-17 ·Kiarie Ndegwa, Tony Chang, Andreas Gros, Vincent A. Landau, David D. Diaz +3 more
Abstract · 30 pages, including appendices

Canopy height models (CHMs) - per-pixel maps of vegetation top height - are essential for forest carbon accounting, wildfire fuel mapping, and tree-level ecology, yet sub-meter CHMs require airborne LiDAR refreshed on multi-year national cycles. We produce 0.5 m CHMs from free 10 m Sentinel-2 imagery, a 20$\times$ cross-modal super-resolution task where optical reflectance must predict geometric height it does not directly observe. Motivated by the perception-distortion trade-off, we derive two operating points with shared conditioning. A deterministic regressor (Anchor) targets low distortion; a conditional flow matching model, trained on the Anchor's residual in a learned CHM latent space (AR-flow), targets perceptual fidelity. Trained on 197k LiDAR-paired tiles spanning 69 ecoregions across the contiguous United States and evaluated on 2,530 held-out tiles against eight baselines, the Anchor achieves 2.83 m MAE using only free satellite input. The AR-flow reduces sliced Wasserstein distance from the Anchor's 1.13 to 0.208, at a higher MAE of 3.68 m. Crown-scale statistics improve, but strict individual-tree detection remains weak: generated sub-pixel detail represents plausible structure rather than measured individual trees. Quarterly continental refresh becomes tractable, letting satellite monitoring inform timing and extent for more expensive data acquisitions.

0.38

SARVIA: Fast-Adapting the Strong SAR Foundation Models That Are Already in Your MLLMs

cs.CV ·2026-11-17 ·Yuxuan Li, Zichen Tian, Qianru Sun
code/data
Abstract

Synthetic aperture radar (SAR) foundation models provide transferable representations for diverse perception tasks in all-weather Earth observation. However, their development often relies on costly large-scale SAR pretraining, raising the question of whether such extensive domain-specific pretraining is necessary. We find that modern multimodal large language models (MLLMs) already provide visual encoders that can outperform dedicated SAR foundation models as downstream initializations. This motivates a new paradigm: reuse general visual capabilities and devote domain training to brief specialization. We therefore introduce SARVIA, an adaptive post-pretraining method that specializes modern MLLMs using self-supervised visual questions with computable answers. These questions are constructed from unlabeled observations through general image tasks that capture spatial relations and SAR-specific tasks that exploit native measurements and SAR--optical correspondence. A shared answer-prediction objective jointly trains low-rank adapters across visual and language components, with loss-driven task sampling. We then export the adapted visual encoder as a standalone backbone. With 600 adaptation steps and 2 GPU-hours, SARVIA outperforms the compared foundation models on all six SAR detection and segmentation benchmarks. Compared with dedicated SAR pretraining, SARVIA reduces domain-training costs by orders of magnitude while improving visual transfer across architectures and scales. This new paradigm shifts the emphasis from repeated, costly pretraining to a reusable specialization process that can benefit from advances in open MLLMs. Dataset and code: https://github.com/zcablii/SARVIA.

0.38

M3OPD-Reasoner: An Open Recipe for Post-Training Multimodal Reasoning with Multi-Teacher On-Policy Distillation

cs.CV ·2026-11-17 ·Nguyen Quang Trung, Hoang Phan, Shuo Sun, Longyin Zhang, AiTi Aw
Abstract

Multi-Teacher On-Policy Distillation (MOPD) enables a single student to absorb the skills of several domain teachers by learning from its own generated responses. We present Multimodal Model Multi-Teacher On-Policy Distillation Reasoner (M3OPD-Reasoner), a fully open recipe for post-training vision-language models (VLMs) over three reasoning domains: image STEM, image perception, and video. We curate two datasets, M3OPD-SFT-1.35M for supervised fine-tuning (SFT) and M3OPD-RL-193K for reinforcement learning (RL) with verifiable rewards. On these data, we compare post-training recipes that start from Qwen3.5-9B: training one model on the mixed data of all domains (MixSFT and MixRL), and training one teacher per domain with Group Sequence Policy Optimization (GSPO), directly or after an SFT warm-up, and distilling the teachers into one student with MOPD. We evaluate on 21 image STEM, image perception, and video reasoning benchmarks after we decontaminate the training data against them. The comparison establishes MOPD of teachers trained with GSPO directly from Qwen3.5-9B as the best recipe: the student closely matches each teacher on its own domain and is on average above the mixed-training baselines on the same prompts. Since MOPD already merges the domain capabilities that MixRL would otherwise have to learn, its student is a stronger starting point for RL, and we ask whether the same data can carry the model further from there: we add MixRL on top of the MOPD student, on the same prompts and with the same settings, which, to our knowledge, no prior VLM recipe has studied. The same prompts still improve the MOPD student, with no collapse, and give our main model, M3OPD-Reasoner-9B. MOPD and MixRL can therefore work as consecutive stages that get more out of the same data. We release all data, the training code, and the checkpoints of every stage.

0.38

Can Vision-Language Models Truly Understand Images? A Benchmark Beyond Commonsense Bias

cs.CV ·2026-11-17 ·Ling Lin, Heng Su, Xiangzeng Liu, Yang Bai, Congcong Zhu +1 more
benchmarkcode/data
Abstract · 10 pages

Multimodal large language models (MLLMs) have achieved impressive performance on a wide range of vision-language tasks. However, it remains unclear whether their predictions are grounded in the visual content itself or are largely driven by statistical co-occurrence patterns and commonsense correlations learned from large-scale image-text corpora. Existing vision-language benchmarks predominantly consist of samples that conform to human commonsense, allowing models to achieve strong performance by exploiting dataset priors rather than genuinely understanding image content. To address this limitation, we present ContraSense, a systematically constructed benchmark for evaluating faithful vision-language understanding. ContraSense consists of carefully curated counter-intuitive image-text pairs that deliberately violate common expectations, substantially reducing the effectiveness of statistical biases and requiring models to rely on visual evidence for reasoning. We conduct comprehensive evaluations on a diverse set of representative open- and closed-source MLLMs. The results reveal a substantial performance gap between current MLLMs and humans: even the strongest closed-source model achieves only about 60\% accuracy, compared with 96.04\% for human participants. These findings demonstrate that current MLLMs still struggle with faithful visual understanding and highlight the need for more robust multimodal reasoning grounded in image content. The ContraSense benchmark is publicly available at https://huggingface.co/datasets/LingLin00/ContraSense.

0.37

TANGO: Real-World Unregistered RGB-Guided Hyperspectral Super-Resolution with Two-Camera Acquisitions

cs.CV ·2026-11-17 ·Matteo Kolyszko, Marco Buzzelli, Simone Bianco, Raimondo Schettini
benchmarkdatasetcode/data
Abstract · 18 pages, 11 Figures, 10 tables. Submitted to CVPR 2027

Unregistered hyperspectral image super-resolution aims to recover a high-resolution HSI from a low-resolution hyperspectral observation and an unregistered high-resolution RGB reference. Although recent methods have begun to address the geometric misalignment between the two modalities, their evaluation is still constrained by the lack of a large-scale public benchmark based on real two-camera RGB--HSI acquisitions with a high-resolution HSI reference. To address this gap, we introduce DUET, the first large-scale public dataset for unregistered RGB-guided hyperspectral super-resolution, comprising 403 indoor and outdoor scenes captured with two separate RGB and hyperspectral cameras. We further propose TANGO (Two-camera AligNment with Gated Occlusions), which first compensates for the dominant cross-camera displacement through end-to-end global alignment and then refines the remaining inconsistencies with a coarse-to-fine local feature alignment. A learned gating mechanism further attenuates the RGB guidance where the two views cannot be consistently matched. Experimental results show that TANGO achieves state-of-the-art performance on both DUET and REAL across all evaluated scale factors, outperforming the strongest competitor on DUET by up to 2.29 dB in PSNR while requiring 2.5$\times$ fewer FLOPs. The code and dataset will be available at https://github.com/FagiolottoConSombrero/TANGO-DUET

0.37

SlideJev: Order-Stable and Tissue-Grounded Jev Decisions for Whole-Slide Pathology

cs.CV ·2026-11-17 ·Lei Wu, Jiashuai Liu, Di Zhang, Chunze Yang, Jiusong Ge +7 more
code/data
Abstract · 10 pages, 6 figures, 3 tables

Whole-slide visual question answering (VQA) offers a natural-language interface to pathology, allowing diagnostic and phenotypic questions to be answered directly from tissue slides. However, current vision-language models (VLMs) rely heavily on linguistic priors: their answers are sensitive to the order of the candidate answers, and can be reached through shortcuts that bypass the tissue evidence. To address these challenges, we introduce SlideJev, a framework that separately verifies candidate answers against the slide and learns to ground these judgments in tissue evidence. To connect parallel candidate evaluation with visual evidence learning, we adopt Jev and score each candidate by the difference between support and rejection log-likelihoods from the original language-model head. We compute these scores in parallel using separate prompts with the same format and no option-position labels, then normalize them for answer selection. To ground these scores in tissue evidence, candidate-conditioned reading and comparison guide gated updates to visual tokens across decoder layers. With the backbone frozen, we train these visual modules through the same scores using question-answer and paired-image supervision. Across three whole-slide pathology VQA benchmarks, SlideJev achieves the highest mean overall accuracy and 100% answer consistency across four option orders. Image-control experiments further show that SlideJev achieves the largest average accuracy gains from correct images over mismatched and image-free inputs across the three datasets. Our code is publicly available at https://github.com/leowu-ai/SlideJev.

0.37

Rescaling OLS-Merged Adapters To Improve Personalized Image Generation

cs.CV ·2026-11-17 ·Leander Girrbach, Niklas Schweiger, Zeynep Akata
Abstract

Merging multiple adapters into a single image generation model reduces deployment costs while preserving diverse task capabilities or personalization modes. State-of-the-art methods such as RegMean, Mix-of-Show, and Subspace Routing fit the outputs of the merged model to those of the individual adapters using ordinary least squares (OLS). We show that this reconstruction objective systematically shrinks task-specific signals whenever the activation distributions of different tasks overlap, producing a merged adapter that is too weak for downstream use. We derive a decomposition that quantifies this shrinkage and propose rescaling the merged update to restore signal strength. On multi-subject personalization with up to 30 merged adapters, rescaling substantially improves subject fidelity over unscaled baselines. Our approach also improves results on style transfer, concept erasure, image classification, and LLM-based text classification.

0.37

Inertial Video World Models

cs.CV ·2026-11-17 ·Yuan Gao, Fan Xu, Yu Guo, Yuxu Lu, Zhengru Fang +2 more
code/data
Abstract

The physical world has inertia: it evolves from what already exists rather than being recreated at every instant. However, many existing video world models lack an explicit mechanism for inertial state evolution, leaving the continuity of motion and structure implicit in video prediction. During autoregressive rollouts, repeated reconstruction can accumulate errors and destabilize the predicted dynamics. To address this limitation, we introduce InertialWorld, an Inertial Video World Model that makes inertia a structural prior for autoregressive prediction: existing scene evidence is carried across time rather than reconstructed at every step. Central to this formulation is Inertial Memory, which turns memory from passive predictive context into an evolving video state with two complementary views: a dense Eulerian grid preserves scene-aligned evidence, while persistent Lagrangian carriers bind local content to learned canonical coordinates. At each step, the model transports the carrier content according to action-conditioned motion, predicts the complete next video state from the transported carriers and dense grid, and uses a learned gate to combine transported and newly predicted content into the memory for the next step. Extensive experiments across multiple domains, including robot manipulation, 3D game simulation, and open-world navigation, demonstrate that InertialWorld achieves competitive autoregressive prediction performance against strong video world model baselines. Codes link: https://github.com/YuanGao-YG/InertialWorld

0.36

You Only Encode Once: Sub-Pixel-Shifted Frames for Arbitrary-Scale Remote Sensing Super-Resolution with 2D Gaussian Splatting

cs.CV ·2026-11-17 ·Canjie Liu, Liuqian Wang, Qi Ming, Wei Lu, Xiaohan Hao +3 more
Abstract

Arbitrary-scale super-resolution (ASSR) of remote sensing images remains challenging due to the limited high-frequency information in a single low-resolution (LR) observation. While multi-frame SR can exploit complementary sub-pixel information, existing approaches often require real multi-frame training sequences and repeated feature encoding, increasing data acquisition demands and computational costs. To address these limitations, we propose SPS-GS (Sub-Pixel-Shifted Gaussian Splatting), an encode-once framework that learns from multi-phase LR observations synthesized from individual high-resolution images, without requiring real multi-frame training sequences. Specifically, Phase-Coherent Aliasing Demodulation (PCAD) exploits shift-dependent aliasing through Fourier-based alignment, phase-aware residual fusion, and least-squares aliasing estimation, enabling multi-frame information to be integrated before a single deep encoding pass. Gaussian Projection Alignment (GPA) then iteratively refines Gaussian centers and colors using multi-observation back-projection to promote observation consistency. The refined Gaussians support continuous rendering at arbitrary scales. Extensive experiments on five remote sensing datasets demonstrate strong reconstruction performance against single-image and multi-frame SR baselines, while maintaining competitive inference efficiency at common scales. The code will be publicly released.

0.36

Wall-to-wall maps of canopy height and aboveground carbon across the Atlantic Forest

cs.CV ·2026-11-17 ·Matheus Pinheiro Ferreira, Thiago Carmuega Rabaçal, Giulio Brossi Santoro, Giovanna de Andrade Ferreira, Rodrigo Costa Pinto +13 more
Abstract · 56 pages, 13 figures, 8 tables, 2 supplementary tables. Submitted to Scientific Data. Maps and code openly available via Zenodo and GitHub

Global canopy height maps are increasingly used to estimate forest carbon, yet their regional accuracy is rarely tested. We present wall-to-wall maps of canopy height (10~m) and aboveground carbon density (30~m) for the Brazilian Atlantic Forest in 2024. Height was predicted by a deep learning network fusing Sentinel-2 reflectance, Sentinel-1 and ALOS PALSAR backscatter and ALOS elevation, trained on 7{,}179 and tested on 1{,}793 patches (58{,}800~km$^2$) from a statewide airborne LiDAR campaign over São Paulo state, Brazil. On held-out test patches, the root mean square error is 3.05~m at 10~m and 2.58~m at 30~m, and the model was validated at two LiDAR sites outside São Paulo. Over the same pixels, the Meta and World Resources Institute global product has more than twice this error. A power law fitted to 124 inventory plots converted canopy height to above ground carbon density. The remaining forest holds 1{,}944~Tg~C over 28.3~Mha. Global products therefore need local validation before use in carbon accounting.

0.36

Disagree to Prune: Reallocating Visual Tokens Where a Small VLM Dissents

cs.CV ·2026-11-17 ·Yanchao Li, Jiaqing Xie, Ben Gao, Yusu Zhang, Beilun Wang +5 more
code/data
Abstract · 45 pages, 8 in the main text. Code and data: https://github.com/LiZaiyuan0619/PRUDENT

Visual token pruning speeds up large vision-language models (VLMs) by keeping only a part of their visual tokens. Most existing methods improve the ranking of the tokens and then answer every question once with the same budget. With strong rankings, pruning hurts only a few questions, and signals computed on the kept tokens see little of what it removed. We find that, on most of these questions, a small VLM given the full image disagrees with the pruned answer, since it still sees the missing evidence. We propose PRUDENT, which first answers with about half the budget and spends the saved tokens on re-answering the disagreeing questions at twice the budget. We show when such re-answering pays and why the small VLM should read the full image. On 18 benchmarks, PRUDENT outperforms eight existing pruning methods at the same average budget of visual tokens. Its lead widens as the budget tightens, and holds across five InternVL sizes. PRUDENT also has the highest average in two more model families. The disagreement of the small VLM flags most pruning failures of all eight methods, and re-answering on it also improves the two strongest with their own rankings. Other checkers give similar gains, and the extra tokens go where the first pass hurts the answer. These results suggest that identifying which pruned answers are trustworthy provides an additional lever for visual token pruning beyond the ranking of tokens. Code and data are available at https://github.com/LiZaiyuan0619/PRUDENT.

0.36

Tracking Ethological Behaviors of Broiler Chickens Through Computer Vision (YoloV11) Using an Android Mobile Device

cs.CV ·2026-11-17 ·Ryan Davi Oliveira de Meneses, Glauber da Rocha Balthazar
published
Abstract

This study developed and evaluated a computer vision model for automated ethological behavior classification of broiler chickens in climate-controlled poultry houses. Using publicly available videos preprocessed and manually annotated into three classes (sitting, standing, and feeding), the YOLOv11 model was trained and deployed on an Android application via TFLite conversion. Exploratory analysis ensured dataset consistency and visual quality, while evaluation showed class-dependent performance: higher robustness for static postures (sitting, AP=0.531; precision=100%) and lower accuracy for dynamic behaviors (feeding, AP=0.292; recall=0.20). Tests with unseen images demonstrated significant improvements, achieving accuracies of 98.04% for sitting, 83.33% for standing, and 67.65% for feeding. The model showed effective generalization in real farm environments, supporting its potential for continuous behavioral monitoring, early welfare assessments, and decision-making in precision poultry farming. These findings establish a technical foundation for future mobile-based applications at commercial scale, reinforcing the role of computer vision in automated animal welfare evaluation.

0.36

LeViT-UNet++: Towards Balanced Fine-Grained Details and Context for Efficient Medical Image Segmentation

cs.CV cs.AIcs.MM ·2026-11-17 ·Guoping Xu, Jinling Zhu, Liuyue Li
code/data
Abstract · 24 pages, 5 figure

Efficient medical image segmentation is essential for clinical applications requiring fast and accurate delineation, such as surgical navigation, radiation therapy, and real-time image-guided intervention. Despite substantial advances in convolutional neural networks and Transformers, attaining high accuracy under limited computational resources remains difficult, because reducing model complexity often sacrifices fine-grained structural details and global context modeling, yielding an unfavorable accuracy-efficiency trade-off. To address this, we propose LeViT-UNet++, an efficient encoder-decoder architecture that better balances local detail preservation and global context modeling. Specifically, we introduce a Haar Wavelet Multi-Scale Downsampling (HWMD) module that decomposes feature representations into low- and high-frequency components to extract multi-scale structural information, thereby mitigating detail loss during downsampling. We further develop a depthwise convolution-based decoder that integrates multi-scale global representations from Transformer blocks with local spatial features, enabling progressive feature refinement while reducing computational overhead. Evaluated on four benchmarks (Synapse, ACDC, LA2018, and MSD Heart) under both 2D and 3D settings, LeViT-UNet++ achieves competitive accuracy with a compact parameter footprint and fast inference. On Synapse, for instance, it attains a DSC of 76.57% with only 8.025M parameters and 241.72 FPS, confirming a favorable accuracy-efficiency balance. These results demonstrate the effectiveness of combining wavelet-based multi-scale feature extraction with lightweight local-global feature fusion for efficient medical segmentation. The source code is available at https://github.com/apple1986/levit_unet_pp.

0.36

Déjà View: A Shared Working-Memory Model for Visual Tasks

cs.CV cs.LGq-bio.NC ·2026-11-17 ·Mranmay Shetty, Gabriel Kreiman
hot direction
Abstract · 6 figures; includes appendices

A working-memory model must remember stimuli and use its memory flexibly to perform different tasks. Here we introduce a model of working memory for images that combines dynamic activity with temporary associations between image features. We train the model on five tasks, including remembering a sample while other images are presented, categorizing distractor images while remembering the sample, and determining whether an image is new or has been presented before. The same trained model performs all tasks, guided by explicit cues about which response is required. We use this flexible memory model to ask how delay, distraction, and image history affect performance. Delays have little effect on sample-memory performance for the delays we tested, while seeing more images leads to more false alarms even when the target image appeared equally recently in all cases. On the sample-memory-plus-categorization task, the probability of correct responses on all questions decreases as more questions are asked, despite robust sample-memory performance. Finally, the model maintains stable performance across 40 brief trials of each task under the tested conditions. Together, these results demonstrate how a single working-memory system can be flexibly used to perform different tasks, and how its errors depend on both previous images and task demands.

0.35

SpikWM: Efficient End-to-End Spiking World Modeling for Visual Planning

cs.CV ·2026-11-17 ·Shuhan Ye, Hongbin Yu, Chong Wang, Hui Lu, Jiaming He +3 more
code/data
Abstract

Goal-conditioned planning with image observations and goals requires predictions that distinguish visual states and action consequences. Spiking neural networks offer sparse computation, but single-step spikes can obscure motion cues and merge actions with different physical outcomes. We propose the Spiking Latent World Model (SpikWM), an end-to-end joint-embedding predictive architecture (JEPA) with one simulation step per image and one neural update per latent position. Spiking Regional Encoding retains spatial cues through regional pooling. Bi-Spike Action Encoding combines positive and negative binary responses, while Causal Membrane Prediction carries subthreshold evidence across the latent history with learned group-wise time constants. The encoder and predictor are trained jointly on reward-free image--action trajectories using next-embedding prediction and SIGReg. With a common planner, SpikWM achieves mean success rates of $88.67\%$, $97.33\%$, $99.33\%$, and $86.67\%$ on Push-T, Two-Room, Reacher, and OGBench-Cube, respectively, outperforming two spiking baselines. Push-T ablations support each module, and controlled replays examine responses to appearance and physical state changes. Under a matched planning budget, SpikWM reduces estimated arithmetic energy by $75.6\%$ relative to LeWM. Code will be released at https://github.com/ShawnYE618/SpikWM.

0.35

SpatialMind: Adaptive Multi-Level Geometric Fusion for Spatial Reasoning in Multimodal Large Language Models

cs.CV cs.CG ·2026-11-17 ·Shibin Zhao, Hao Tan, Fan Yang, Haixin Wang, Feilong Chen +4 more
code/data
Abstract · Includes supplementary material. Code available at https://github.com/zhaosb-star/SpatialMind

Multimodal Large Language Models (MLLMs) demonstrate promising capabilities in 2D visual understanding, yet remain limited in 3D spatial reasoning. Recent studies explore integrating 3D reconstruction models into MLLMs. However, these approaches either use a single geometric granularity across all tasks or adopt a fixed multi-level fusion scheme, regardless of task-specific reasoning requirements. Through systematic experiments on multi-level geometric fusion, we find that spatial tasks at different cognitive levels (e.g., distance estimation vs. route planning) require geometric representations of different granularities, revealing that fixed fusion cannot accommodate diverse spatial reasoning requirements. To bridge this gap, we propose SpatialMind, a task-adaptive framework for dynamically integrating multi-level geometric representations. SpatialMind employs GeoFus, a multi-layer geometric fusion module built around a gating network jointly conditioned on textual instructions, visual semantics, and geometric representations. This network adaptively modulates the contributions of shallow, intermediate, and deep geometric features across the layers of the LLM decoder. Furthermore, we introduce a negative-entropy regularization loss to mitigate gate collapse. Gate visualizations indicate that GeoFus adapts geometric granularity to task requirements. Experiments demonstrate that SpatialMind achieves competitive performance across spatial benchmarks, attaining a state-of-the-art score of 72.3 on VSI-Bench. The source code is publicly available at https://github.com/zhaosb-star/SpatialMind.

0.35

OVET-Consensus: Rank-Preserving Counterfactual Event Verification for Video Moment Grounding

cs.CV ·2026-11-17 ·Alibay Osmanli, Shaogang Gong
Abstract · 22 pages, 2 figures, 17 tables

Modern video moment retrieval systems often propose windows that overlap the target event, yet still select the wrong one. We study this scorer bottleneck on QVHighlights with a fixed QD-DETR-derived temporal expert: the parent-aware temporal expert reaches 64.71 R1@0.5, while diagnostic oracles over the same candidate family reach 75.68 R1@0.5 within the parent-reranked pool and 94.97 R1@0.5 over the full 100-candidate evidence pool. We introduce OVET-Consensus, a rank-preserving counterfactual event verification layer that keeps the temporal expert's ranked list as the default and permits only high-confidence rank-1 overrides from a policy calibrated on held-out training videos. The verifier uses parent-aware interval evidence, structured Qwen3-VL event judgments, VideoLLaMA3 pairwise temporal signals, and dense temporal evidence, but never lets those signals freely reorder the top-K list. On QVHighlights validation, OVET-Consensus improves the parent-aware temporal expert from 64.71 to 65.35 R1@0.5 with a positive paired-bootstrap interval, while mAP is largely unchanged, from 42.67 to 42.74; it makes 20 overrides, with 15 helpful corrections and 3 harmful corrections. The results point to a practical constraint on verifier design: semantic compatibility is useful evidence, but temporal grounding still requires deciding where the event starts, changes, and ends inside a visually compatible scene.

0.35

Fiberpose: state-of-the-art fiber morphometry and instance segmentation with a distance-transform neural network

cs.CV q-bio.QM ·2026-11-17 ·Viktor A. Balashov, Daniil S. Antonenko
unusual
Abstract

Analysis of fibrous structures in SEM images is a routine task in biomedical engineering and materials science. However, most available methods for fiber quantification still rely on classical image binarization followed by simple post-processing. These approaches often produce unreliable results in images with weak boundaries, heterogeneous contrast, and densely packed fibers. Here, we introduce Fiberpose, a framework that reformulates fiber analysis as direct regression of a smooth distance-transform (DT) field, together with a novel set of geometry-based algorithms specifically adapted to this representation. The predicted field directly encodes fiber geometry, while the accompanying algorithms enable instance reconstruction and measurement of diameter and orientation within a unified workflow. The framework is implemented as a standalone Python desktop application and a web tool https://fiberpose.org. On an independent dataset of fibrous SEM images, Fiberpose achieved state-of-the-art instance-segmentation accuracy in comparison with SAM2 and Cellpose-SAM, and more accurate diameter measurements than DiameterJ, Fibar and GIFT. The method remains effective in dense and weakly contrasted regions and shows qualitative generalization across microscopy modalities not represented in the training data. Fiberpose therefore provides an interpretable geometric representation and an automated pipeline for reproducible quantitative characterization of fibrous materials.

0.34

UnrealEvent: Synthetic Dataset Generation and Annotation for Event-Based Vision

cs.CV ·2026-11-17 ·Yentel Schoovaerts, Zijie Ning, Patrick Vandewalle
published
Abstract · Accepted at the Workshop on Neuromorphic Vision (NeVi), ECCV 2026

Event-based cameras provide high temporal resolution, low power consumption, and a high dynamic range, but progress in event-based object detection remains constrained by the scarcity of annotated training data. To solve this issue, this paper introduces the tool UnrealEvent, a synthetic dataset-generation pipeline implemented in Unreal Engine 5.5. The pipeline combines a V2E-inspired event-frame generator with EasySynth semantic-rendering passes, allowing rendered level sequences to be converted into paired event representations and YOLO-format bounding-box annotations without manual labeling. Events are generated from per-pixel log-luminance differences between consecutive frames, with contrast thresholds sampled independently from a normal distribution to emulate the fabrication-induced threshold mismatch of real sensors. Using this pipeline, we provide a synthetic dataset and evaluate YOLOv8 detectors under synthetic, zero-shot real, and real fine-tuning training on eTraM. The results show strong in-domain synthetic performance but a substantial synthetic-to-real gap, especially for pedestrians. Fine-tuning on a small real subset improves real-domain performance, while the remaining gap highlights the importance of asset diversity and sensor calibration.

0.34

SRED: Sparse Regional Evidence Distillation Across Magnifications in Whole-Slide Imaging

cs.CV ·2026-11-17 ·Zhiyuan Yang, Jiahao Cheng, Vincent Quoc-Huy Trinh, Mahdi S. Hosseini
published
Abstract · CVPR 2027 submission

In computational pathology, high-magnification whole-slide image analysis preserves tissue detail but requires encoding many patches, whereas lower magnification reduces computation by covering a higher field of view per patch while sacrificing detail. Existing cross-magnification distillation matches features that often encode detail unreachable by low-magnification input, sometimes requiring slide-level labels that limit scalability. To address those issues, we introduce Sparse Regional Evidence Distillation (SRED), which transfers regional event occurrences that encode reachable high-level interactions between details without slide-level labels. We use a sparse autoencoder (SAE) to learn an event dictionary from 20x teacher features. Logical OR pooling records whether each event is detected in any spatially aligned patch, retaining both localized and shared evidence within the corresponding 5x region. A pretrained 5x encoder is fine-tuned to predict these regional targets without slide-level labels. Visual analysis identifies dictionary components responsive to recognizable histological patterns, supporting target interpretation. Compared with native 5x features, SRED improves average macro-F1 by 1.1 points across seven classification and grading cohorts and average C-index by 1.7 points across five survival cohorts. Runtime matches native 5x, with an average 11.58x speedup over 20x.

0.34

Code World Model 2.0: Towards Infinite Controllable Video Generation

cs.CV ·2026-11-17 ·Ruoyu Wang, Hao Zhong, Yanming Yang, Jiahao Tian, Lin Liu +6 more
code/data
Abstract · Project Page: https://www.inclusion-ai.org/CodeWorldModel/

Visual proxies decouple explicit world-state simulation from photorealistic video synthesis, offering a controllable interface for both offline clip creation and real-time interactive streaming. Yet a practical renderer must address two challenges. Upstream proxies vary from coarse layouts to fine articulated geometry depending on the interaction, and streaming inference must operate in a few causal steps without accumulating drift across self-generated history. We present Invoker-Render, pairing a bidirectional creation model with a causal four-step streaming model under a shared conditioning interface over structural proxies, an appearance reference, and language. Trained on an aligned approximately 200-hour multi-granularity corpus, Invoker-Render honors fine geometric constraints when supplied while synthesizing plausible local motion under coarse proxies, enabling runtime selection of input detail. To bridge bidirectional creation and real-time streaming without compounding step compression and rollout drift, we introduce Progressive Forcing (ProForcing). ProForcing first trains a causal predictor and distills it into a four-step renderer using histories derived from training videos. Training then shifts to continuous self-generated rollouts. A pipelined serving stack supports efficient streaming inference. Experiments show versatile control across proxy families, granularities, and camera paths, consistent gains over recent streaming baselines, and stable 30-minute racing rollouts that preserve scene and subject appearance.

0.34

LoopZip: Visual Token Compression for Efficient Looped Vision-Language Models

cs.CV cs.LG ·2026-11-17 ·Weizhong Huang, Zhicheng Liu, Yan Zhang, Fei Chao, Xiawu Zheng +1 more
code/data
Abstract · 11 pages, 10 figures

Looped vision-language models (VLMs) deepen computation by reusing shared layers, approaching larger models' capabilities with fewer parameters and easing the storage and deployment burden of stacking independent layers. However, repeatedly processing long visual sequences still incurs substantial compute and GPU memory costs. Visual token compression reduces these costs by shortening sequences. Existing methods mainly select original tokens directly or combine selection with aggregation. In looped models, the former may overlook changing token importance across calls and discard information needed later; the latter lack joint consideration of output changes across calls, limiting information preservation. We propose LoopZip, a training-free visual token compression algorithm addressing both issues. Visual aggregation builds a compact candidate pool before the looped backbone, and cross-invocation selection averages instruction attention from the first two existing calls of the same shared attention layer to retain tokens needed by both states. Weight-tied value correction then jointly fits shared visual-value adjustments to both pre-selection layer outputs, reducing output deviations across states. To validate LoopZip's effectiveness, we conduct extensive experiments covering two open-source looped VLMs, eleven benchmarks, and three pruning ratios. On LoopVL-1B, LoopZip retains 97.16% of the original model's performance at 50% pruning ratio and exceeds the strongest baseline by 5.36% at 87.5% pruning ratio. Across Hugging Face Transformers, vLLM, and SGLang, LoopVL-1B with LoopZip reduces peak allocated GPU memory by 58.36%-58.98% and achieves 2.01-2.47x end-to-end speedups over the original model.

0.33

Impact of Pretraining Length, Model Size, and Annotation Budget on Downstream Clinical Tasks of Foundation Models in Radiotherapy

cs.CV ·2026-11-17 ·Alejandro Cortina Uribe, Sidsel Højklint Poulsen, Ivan Richter Vogelius, Jens Petersen
code/data
Abstract · 9 pages, 5 figures, 2 tables. Supplementary Material included

Cone-beam CT (CBCT) scans acquired during radiotherapy capture changes such as tumor shrinkage and radiation-induced tissue changes, offering great potential for foundation models that help monitor individual treatment responses. However, to our knowledge, no large-scale pretrained models for radiotherapy CBCT exist, and evaluating modeling choices is difficult since downstream tasks are not known in advance. We examine how pretraining length, model size, and finetuning strategy relate to downstream clinical performance, and how metrics tracked during pretraining correlate with it. We pretrain vision transformer encoders with masked image modeling on about 150,000 raw 3D radiotherapy CBCT scans, tracking representation quality via validation loss, effective rank, and linear probing. Downstream tasks are classification of feeding tube presence ($n=51,135$ CBCTs, $1,562$ patients) and clinical characteristics (T-, N-stage, sex, age, tumor site; $n=1,278$ patients) in head and neck cancer. We compare low-rank adaptation (LoRA) and full finetuning against no pretraining, varying annotated dataset size. Pretext validation loss was the most consistent predictor of downstream performance; effective rank and linear probing were also associated, but not consistently across model sizes and finetuning methods. LoRA performed comparably to full finetuning while training only 0.5-2.5% of the parameters. Fully finetuned models consistently outperformed models without pretraining across all sizes, with AUROC 0.91-0.97 vs. 0.58-0.85 for feeding tube classification and 0.99-1.00 vs. 0.71-0.97 for sex classification. Model size had only a small, inconsistent effect, suggesting smaller models suffice at lower computational cost. Code and pretrained models (non-commercial license): https://github.com/alexCortinaU/RT-CBCT_SSL

0.33

Fast Training-Free Low-Light Video Enhancement with a Frozen Image Diffusion Prior

cs.CV ·2026-11-17 ·Hon Ming Yam, Chun Pong Lau
Abstract · 15 pages, 12 tables, 4 figures

Pretrained image diffusion models can enhance low-light video without low-light training data and without retraining for a new camera. ZVRD, the only published training-free method for low-light video, is too slow to use: our implementation needs over a minute per $256 \times 256$ frame. We present a causal, training-free pipeline that adds two temporal components to a frozen MR. Illuminate image prior. Temporal input fusion recursively averages motion-compensated low-light frames before diffusion inversion. Late-start offset propagation runs the full sampling trajectory only on key frames. Each non-key frame inverts the low-noise part of its own trajectory and adds an enhancement offset warped from the latest key frame. At ZVRD's own resolution the pipeline is $261\times$ faster than ZVRD and scores better on all five metrics we report. On 13 held-out SDSD clips at $1280 \times 720$, it improves on framewise eight-step processing by 0.55 dB in PSNR and 0.034 in LPIPS and halves the warping error, in 10% less time. Measurements inside the model show why, and a linear-Gaussian model of inversion explains them. Fusion removes frame-to-frame noise before inversion amplifies it. A flow error perturbs the transported offset ten times less than it perturbs a transported state. The perceptual gain does not transfer to low resolution: at $346 \times 260$, LPIPS is worse by 0.028. At 0.65 frames per second, the pipeline is an offline method.

0.33

$π^4$: A Foundation Model for 4D Scene Geometry and Motion

cs.CV ·2026-11-17 ·Yifan Wang, Weiguang Zhao, Ni Yang, Haoyu Guo, Junting Dong +2 more
code/data
Abstract

$π^4$ is a unified foundation model for 4D scene perception. From diverse, uncalibrated RGB observations in the wild---a single image, an unordered collection, or a video sequence---it recovers dense geometry, camera poses, and metric scale, together with 2D/3D point tracks and moving-object masks for multi-image inputs. It couples a geometry branch with a geometry-conditioned motion branch, which predicts image-space correspondences that are lifted into 3D using reconstructed target-view geometry. Across diverse benchmarks, $π^4$ achieves the best overall average rank in geometry evaluation, alongside competitive performance in single- and multiview 2D/3D point tracking and moving-object segmentation. With a shared backbone, the full geometry-and-motion system contains approximately 0.5B parameters---less than half the size of most baselines. It produces joint reconstructions and tracks in one forward pass, including large-scale scenes with thousands of input images despite training with contexts of at most 64 views. We further present the data and training recipe underlying these capabilities, together with empirical studies of key design choices. Code and models are available at https://github.com/yyfz/Pi4.

0.33

Social Signal Bench: Evidence-Weighted Evaluation of Co-occurring Social Signals in Video and Audio

cs.CV cs.LG ·2026-11-17 ·Siddharth Ravi, Anders R. Bargum, Hugo Anselme, Patrizio Cugia di Sant'Orsola, Alberto Cardaci +3 more
benchmark
Abstract · 25 pages, including appendix

Recognizing social signals such as interest and hesitation requires interpreting context-dependent behaviors that may co-occur in conversation. Existing affective computing benchmarks rarely evaluate such co-occurring signals, and conventional multi-label F1 overlooks semantic relationships and annotator disagreement. We introduce Social Signal Bench, a 47-hour densely annotated benchmark for the evaluation of social signal recognition and engagement, comprising two datasets: Social Signal Bench Video (25 hours), covering ten social signals in audiovisual recordings, and Social Signal Bench Audio (22 hours), covering six signals in audio-only recordings. Each dataset pairs a public training set of online conversational video, of which 10 hours are expert-annotated and the remainder pseudo-labeled, with a private expert-labeled test set spanning four conversational domains. We also propose Evidence-Weighted F1 (EW-F1), which weights reference labels by annotator evidence and assigns graded credit to semantically related predictions through optimal per-clip matching. Controlled perturbations demonstrate that EW-F1 distinguishes semantically related errors from unrelated ones, unlike conventional F1. Under a shared evaluation protocol, 7B open models post-trained on Social Signal Bench outperform general-purpose frontier models on video and match them on audio.

0.33

Say Less, Hallucinate Less? Diagnosing and Restoring Object Coverage in Preference-Optimized Vision-Language Models

cs.CV cs.AI ·2026-11-17 ·Hansang Lee
Abstract · 18 pages, 6 figures, 19 tables. Main text and appendices

Preference optimization (PO) is the standard way to reduce object hallucination in large vision-language models. We find that four public PO checkpoints of LLaVA-1.5-7B also lower object coverage. They name fewer object categories, and their lower hallucination rates bring no significant gain in set-level object F1 on CHAIR. We therefore ask how much of the PO gain the base model can reproduce by stopping earlier. A positive end-of-sequence (EOS) logit bias lets the base model reach the recall of each PO model, and this early stopping alone reproduces 58 to 86 percent of the CHAIRi gain of three checkpoints on two disjoint COCO sets. Writing longer does not bring the lost categories back. Under greedy decoding, a negative EOS bias recovers only 2 to 9 percent of the lost CHAIR recall of these checkpoints, and 94 to 96 percent of the added object mentions repeat categories already named. On CHAIR, the base caption of the same image still names about half of the missed object categories. We therefore propose base-guided object-category restoration, which appends high-confidence base-only categories to the PO caption without training. With thresholds fixed on a development set, it brings three checkpoints within one point of the base recall on a held-out COCO set with a lower distinct-object hallucination rate than the base model. Against the recall-matched base stopping curve, its gain holds for RLAIF-V near the base recall and for RLAIF-V and CHAIR-DPO at an intermediate recall. We will release the code of the recall-controlled evaluation.

0.33

ExplicitIR: Explicit Spatial Policy Fields for All-in-One Image Restoration

cs.CV ·2026-11-17 ·Dezhi Zhang, Shangrong Yang, Yudong Liang, Wenjian Wang
Abstract · 13 pages, 10 figures, 4 tables

All-in-one image restoration aims to recover clean images blindly from inputs with various degradation types and levels using a single model. However, the adaptation policy in most existing methods remains implicitly encoded in feature representations, leaving the restoration decision neither explicitly represented nor independently controllable. Since real-world degradations vary spatially, applying uniform updates across the entire image leads to over-smoothing of clean areas and under-correction of localized degradations. To address these limitations, we present ExplicitIR, which represents the restoration policy as two explicit per-pixel fields: Budget, which determines where and how strongly to update, and Style, which selects between narrow and wide context windows. Specifically, ExplicitIR incorporates a Spatial Policy Estimator (SPE) to predict these fields, explicitly representing the restoration policy and capturing spatially-varying degradations, which are then applied by Policy-Adaptive Decoder Blocks (PADB) through adaptive fusion of narrow- and wide-window responses, enabling independent control and precise local modulation. This design enables spatially-adaptive restoration in a single forward pass, without requiring task labels, prompts, or test-time degradation estimators. Extensive experiments on three-task and five-task benchmarks demonstrate that ExplicitIR achieves performance competitive with the strongest all-in-one baselines.

0.33

Check Your Blind Spots: Decision-Guided Perception by Probing the Planner with Hypothetical Objects

cs.CV cs.AI ·2026-11-17 ·Hoang Pham Cong, Long D. Nguyen, Ho Viet Duc Luong, Nguyen Tran Minh Nhat, Binh P. Nguyen
theory
Abstract

Not every perception error changes the planner's action. Yet adaptive-resolution perception chooses where to spend high-resolution computation from image cues, which can neither reliably point to an object that low resolution missed nor tell which objects matter to the planner. We observe that the planner can say where an object would matter before anything is seen: inserting a hypothetical object at a location tests whether its action would change. Decision-Guided Speculative Perception (DGSP) treats the low-resolution scene and its action as a draft and probes the planner in this way to obtain a decision-sensitivity field, the locations where an unseen object would change the action. Only this field and the support, the detections on which the current action rests, are verified at high resolution, repeatedly until the action is stable. For two planner classes, we prove that DGSP returns exactly the action of whole-frame high-resolution perception if crop detections match whole-frame ones and the crops cover the field, conditions that real images meet approximately. On three driving datasets, it reproduces this action on about 98% of frames with about a quarter of the pixels. At the pixel cost of uniform downscaling, it repairs more of the draft's disagreements with this action, brakes falsely less often, and recovers many more of the decision-relevant objects that the draft misses.

0.32

Spool: Low-Latency Streaming Inference for Hybrid Vision-Language Models

cs.CV cs.LG ·2026-11-17 ·Rui Ning, Om Chabra, Hari Balakrishnan
benchmarkcode/data
Abstract

Applications such as visual accessibility agents and robotics require answering questions about recent video content promptly, over a bounded window of past frames (e.g., the last two minutes). Hybrid vision-language models (VLMs), which interleave attention with linear recurrent layers such as Gated DeltaNet (GDN), are attractive for this setting, but maintaining a sliding window is expensive. The reason is that recurrent layers combine all past frames into a shared state, so expiring old frames normally requires re-running the entire window through the model. We observe that the recurrent update is affine in the state with input-dependent coefficients, so the effect of each frame can be captured as a transformation that composes associatively. We present Spool, a training-free framework that caches per-frame state transformations and composes them efficiently as the window advances, removing expired frames from each recurrent layer's state without replaying frames through the backbone. Removal is exact within each layer; as with key-value eviction in attention, expired frames can retain residual influence through the cached inputs computed by earlier layers. We design a dedicated data structure and GPU kernels to achieve low caching and composition overhead. Across five open-source hybrid VLMs and four video reasoning benchmarks, Spool achieves the lowest p95 time-to-first-token in all 20 model-benchmark tests, reducing it by a geometric mean of $6.6\times$ over the faster of two recent vLLM baselines, while improving accuracy by 2.3 percentage points on average.

0.32

M$^3$-Test: Coverage Gaps in Joint-Perturbation Testing of Visual Question Answering

cs.CV ·2026-11-17 ·Enhui Wan, Xiaowan Wang
Abstract

A visual question answering (VQA) model can pass every image-only and question-only test yet fail when the same changes occur together. Standard robustness reports average over conditions and therefore cannot say which accepted requests remain vulnerable. We introduce M3-Test, a coverage audit that aligns each clean input with its single-modality variants and a complete 5x5 family, 4x4 strength joint library, so that any reduced test suite can be replayed on the same requests. Across 2.6 million responses from four vision-language models, joint failures affect 26-29% of the Qwen questions that pass all 40 single-modality tests, although fewer than 0.6% of their joint records fail. Passing the 25 maximum-strength corners does not close this gap: 11-14% of the accepted questions still fail elsewhere in the library. Re-executing the archived failures reproduces about 90% of them, against a 0.3-0.6% flip rate among matched passing controls, and re-checking the most fragile failure records supports 80-89% of them. On the evidence examined, neither execution noise nor grading error can fully account for the gap. The same aligned grids make previous evaluations reusable. On held-out images, a risk order learned from other models discovers about 80% of affected Qwen questions within 25 additional queries, compared with 65-68% for maximum-first and 53-57% for pairwise covering arrays. M3-Test thus turns a passing robustness suite into a request-level account of what remains untested, supporting acceptance, regression review and test allocation.

0.32

Learning Agent-Centric Vision-Language-Action World Models for Autonomous Driving

cs.CV cs.AI ·2026-11-17 ·Guoqing Wang, Pin Tang, Xiangxuan Ren, Chao Ma
top venueposition
Abstract · Accepted by NeurIPS 2026

Recent progress in integrating vision-language-action (VLA) models with world modeling has significantly advanced end-to-end autonomous driving by allowing systems to reason about imagined futures instead of merely reacting to current observations. However, most world VLA models generate future frames at the pixel level with plenty of dense visual tokens. Although such generated futures can be visually plausible, they preserve many driving-irrelevant appearance details, i.e., only a small portion of the tokens directly encode critical planning-relevant factors, such as agent geometry, motion dynamics, and interactions. This introduces a significant performance bottleneck. An important question thus arises: what world state should a world VLA model reason over to support planning beyond pixel-level prediction? We argue that planning-oriented world modeling should focus on a compact and structured agent-centric latent state that captures the interactions between objects in the driving scene over time. As such, we introduce AgentWorld, an agent-centric VLA world model that uses discrete agent tokens jointly for reasoning and trajectory planning. Each token is dynamically grounded to a surrounding agent associated with its object's state, such as semantic identity, 3D geometry, motion, and interaction context. AgentWorld learns these tokens through three key processes: explicit geometric grounding, reasoning-oriented supervised fine-tuning, and agent-oriented reinforcement learning. Extensive open-loop and closed-loop experiments under diverse driving scenarios show that AgentWorld enhances planning capabilities, fully exploits visual evidence and agent dynamics, and yields interpretable agent-centric world states.

0.31

TextSphere: Efficient Text-to-Image Generation via Shared Hyperspherical Alignment and Uniformity

cs.CV cs.CL ·2026-11-17 ·Yixia Zhang, Girish A. Koushik, Xilu Wang, Wenwu Wang
Abstract · 21 pages, 8 figures

Recent advances in text-to-image generation have substantially improved image quality, particularly through diffusion-based methods that require multiple inference steps. In contrast, Sphere Encoder offers an efficient generative framework capable of producing images with fewer forward steps. However, extending it to free-form text is nontrivial, because its class-conditioned design lacks cross-modal alignment. To address this challenge, we propose TextSphere, an efficient text-to-image generative framework that extends Sphere Encoder with language conditioning and cross-modal alignment. We incorporate pretrained CLIP text embeddings through adaptive layer normalization (AdaLN) to guide image generation while retaining the original spherical sampling mechanism. To bridge the cross-modal representation gap, we introduce a Spherical Semantic Alignment and Uniformity (SSAU) objective that jointly promotes semantic alignment and representation uniformity on the hypersphere. Experiments on CUB-200-2011, MS-COCO-2014, and Flickr30k demonstrate that TextSphere consistently reduces generation FID (gFID) over the text-conditioned baseline while achieving reconstruction FID (rFID) below 1.0. With only 4 and 6 sampling steps, TextSphere achieves lower gFID than the compared diffusion- and GAN-based models trained from scratch on CUB-200-2011 and MS-COCO-2014, respectively. We further observe that improved reconstruction fidelity does not necessarily yield better generation quality. In addition, we use Partial Least Squares Singular Value Decomposition (PLS-SVD) to analyze cross-modal alignment and modality gap, offering insight into the effectiveness of our approach.

0.31

When a Larger Guidance Scale Repairs Text-to-Video Failures and When It Does Not

cs.CV ·2026-11-17 ·Taewon Kang, Matthias Zwicker
unusual
Abstract · 29 pages, 15 figures

When a text-to-video diffusion model ignores part of a prompt, the usual first fix is to raise the classifier-free guidance scale. This paper is a measurement study of when that fix works. On Mochi, the same increase repairs failures on rare compositions ("a snowy beach") beyond what generating again does, does not repair negations ("a highway with no vehicles"), where it fixes fewer failures than generating again does, and changes nothing for events that must happen late in a clip ("a sandcastle that a wave washes flat"). Which of these happens can be known in advance, from the prompt and one inexpensive score. The score, which we call demand, measures how strongly the denoiser reacts to the part of the prompt that the failed video leaves out, and it is read from the first quarter of denoising, before any video exists. Our main finding is a condition: raising the scale is effective only when the prompt contains no negation and no event tied to a specific time, and when demand is sufficiently high, with recovery clearly higher above a fitted threshold than below it. Published corrections for the three failure types, used as fixed probes, behave as this account predicts. A short decision rule follows: before generating again, a user can tell whether the most common fix is worth trying, and what to use when it is not.

0.31

CueClash: Frontier Vision-Language Models Judge 3-D Relations Close to the Picture Plane

cs.CV cs.AIcs.RO ·2026-11-17 ·Soham Dan
hot direction
Abstract

A vision-language model (VLM) can judge 3-D relations from the picture or from the ground layout. In photographs the two usually agree. We separate them with CueClash, rendered scenes in which the 2-D cue for each of five relations is right on exactly half the statements. The shortcut gap is accuracy when the cue is right minus accuracy when it is wrong. For Claude Opus 5.5, GPT-5.5 and Gemini 3.6 Flash at low reasoning effort it is +18 to +37 percentage points. It is +7 to +29 without two relations whose cue-wrong items are confounded with borderline cases or occlusion. Given ground coordinates and each relation's definition, they answer 99 to 100% correctly. Pixel measurements with camera depths remove most of the gap for objects fitting through openings. The failure lies in the step from the picture to the layout. The maps that the models draw of the scene from above mostly keep the depth order but, like the picture, shorten depth. GPT-5.5's maps stay close to a copy of the picture. Opus 5.5's and Gemini's maps undo about 0.6 of this shortening. Gemini's then shrink all depths by a further uniform factor. Opus 5.5 follows the picture mainly on fits, and its maps correct depth about as much as human observers of real scenes do. The three models also lose accuracy on CV-Bench photographs where a reliable 2-D cue misleads. In two open models the layout information reaches the language model, but their maps lose much of it.

0.31

FrameSeg3D: Learned Local Frames for Segmenting Small Parts on High-Resolution 3D Scans

cs.CV cs.GR ·2026-11-17 ·Xinwen Zhang, Tongyan Wang, Huawei Lin
Abstract · 31 pages, 6 figures, 16 tables

Segmenting small parts on high-resolution 3D scans, such as nails on hand scans and teeth on intraoral scans, poses three challenges: 1) each part covers a tiny fraction of the scan, 2) its boundary is only a subtle change of height and curvature, and 3) it faces its own direction. Existing methods fall into two groups, and both fall short. The first group processes the whole scan at once and sees each part too coarsely. Point and mesh networks must shrink the scan to fit in memory, so a small part keeps only a few of its points. Multi-view methods render images of the whole object, so each part fills only a few pixels and a tilted part is seen from the side. The second group first finds each part and then segments it up close. But these methods find the parts with knowledge about one kind of part, for example the shape of the dental arch for teeth. Such knowledge does not carry over to other kinds of parts. We find that accuracy is limited by how a part is seen, from which direction and at what scale, not by how many points land on it. We therefore present FrameSeg3D. It first predicts a local frame for every part, its center and the direction it faces, from a low-resolution view of the whole scan. It then segments the part in views around its frame normal at the scanner's resolution, so covering the whole scan no longer costs resolution. Because every part arrives centered, facing the camera, and at a fixed scale, one segmentation network can serve different kinds of parts and learn from few labeled masks. We test FrameSeg3D on two kinds of parts, nails and teeth. It outperforms state-of-the-art methods on hand scans (mean Dice 0.970 on nails) and on the Teeth3DS+ test fold (per-tooth IoU 96.5), with the most accurate tooth boundaries (boundary IoU 70.5).

0.31

Lost in Read-out: Locating the Image Quality Gap in Vision-Language Models with a Closed-Form Edit

cs.CV ·2026-11-17 ·Zhuangzi Li, Shilv Cai, Jian Jin, Zhiyi Pan, Haihong Xiao +1 more
hot direction
Abstract

Vision-language models (VLMs) are widely used as zero-shot image quality assessors, yet their good-vs-poor answers agree only moderately with human opinion. We ask whether these models fail to see quality or fail to say it. Layer-wise probes on five VLMs point to the second: quality is linearly decodable from the vision encoder and still from the last language layer, far better than the answer reveals, and about a hundred labels recover most of it. The loss lies mainly in the final read-out, which partly reads something else: relative to human opinion, it over-rewards colorfulness and under-rewards sharpness on all five models. We test this by intervention: if the read-out is at fault, a closed-form edit of a single last-layer neuron, fitted to many prompts at once, should close the gap and no more. Both hold. Under prompts never used for fitting, the edit recovers 86--100\% of what a ridge head on the last-layer state, fitted under the same prompts, finds beyond the answer on KonIQ. Where that gap is small, it gains little, too little for our transfer criterion on the two largest models; where the head itself is below the answer, it lowers the score; and it works only when written from the middle of the language model on. Unlike unconstrained fine-tuning, the edit leaves perception question answering intact. It does not carry over to yes-or-no or rating questions, and it reverses the good--poor contrast after sentiment-laden sentences. Code will be made available.

0.30

Sample-Aligned Decoders for Latent Generative Models

cs.CV cs.LG ·2026-11-17 ·Yu Lei, Yi Li, Hongcheng Gao, Yipeng Wei, Liang Li +2 more
code/data
Abstract · Project page: https://torpedo2003.github.io/SAD/

Latent generative models synthesize images by decoding generated latents, yet their decoder is trained only to reconstruct encoder latents and never sees the latents it decodes at test time, so the residual errors of the generator are rendered directly into the final image. Existing remedies retrain the generator, add cost at every sampling step, or make the decoder robust to perturbed real latents, assuming that generator errors resemble such perturbations. We find that they do not. Generated latents lie on the encoder latent manifold as closely as real ones, yet follow a different distribution. The decoder should therefore be adapted to the distribution of generated latents, which raises two challenges: (1) unpaired supervision, since a generated latent has no ground-truth image, and (2) latent-space drift, since a decoder freed from paired targets forgets the latent space it must decode. We propose Sample-Aligned Decoders (SAD), a post-training framework that fine-tunes only the decoder of a frozen generator at no inference cost. Generated-Sample Distribution Alignment matches decoded generated latents to unpaired real images in a frozen foundation-model feature space, and Encoder Anchor Preservation keeps the decoder close to the original one on encoder latents. Our analysis shows that the alignment is effective only on generated latents, where it corrects each sample in proportion to how much more often the generator produces such latents than the encoder; on encoder latents it can only work against the anchor. SAD improves four state-of-the-art ImageNet generators across three tokenizers, e.g., FID-50K from 1.15 to 0.98 for REPA-E, complements discriminator guidance, and extends to text-to-image generation. Our code and models are available at https://torpedo2003.github.io/SAD/.

0.30

TRIDENT: Efficient Triple-Task Learning of Dehazing, Depth, and Uncertainty Estimation for Underwater 3-D Robot Visual Perception

cs.CV cs.RO ·2026-11-17 ·Geonmo Yang, Younggun Cho
published
Abstract · 13 pages. (c) 2024 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any curren…

Underwater visual systems often suffer from blurry textures and low color contrast due to the inevitable light propagation. These issues can significantly degrade the perception of stable robotic operations. In this paper, we introduce a novel learning-based sensing system that tackles the multidimensional vision tasks in underwater; concretely, we deal with image enhancement, depth estimation, and uncertainty for 3-D visual systems. Also, we propose a TRIDENT model in a fast and lightweight manner; TRIDENT consists of three parallelized decoders and one backbone structure for efficient feature sharing. In addition, it is designed to be trained to express complex parameterization. In experimental evaluation on several standard datasets, we demonstrate that TRIDENT significantly outperforms other existing methods on image enhancement and depth estimation. Despite performing three tasks, our model has better efficiency than the others for both memory size and inference time. Finally, our joint learning approach demonstrates robustness in feature matching and seamlessly extends from 2-D to 3-D vision tasks. Supplements are available at https://sites.google.com/view/underwater-trident/home

0.30

FVScore: Specification-Driven Neuro-Symbolic Video Evaluation

cs.CV cs.AI ·2026-11-17 ·Do Xuan Long, Long T. Le, Yiwen Song, Min-Yen Kan, Tomas Pfister +1 more
code/data
Abstract · Project page: https://dxlong2000.github.io/FVSCORE/

We introduce Faithfulness Video Score (FVSCORE), a neuro-symbolic framework for specification-driven video generation evaluation. Existing evaluators rely on holistic MLLM judgments or vision-language metrics, yet aligning their scores with human judgments and tracing them to specific temporal violations remain open challenges. FVSCORE addresses these limitations by compiling each video into explicit propositions capturing prompt requirements and implicit propositions capturing unstated physical, temporal, and cross-modal constraints. It grounds each proposition using specialized perceptual tools and evaluates the resulting confidence traces with soft temporal operators inspired by Linear Temporal Logic. A Significance-Aware Calibration step then aligns verification with human perception by retaining only perceptually significant violations before producing a timestamp- and evidence-grounded report. We conduct extensive experiments across public and complex multi-event benchmarks with the strongest video models to date, demonstrating that FVSCORE outperforms holistic and agent-based evaluators in human correlation while providing highly accurate and interpretable feedback.

0.30

OmniTongue: A Comprehensive MLLM Benchmark for Automated Tongue Diagnosis in Traditional Chinese Medicine

cs.CV ·2026-11-17 ·Liuyi Xu, Yun Guo, Shusheng Li, Xuewen Zhang, Yining Qian +1 more
benchmarkcode/data
Abstract · 11 pages, 2 figures, 3 tables

Multimodal Large Language Models (MLLMs) have excelled in Western medical imaging, yet their potential in Traditional Chinese Medicine (TCM) tongue diagnosis remains systematically unevaluated. Tongue diagnosis requires sophisticated global and local perception, serving as a rigorous touchstone for assessing the visual understanding depth of MLLMs. However, existing datasets are limited by single dimensions, coarse granularity, and ambiguous labels, hindering the evaluation of authentic clinical capabilities. To bridge this gap, we introduce OmniTongue, the first comprehensive, fine-grained Visual Question Answering benchmark for TCM tongue diagnosis. This benchmark integrates 1,351 expert-validated images and 21,616 instruction pairs. Its core design features 16 clinical dimensions for multidimensional evaluation, region-specific annotations for pathological localization, and 66 fine-grained labels for precise diagnostic representation. Extensive evaluations of existing MLLMs reveal significant performance gaps, demonstrating that medical-specific models lack domain advantages, while universally poor Macro-F1 scores expose a reliance on statistical shortcuts rather than genuine understanding of tongue pathology. These findings not only highlight the current limitations of MLLMs in interpreting regional medical features but also underscore OmniTongue's significance for developing robust automated tongue diagnosis models. The code and dataset will be made publicly available in a future release.

0.30

ATS: Anti-Aliased Triangle Splatting with Exact Per-Primitive Integrals

cs.CV cs.GR ·2026-11-17 ·Brent Zoomers, Joni Vanherck, Leander Winters, Jeroen Put, Lode Jorissen +1 more
theory
Abstract · 17 pages, 7 figures, 9 tables

Triangle Splatting represents scenes using semi-transparent triangles, but it evaluates each triangle at the pixel center and culls any triangle with an inradius smaller than one pixel. The former causes severe aliasing, while the latter culls most of the scene at lower resolutions. We show that the integral of the Triangle Splatting window over a pixel admits an exact closed-form solution. The angle bisectors split the triangle into three sub-triangles, in each of which the window is a power of the distance to one edge. Each piece therefore has a closed-form antiderivative, and Green's theorem reduces the integral and its gradients to one term per side of the sub-triangle clipped against the pixel. The integral over the whole triangle, which we show has a closed form, then allows conservative culling, removing only triangles that cannot meet the renderer's alpha cutoff at any pixel. This exact formulation anti-aliases existing models without retraining, exhibits less temporal aliasing than 2x2 supersampling, and enables multi-scale training, matching or outperforming supersampled training at reduced resolutions.

0.30

Breaking Visual Conventions: Spectrally Coupled Weight Perturbation for Creative Image Generation

cs.CV ·2026-11-17 ·Fu Feng, Ruoheng Zhou, Yucheng Xie, Xu Yang, Jing Wang +1 more
Abstract

Text-to-image (T2I) models often produce high-quality images that follow familiar visual conventions, even in response to open-ended prompts such as ``a creative [CLS]''. Existing methods encourage novelty by exploring semantic conditions or modifying features and guidance during denoising. Pretrained weights offer a complementary source of creative variation, but unstructured perturbations can have little visual effect or degrade image quality. We introduce Coupled Spectral Perturbation (CosPert), a training-free method that targets attention projections in layers identified by a one-time sensitivity analysis. For each projection, CosPert uses a shared random transformation and its transpose to couple updates to the singular-vector factors while perturbing the spectral gains. A common relative budget calibrates the magnitude of weight changes across projections, and a sigmoid schedule concentrates their influence in early denoising and gradually attenuates them during refinement. CosPert provides plug-and-play creativity enhancement at inference time. Once the perturbations are prepared, sampling requires no auxiliary networks or additional model evaluations. On both FLUX.1 and Stable Diffusion 3.5, CosPert enhances visual creativity, producing more varied object structures while preserving category identity. Across 50 object categories, CosPert achieves the highest mean CREward among the evaluated methods while maintaining strong prompt alignment.

0.30

Propose & Verify: Fast Reflection Symmetry Detection via Harmonic Fields

cs.CV ·2026-11-17 ·Yishai Lavi, Shai Avidan
benchmarktheory
Abstract

Detecting multiple reflection symmetries in 3D shapes requires aggregating geometric evidence over a large space of candidate planes, making traditional voting approaches computationally expensive. We introduce the Fourier Accelerated Symmetry Transform (FAST), a training-free framework that combines geometric pair voting with efficient spectral computation. Exploiting the fact that global reflection planes pass through the shape centroid, we reduce the search to plane orientations, while geometric consistency filtering and independent witnesses suppress unreliable pair hypotheses. Our key insight is to represent these votes as a signal over plane orientations and efficiently compute a continuous, smoothed symmetry field in the spherical-harmonic domain. Specifically, we exploit the Funk-Hecke theorem, the spherical counterpart of the Fourier convolution theorem for rotationally symmetric kernels, to transform Gaussian smoothing into multiplication of harmonic coefficients. This enables efficient field evaluation, from which symmetry candidates are extracted, refined, and geometrically verified. Across two benchmarks, including a new benchmark of highly symmetric shapes, FAST achieves the highest clean multi-plane detection $F_1$ scores among evaluated methods (0.861 and 0.947), with GPU inference below 4 ms per shape: over $100\times$ faster than the next-fastest evaluated multi-plane GPU baseline. FAST also demonstrates strong robustness to geometric noise and competitive single-plane detection, achieving 94.9% success on clean single-plane shapes.

0.30

When Anchors Shift the Target: Advantage-Gated Adaptation for Diffusion Reinforcement Learning

cs.CV cs.LG ·2026-11-17 ·Pu Cao, Min Zhou, Ying Shu, Tiezheng Ge, Qing Song +1 more
Abstract · 12 pages, 4 figures, including supplementary material

Forward-process reinforcement learning adapts diffusion and flow-matching models by re-noising rewarded generations into regression targets. Because negative advantages induce concave updates with unbounded repulsive gradients, existing methods stabilize training using cached predictions from a moving policy copy, yet these anchors also alter what positive feedback teaches the model to fit. We formalize this conflict as anchor coupling: after separating frozen-reference regularizers, affine moving-anchor rules cannot eliminate negative curvature while preserving every positive regression target. We propose Advantage-Gated Adaptation (AGA) to decouple these roles. Positive samples retain standard target fitting, whose self-limiting update vanishes at the observed target independently of the moving anchor. Negative samples instead use the first-order tangent of the concave objective at the moving-policy prediction, preserving the local repulsion direction with a prediction-independent gradient. Within the quadratic-anchor family, these requirements uniquely determine the parameter-free gate $ω(a)=\max(-a,0)$. Because the moving copy provides local linearization rather than retention, AGA assigns long-horizon behavior preservation to a frozen pretrained reference regularized at its operating classifier-free guidance scale, interpreted through an implicit-reward lens. Across compositional generation, text rendering, and preference alignment, our experiments show that displacing positive targets degrades visual quality and heightens sensitivity to policy-copy schedules, whereas AGA consistently improves the reward--quality trade-off.

0.29

Decoding Without a Decoder: Direct Pixel Prediction from Diffusion Transformers

cs.CV ·2026-11-17 ·Hansheng Chen, Shengqu Cai, Jan Ackermann, Rayhan Zirvi, Gordon Wetzstein +2 more
code/data
Abstract · Project page: https://hanshengchen.com/dwd

Latent diffusion models rely on a decoder to map denoised latents to pixels, adding complexity and inference cost. We introduce Decoding Without a Decoder (DWD), which lets a finetuned diffusion transformer produce final pixels directly through a linear head. Unlike previous methods, where the denoising output leads to final pixels, DWD separates the two roles: the original denoising head continues to drive sampling, then the new DWD head predicts the pixels at the final step. We extend pretrained latent models with DWD efficiently by finetuning only the last few transformer blocks, training the DWD head with pixel quality losses while anchoring the denoising head to the pretrained model to preserve the sample trajectory. Across latent image and video generators, DWD removes the decoder and reduces inference cost while maintaining or even improving pixel quality. On Wan video generation, it visibly reduces detail distortion and flickering, lowering warping error by 23%. Remarkably, the same separation also improves pixel diffusion models that do not even use a decoder: a final-step DWD head produces substantially better texture detail in text-to-image generation, and improves ImageNet FID from 1.55 to 1.43, a new state of the art among pixel flow models.

0.29

Plausible Videos Can Lead to Poor Decisions: Diagnosing and CURE-ing World Models for Planning

cs.CV ·2026-11-17 ·Raphael C. Kim, Michele Santacatterina, Jingsen Zhu, Ramin Zabih
Abstract

Action-conditioned world models are increasingly used as world models for planning, yet a world model that makes plausible, realistic-looking videos that are consistent with the training data can produce poor decisions. We address two key questions: (1) when can we trust a world model for planning, and (2) what should we do when we cannot? We introduce CURE (Certify, Upweight, Rollout, Explore), a certificate-guided technique that assesses whether a world model can be trusted for planning and determines what to do next: whether to upweight the training data to debias the model; roll out more samples at test time; or explore by collecting new data where support (how well the training data cover the planner's decisions) is missing. We deploy CURE across multiple domains, spanning simulated physics based on PhyWorld, MuJoCo arm control, and Language-Table pushing in simulation and on real videos. The decisions our certificate trusts have lower regret than those it does not (e.g., 0.06 vs. 0.38 on Language-Table sim and 0.07 vs. 0.38 on MuJoCo contact), whereas video perceptual quality does not identify good decisions. By focusing on decision quality rather than perceptual quality, CURE lowers regret over the best pessimistic decision rule by 10-25% across support levels, and over world-action models, search and policy learning by 41-64%, given the same data.

0.29

Propose with the Gradient, Accept with the Loss: One Hedged Attack for Pixel-Sparse Robustness

cs.CV cs.CRcs.LG ·2026-11-17 ·Hon Ming Yan, Chun Pong Lau
Abstract · 22 pages, 13 tables, 4 figures

A pixel-sparse adversary may change a few pixels of an image to any value. Robustness to such attacks is currently measured with an ensemble, sparse-AutoAttack (sAA), because its members fail in different places: sparse projected gradient descent (sPGD) trusts a first-order model that a full pixel change breaks, and sparse random search (Sparse-RS) ignores the gradient even where it is informative. We show that both attacks are local searches, each a proposal distribution plus an acceptance rule, and ask what such a search can learn from the gradient. A Taylor bound says what to measure, and measurements at the states the search visits give five rules. The sign of the gradient is its most reliable use at a single pixel and sets a new pixel's value. Its ranking of candidate pixels should be followed loosely, and its ranking of the pixels to remove, which is often more reliable, can be followed more closely. A pixel should not be proposed twice before a move is accepted, and acceptance should be decided by the loss alone. These rules define Gradient-guided Random Search (GRS), which runs once and costs one forward and one backward pass per step. On 13 CIFAR-10 configurations and 5 ImageNet models, GRS matches or beats the best single member of sAA and is within one point of sAA itself or stronger on 17 of 18 settings, and it is significantly stronger than sAA on 8. It uses 3 instead of up to 7 forward-pass units, and on CIFAR-10 it runs 3.7 to 5.0 times faster than the members of sAA run in full.

0.29

Beyond the Camera: A Tri-Modal Driving World Model for Future RGB, Thermal, and LiDAR Forecasting in Adverse Weather

cs.CV ·2026-11-17 ·Faysal Mehrab Chowdhury, Md. Mahfuzur Rahman, Ahmed Rafi Hasan, Kishor Datta Gupta, Mohd Ariful Haque +1 more
Abstract

Driving world models forecast what a vehicle's sensors will record next. Most forecast camera video and some add LiDAR, but thermal cameras, which still see in darkness and some adverse weather, are rarely a forecast target, and to our knowledge no world model forecasts all three sensors. We study joint forecasting of RGB, thermal, and LiDAR on real adverse-weather sequences of the DSERT-RoLL dataset. We post-train a pretrained video world model for each sensor and represent LiDAR as a depth video aligned with the RGB camera, with densification and encoding chosen by what the model's video tokenizer can reconstruct. A tri-sensor model sees all three sensors as one video. We score LiDAR in meters against measured returns and compare against persistence baselines passed through the same tokenizer as the model. With this matching, an apparent loss of the thermal model to persistence in fog disappears. On all 58 validation sequences in six weather conditions, post-trained RGB and thermal forecasts beat compression-matched persistence. Joint training lowers the LiDAR error at 2s by 2.1m at a small cost in camera fidelity, and only the tri-sensor model's LiDAR forecasts beat persistence in mean error. Blanking one sensor degrades the forecasts of the others more in snow than in fog.

0.29

LOCUS: Causal Support Windows for Attribute Control in Autoregressive Image Generation

cs.CV ·2026-11-17 ·Zijun Huang
Abstract

Unified autoregressive (AR) image generators struggle to satisfy multiple attribute constraints in a single prompt. Although these attributes are decodable from image-token latents, editing those latents at test time remains difficult. Adding a fixed attribute direction to the prefix can reinforce a requested attribute but rarely overrides a conflicting one, such as turning a car red when the prompt specifies blue. We investigate how the effect of an intervention depends on where it acts in the image-token sequence. We introduce the causal support profile, which measures, for each attribute, how often patching image-token states within a given window flips the rendered attribute. Structural attributes (count and position) are most responsive in early, narrow windows, while appearance attributes (color, size, and material) respond over later, broader windows. A budget-matched comparison of early and middle windows shows that the effect of window location depends on the target attribute, and two-dimensional support maps show that the responsive regions track spatial content. Based on these profiles, we propose LOCUS (Latent Optimization in Causal Support windows), which diagnoses which attribute failed and routes test-time latent optimization to its corresponding support window. With the generator, optimizer, reward, verifier, and token budget held fixed, support-routed windows outperform late windows by 23.6 percentage points in GenEval++ macro accuracy. With the generator frozen, LOCUS raises Janus-Pro's GenEval++ accuracy from 24.6% to 46.8% overall and from 19.5% to 40.5% on multi-attribute tags. On GenEval, LOCUS repairs 48.1% of failed generations in one rollout, compared with 37.4% for prefix optimization, and needs 14% fewer rounds on average to succeed.

0.29

CodeReel: Accelerating Code-to-Video Agents with Exact Frame Reuse and Blocking-Aware Scheduling

cs.MM cs.AIcs.CVcs.DC ·2026-11-17 ·Junyi Shen, Yao Lu
benchmark
Abstract · 19 pages, 8 figures, 14 tables

Coding agents make videos by writing and refining programs in a build--inspect--edit loop. A finished film costs a median of $2.85 in model calls and 11 minutes, over a third of that time in builds that block the agent. We find that much of this time is avoidable, for two reasons. First, 37.5% of the frames of a session's later films were drawn before, many shifted in time or inside an edited scene. Second, half of the builds stop at a check, where the agent needs only the verdict, but which builds stop is unknown until the checks finish. We present CodeReel, a training-free execution layer behind agents' existing commands. Exact frame reuse keys each frame by its drawing commands, regardless of its position in time, and reuses each frame whose key recurs. Blocking-aware scheduling runs the checks first and starts the render alongside them, at idle priority, unless the session's recent builds make a pass unlikely. Scheduling spends CPU to answer sooner and reuse returns it; exact keys let later versions use frames rendered for a failed one as the agent thinks. We introduce EditLoop, a benchmark of replayable agent sessions whose films are checked against renders from scratch. On it, CodeReel cuts the agent's wait by 36.3% against Rebuild, a warm pipeline with parallel rendering, and by 57.9% against Sequential, today's step-by-step pipeline. On time-shifting edits its reuse renders 43.9%, close to the 43.0% whose pixels are new, while Manim-style caching renders 99.8%. With real agents, the median wait per build is 0.62x Rebuild's in Claude Code, on topics never used in development, and 0.67x in Codex CLI. The gain holds for five models and two frameworks.

0.29

The Rhythm of Generated Video: Latent Chunking Leaves Periodic Traces in Compressed Bitstreams

cs.CV ·2026-11-17 ·Chen Ye Gan, Bryan Zhen Zhang, Gene Wen, Yuxing han
Abstract · Manuscript, 9 pages, Computer vision, AI-generated video detection, compressed-domain analysis, video coding,

Most AI-generated video detectors analyze decoded pixels and overlook how videos are produced. We show that causal video autoencoders leave an architecture-specific rhythm in compressed bitstreams: their temporal stride concentrates unpredictable content at latent-chunk boundaries, so rate-distortion optimized encoders allocate bits periodically. Without decoding pixels, we observe four-frame pulses in CogVideoX and Wan2.2 and eight-frame pulses in LTX-Video, which are absent from content-matched camera footage. Passing camera footage through a generator's autoencoder alone induces the pulse, supporting its proposed origin. A training-free peak-ratio statistic separates generated from camera video (AUC 0.84-0.94) and persists across tested quantization settings and H.264/H.265 encoding. Frame-size traces also identify four commercial generators, with 89.2%, 84.1%, and 83.0% accuracy under H.264, H.265, and AV1, beating the pixel-based UnivFD baseline (72.7-73.6%). Our results directly link generative architecture to measurable coding demand.

0.29

RefCloth: Transferable Adversarial Tablecloths for Vision-Language-Action Models

cs.CV ·2026-11-17 ·Shaozhen Liu, Zhiwei Li, Jiayang Sun, Shuangjun Du, Yi Zhong +4 more
Abstract

Closed-loop vision-language-action (VLA) policies rely continuously on multi-view cameras, yet existing visual attacks typically hijack sensor streams with 2D patches or alter target object meshes that require retraining when manipulation objects change. We introduce RefCloth, a framework that optimizes one persistent adversarial tablecloth on a frozen source policy and deploys it zero-shot across unseen tasks and target models without modifying task objects or cameras. Throughout manipulation, static third-person and moving wrist views observe this shared world-bound texture under evolving occlusions. To generalize across shifting surface views while bridging the gap between differentiable and native renderers, RefCloth combines bounded texture parameterization with shared surface UV transformations and a same-path clean reference that removes static rendering offsets at initialization. Across four LIBERO suites in native simulation, one frozen texture transfers from OpenVLA-OFT to $π_0$ and $π_{0.5}$, reducing success relative to an unoptimized cloth by 62.50 and 43.75 percentage points on average and far exceeding amplitude- and variation-matched random controls. Dual-view audits and target closed-loop camera interventions show that moving wrist exposure drives over twice the standalone drop of the main view with strong synergy, while ablations confirm the roles of UV diversification and path calibration.

0.28

LUTBench: Benchmarking Text-Driven Photographic Look Transfer

cs.CV ·2026-11-17 ·Chuanzhi Xu, Boyu Wei, Chengkun Yue, Xuanhua Yin, Yiyang Li +3 more
Abstract

Photographic color grading shapes the tone and color of a photograph to achieve a desired look. Text-driven photographic look transfer lets users specify these adjustments in natural language, with the goal of reproducing the requested look across photographs while preserving their content. However, prior research on text-driven color grading uses different datasets, text instructions, target definitions, and metrics. These differences make it difficult to compare how faithfully the methods reproduce the same look. To compare methods under a shared protocol, we introduce LUTBench. We collect 91 three-dimensional lookup tables (3D LUTs) and define one photographic look per LUT with a specified input conversion. Applying these looks to 5,000 MIT-Adobe FiveK photographs produces 455,000 input-target image pairs with reproducible target images. For each look, we use templates to turn measured tone and color changes on fixed color samples into candidate clauses. We use training image measurements to verify applicable clauses before composing the text description. The evaluation measures target fidelity, tests sensitivity to predefined input conversions, and separates generalization to unseen looks from reproduction of seen looks on new photographs. Across 16 methods, stronger text alignment does not guarantee target fidelity. However, most supervised variants improve fidelity on both seen and unseen looks.

0.28

Reliability-Driven Feature Correction for Hyperspectral and LiDAR Classification

cs.CV ·2026-11-17 ·Lianming Zhang, Xiaolu Wang, Qiya Song, Jinyang Liu, Renwei Dian +1 more
Abstract

Hyperspectral image (HSI) and LiDAR data provide complementary spectral and elevation information for land-cover classification. However, the spatially varying class-discriminative reliability of modality features is rarely modeled explicitly, allowing weak or ambiguous features to adversely affect cross-modal interaction and fusion. To address this issue, we propose ReDF, a reliability-driven framework that quantifies the reliability of each modality's features at each spatial location through the strength of class-related evidence they provide, and uses it to guide feature correction and fusion. Specifically, a Reliability-Guided Feature Correction module introduces a class prototype bank for each modality and converts feature-prototype similarities into class evidence for reliability estimation. The estimated reliabilities are used to construct a reliability-weighted shared representation and to guide bidirectional cross-modal feature correction based on neighborhood agreement, refining modality features with weak class evidence through information from the other modality. Building on these outputs, a Reliability-Guided Feature Reweighting module conditions a low-rank modulation on the estimated reliabilities and the cross-modal disagreement, and applies it to the corrected modality-specific descriptors to adaptively rescale each modality's contribution to the fused representation used for final classification. Experiments on three public HSI-LiDAR datasets show higher mean overall accuracy and Kappa than the compared methods.

0.28

CoSphere-SLAM: Multi-Agent Panoramic Gaussian Splatting SLAM with Spherical Loop Closure

cs.CV ·2026-11-17 ·Qi Shao, Zhihao Cao, Shuhao Zhai, Taoyu Wu, Zhuoxiao Li +6 more
Abstract · 9 pages, 4 figures, 7 tables. Project website: https://cosphere-slam.github.io

Dense collaborative mapping is essential for search and rescue, building inspection, and warehouse automation, where panoramic cameras capture the entire surroundings in a single frame. Unlike pinhole agents that share views only at similar headings, panoramic agents observe overlapping scenes at any 3D orientation, yet searching full rotations also produces mutually consistent false loop closures across distinct symmetric rooms. Moreover, continuous monocular scale drift splits rigidly merged maps at revisited surfaces, while depth noise at short baselines distorts loop translation directions. To address these challenges, we propose CoSphere-SLAM, the first multi-agent Gaussian splatting SLAM framework for panoramic cameras, whose spherical loop closure aligns spherical features over $\mathrm{SO}(3)$ via fast Fourier correlation across arbitrary orientations and rejects symmetric false closures through global $\mathrm{Sim}(3)$ consensus. Building on these closures, directional global optimization solves a per-keyframe $\mathrm{Sim}(3)$ pose graph with log-scale regularization to correct scale drift while separating optical-flow translation directions from depth-based magnitudes. Finally, omnidirectional Gaussian fusion transforms each Gaussian with its keyframe similarity and renders Gaussians at their maximum ray response to jointly refine poses and maps. Experiments on indoor and cross-session multi-agent panoramic benchmarks demonstrate that CoSphere-SLAM achieves state-of-the-art trajectory accuracy and rendering quality among single-agent and multi-agent methods. Project website: https://cosphere-slam.github.io.

0.28

Automated Annotation-Free Synthetic Data Generation for Underwater Litter Detection: A Comparison of Generative Approaches

cs.CV ·2026-11-17 ·Amala Mary Vincent, Bart De Schutter
Abstract

Underwater object detection is often limited by scarce and imbalanced training data and by strong appearance variations caused by turbidity, lighting, and differences between sensors. Existing data augmentation strategies, such as geometric or color augmentation or fully synthetic image generation with Generative Adversarial Networks (GANs), either lack semantic realism, fail to respect scene physics, or require additional annotations. In low-data domains like underwater environmental monitoring, the need for additional annotations is often the core bottleneck. Collecting new footage is hard, but labeling this dataset is even more time-consuming and costly, which limits how quickly a usable dataset can be constructed. We propose a unified fully automated zero-annotation pipeline that inserts class-conditioned trash objects into real underwater frames while preserving the original background context. Starting from sampled bounding boxes, each box is converted into an insertion mask and the object is pasted or generated within it, while the same box coordinates are automatically written as the corresponding annotation. Because labels are derived directly from the sampled boxes, the pipeline generates fully labeled training data for low-data settings without any manual annotation. Within this single pipeline we compare three strategies: copy-paste, Stable Diffusion inpainting, and DALL-E image-conditioned generation. We evaluate bounding-box faithfulness, visual realism, and the impact of the generated samples on downstream detector performance on the SeaClear underwater debris dataset. Results show that DALL-E improves detection performance in four of five domains, with relative gains of up to 96 percent over the real-only baseline.

0.28

SIDUMN: Self-Supervised Image Denoising Under Symmetric Multiplicative Noise

cs.CV ·2026-11-17 ·Nyan Htet Naing, Gyeongha Hwang
Abstract · 10 pages, 4 figures, 2 tables. Submitted to IEEE Signal Processing Letters

Self-supervised denoising is challenging when the corruption is multiplicative and signal-dependent. In this letter, we propose SIDUMN, a symmetric noisier framework for self-supervised denoising under mean-one symmetric multiplicative noise. Given noisy observations and the noise law, the method samples an independent synthetic noise factor from the same law and constructs two complementary noisier images by applying the factor and its reflection about one. The denoising network is trained to predict one noisier image from the other, and at inference the reconstructions of $K$ such antithetic pairs are averaged. We show that the population target of the training loss is the clean conditional mean minus a multiplicative cross-term. We further prove that, for any trained network, antithetic averaging never increases the expected squared error of a single noisier-input reconstruction. Numerical results validate the theoretical findings and the proposed method.

0.28

Is model self-assessment enough for diverse LVLM ensembles?

cs.CV cs.LG ·2026-11-17 ·Michael Smith, Frank P. Ferrie
Abstract

Ensembles of diverse Large Vision Language Models have demonstrated performance beyond that of single models across multiple datasets and tasks. However, the necessary aggregation process usually relies on training, parameter fitting, or complex task-specific pipelines. Furthermore, evaluation is limited, making a fair comparison difficult. We create a unified pipeline where we perform candidate response selection: we build an ensemble by selecting the best answer from a set provided by different models. We comprehensively evaluate across 34 different means of confidence estimation, 24 ensemble configurations, 20 datasets, and two prompt specifications. We show that simple training and parameter-free approaches can be sufficient, even for ensembles of different architectures. These simple self-assessment approaches rely solely on each constituent model's confidence, without explicit calibration, yet are able to perform better than said constituent models. They are also competitive with more involved alternatives on standard metrics such as accuracy, including on open-ended problems. However, we also discover that dataset, ensemble composition, and prompt wording are key confounding variables insufficiently explored by prior work that significantly affect ensemble performance, such that the ensemble can perform worse than its best constituent model. These must be optimized for to get the most out of ensembles, at the cost of limiting their generalizability. We recommend that practitioners consider training-free ensembles only if deployment conditions can be precisely characterized, with Best-of-N configurations of strong models being a good starting point.

0.27

Nunchi: Ears Help Eyes in Multi-Party Conversations

cs.CV cs.CL ·2026-11-17 ·JunYeong Park, Shijing Wang, Seonwook Park, Michael Stengel, Shalini De Mello +2 more
benchmark
Abstract · 10 pages, 5 figures

Identifying whom a person looks at is central to understanding multi-party conversations, yet gaze geometry alone becomes ambiguous as more people appear. Conversations offer cues that geometry lacks, such as who is speaking and what is said, but no gaze dataset provides them for many multi-party scenes. We introduce SORBET, a benchmark of 3,340 samples from 1,578 clips with synchronized speech, transcripts, and attended person labels, covering more multi-party scenes than existing gaze datasets. Prior gaze models struggle to use these cues, which come in different forms, must be linked to the right person, and can conflict with gaze geometry. We propose Nunchi, a framework that constructs a text prompt from the transcript and the outputs of gaze and speaker models, links every cue to the candidates with shared labels, and lets a vision-language model (VLM) weigh gaze geometry against conversational cues to identify the attended person. Nunchi outperforms the strongest existing method by 4.8 percentage points. On SORBET, conversational cues help whereas additional visual cues do not, and the gains hold across three VLMs, with automatic transcripts, and on two existing datasets.

0.27

PhyVGGT: Feed-Forward Articulated Object Reconstruction from Monocular Videos

cs.CV ·2026-11-17 ·Qi Shao, Zhihao Cao, Shuhao Zhai, Anh Nguyen, Hong Liang +2 more
code/data
Abstract · 10 pages, 5 figures, 7 tables. Project page: https://phyvggt.github.io

Reconstructing an articulated object with its 3D movable parts, joint parameters and unseen articulation states from an uncalibrated monocular video is essential for robot manipulation and interactive digital twins. Existing per-scene optimisation methods require minutes to hours and external depth or camera poses, whereas feed-forward 3D foundation models reconstruct dense point maps in one pass without segmenting movable parts or inferring joints. Even on these monocular point maps, regressing or fitting continuous joints directly fails on casual videos, since cross-frame attention blurs moving part boundaries in deep layers, underestimates motion along the viewing ray and leaves continuous axes ill-conditioned under small opening angles. To address these challenges, we propose PhyVGGT, a single-pass framework on a frozen 3D foundation model that anchors moving-part matching and closed-form joint estimation to the undistorted geometry of the stationary base. First, our Part and Motion Decoder fuses intermediate per-frame tokens with global tokens and references part queries to a static-base query to segment 3D parts, align point maps on the static base and predict learned Bingham priors. Second, our Kinematic Inference Layer fits an orthogonal Manhattan frame to static surface normals, selects the joint axis in closed form under a Bingham posterior combining pairwise motion scatter with the learned priors, and solves a linear system for the hinge pivot to drive unseen states. Trained solely on synthetic data, PhyVGGT achieves state-of-the-art monocular performance across five benchmarks, outperforming FAMOS and monocular ArtiPoint, surpassing posed RGB-D ArtiPoint in median error, and running in 6.8 s per 48-frame video against hours of per-scene optimisation in Articulat3D. Project page: https://phyvggt.github.io.

0.27

Distilling World Action Models for Efficient Bimanual Manipulation via Anchored Path Matching

cs.CV cs.RO ·2026-11-17 ·Jialong Liu, Junjie He, Yandu Chen, Bin Zhou, Heng Su +3 more
code/data
Abstract

World Action Models (WAMs) jointly generate future video and robot actions through iterative denoising. Few-step distillation reduces the number of model evaluations, but its coarse updates can move the student away from the teacher's trajectory. We propose Anchored Path Matching (APM), which supervises endpoint and short-transition predictions from states reached by the student. From each state, we continue the teacher to zero noise and use its endpoint and first intermediate state as targets. Modality-routed adapters learn to match both targets from the same starting state. When the two gradients conflict, anchor-priority updates preserve the endpoint gradient and remove the opposing component of the local gradient within each modality. One checkpoint supports 1-, 2-, and 4-step joint generation. The sampling budget can be chosen at deployment without retraining or teacher queries. On RoboTwin 2.0, we train only on Clean demonstrations and evaluate directly on the domain-randomized Random setting. At one step, APM improves closed-loop success over continued base distillation at matched inference budgets by 2.0 percentage points on Clean and 1.4 on Random. Video peak signal-to-noise ratio (PSNR) rises from 30.169 to 32.317 dB, with gains in structural similarity (SSIM) and learned perceptual image patch similarity (LPIPS). On two real-robot tasks, APM improves success over the one-step teacher and reduces control-cycle latency by $21.3\times$ relative to the eight-step teacher.Project page: https://jialonglong.github.io/Anchored-Path-Matching/.

0.27

Centroid Spectral Coupling: Implicit Alignment and Representation Diversity in Self-Supervised Learning

cs.CV cs.LGstat.ML ·2026-11-17 ·M. Hadi Sepanj, Benyamin Ghojogh, Saed Moradi, Paul Fieguth
Abstract

Self-Supervised Learning (SSL) aims to align augmented views while preserving representation diversity and preventing collapse. Existing non-contrastive methods often address these goals using multiple weighted loss terms. We introduce Centroid Spectral Coupling (CSC), a single-term SSL objective that implicitly couples view alignment, representation diversity, and collapse resistance. CSC constructs centroids from independently normalized augmented-view pairs and applies a ridge-based spectral objective to their uncentered second-moment matrix. Preserving centroid magnitudes encodes view agreement while encouraging balanced representation energy across spectral directions. We establish an exact decomposition into nonnegative alignment and spectral-imbalance components, characterize the geometry of global optima, and derive finite-loss guarantees for alignment and diversity. We also show that complete constant-output collapse is a strict saddle in normalized representation space. Experiments on CIFAR-10, CIFAR-100, and Imagenette show that CSC achieves the highest linear Top-1 accuracy among the evaluated non-contrastive methods on all three benchmarks, and the highest kNN Top-1 accuracy on both CIFAR datasets. A ridge-parameter sensitivity study reveals an empirical alignment--diversity trade-off: increasing $κ$ strengthens view alignment but reduces representation effective rank. These results show that a unified spectral objective yields competitive visual representations without explicit contrastive negatives or separately weighted alignment and diversity penalties.

0.26

PCRG: Training-Free Pruning and Caching of Reference Tokens for Video Generation

cs.CV ·2026-11-17 ·Siyuan Wang, Haoran Li, Yi Yue, Tak L. Fung, Quan Chen
benchmark
Abstract · 11 pages, 5 figures

Reference-to-video diffusion transformers pack the variational autoencoder (VAE) latents of all references and the generated tokens into one full-attention sequence, in which reference tokens make up about half. Because attention cost grows quadratically with sequence length, these tokens take up an even larger share of the attention computation, yet most of them are static, smooth or irrelevant to the output. To accelerate reference-conditioned generation, we propose pruning and caching of reference tokens for video generation (PCRG), a training-free, plug-and-play method that uses two signals the model already computes to assign each reference token to one of three tiers, namely pruned, cached or kept. The first is a low-level redundancy r from VAE latents, since a token whose latent barely changes over time or across its neighbors adds little new content. The second is a high-level importance g from the vision-language-model (VLM) conditioner, where the attention that the prompt and any first or last frames pay to a token tells whether the output needs to read it. Tokens that are not redundant are kept, and redundant tokens are cached if important and pruned otherwise. Pruned tokens leave the sequence before the first step with all rotary position embedding positions unchanged. Cached tokens have their keys and values computed once and reused, and kept tokens are processed as usual. On MiniMax-H3, already distilled to 8 timesteps and quantized to int8, PCRG is 1.57x faster end to end on our V2V-Bench and 1.49x faster on the public VACE-Benchmark, while the quality of VBench drops by only 0.24% and 0.17%, respectively. PCRG shortens the sequence rather than changing how attention is computed or which steps are run, so it also stacks on step caching and sparse attention. On V2V-Bench it reaches 1.86x with TeaCache, 1.82x with Spark at nearly unchanged fidelity and 2.15x with Veda.

0.26

Kolibri: Color Finds the Place, Refinement Sets the Pose for RGB-D Localization in Colored Point Clouds

cs.CV ·2026-11-17 ·Leonel Aguilar, Bartu Tuncay, Christoph Hoelscher
Abstract

An agent or robot in a scanned building must locate its camera in rooms that no model saw during training. Point cloud registration methods can align each RGB-D frame with the building's colored point cloud, accurately but at a high cost. We observe that geometric refinement brings nearby initial poses to the same final pose, so the learned component only needs to find the place. Kolibri does this with a graph network of 55,712 parameters that describes 10 cm cubes of points, called cells, by their color, local shape and neighbors. With the same refinement, it matches GeDi on 96 unseen ScanNet rooms (81.4% against 80.3% within 5 cm and 5°) and on a held-out test of 10 further rooms, in 82 ms per frame on an RTX 2080 Ti. The saved time allows for search without a room label. In one point cloud of 100 rooms, Kolibri loses 1.3 points against knowing the room, while retrieving the room first loses 56.0. In a hospital emergency department scene from a real scan, it localizes 86.5% of the frames rendered from the scan in 384 ms each, where GeDi needs 91 minutes per frame on cubes of the scan for a similar recall. The method needs colors that agree between the frame and the point cloud. On gray frames and across cameras with different color responses, point cloud registration methods keep the lead.

0.26

CAST: Calibrating Correspondences for Training-Free Biomedical In-Context Segmentation

cs.CV ·2026-11-17 ·Mingze Jiang, Jiashu Han, Haonan Chen, Zhiyi Li, Yilun Du +1 more
Abstract · 10 pages, 6 figures

Biomedical image segmentation must keep up with new anatomical structures, subjects, imaging modalities, and acquisition protocols, yet dense expert annotations are scarce and models trained on one cohort often struggle to transfer. In-context segmentation (ICS) offers a way out: a new structure is specified by a few annotated exemplars at inference time. However, reliable correspondence remains challenging in biomedical images, where adjacent structures often share similar appearances, and variations across subjects and imaging conditions further obscure their distinctions. We introduce CAST, a training-free framework built on a frozen DINOv3 backbone to calibrate fine-grained correspondences. Its stability test improves feature reliability at both global and patch-specific levels, while its coherence test assesses whether candidate correspondences are supported by consistent local structural context. Together, these tests enable more reliable correspondence across diverse anatomical structures and imaging conditions. Across eight biomedical benchmarks spanning 2D and 3D segmentation, cross-subject, and cross-modality transfer, CAST outperforms the strongest overall baseline by 12.34 percentage points in Dice and ranks first across all six volumetric settings. The versatility of CAST further opens up opportunities for challenging biomedical applications, including 2D-to-3D segmentation and atlas label transfer.

0.26

Pixel is All You Need

cs.CV ·2026-11-17 ·Jie Chu, Haoting Shi, Fucheng Zhang, Jiawei Liu, Junwei Liao +2 more
Abstract · 20 pages, 6 figures, 7 tables, including references and appendices

Text, images, audio, touch, and even robot actions can, in principle, all be rendered onto a single canvas, suggesting an ambitious goal of unifying all modalities within vision. Language provides the natural first test. To subsume it, a visual model must reason directly over text in pixels. Across a wide range of image, video, and unified multimodal generators, however, we find that their visual generation modules cannot perform symbolic reasoning over text on the canvas. They succeed only when an underlying text module or prompt rewriter already supplies the answer, effectively reducing the generator to a mere renderer, a limitation that post-training does little to resolve. To test whether this failure stems from pixels themselves or from current training paradigms, we train a family of purely pixel-space generative models from scratch without any text pathway. With images as their sole input and output, these models read questions, write intermediate steps, and produce answers directly on the canvas. Remarkably, they exhibit hallmark behaviors of large language models. Performance improves predictably with scale following a power law. Writing intermediate steps frame by frame substantially outperforms single-frame answering, echoing chain-of-thought, while reinforcement-learning post-training further strengthens reasoning. Moreover, sharing one canvas allows the same model to reason across modalities, a capability that text-only language models lack. To the best of our knowledge, this is the first systematic study of these phenomena in purely pixel-space generative models free of text tokens.

0.26

Classification of Tissues on MALDI-MSI Image Modality for Pancreatic Cancer Studies

cs.CV ·2026-11-17 ·Abubakar Qahir, Jenny Benois-Pineau, Victoria Bourgeais, Fabien Baldacci, Sandrine Dabernat +4 more
Abstract

Mass Spectrometry Imaging (MSI) spatially maps molecular distributions in tissue sections. Despite the availability of preprocessing tools for MSI data, building end-to-end workflows suitable for deep learning remains challenging. We have developed a fully automated pipeline for transformation of raw Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry imaging (MALDI-MSI) data into deep-learning-ready image patches for binary pancreatic tissue classification (tumoral vs non-tumoral) and built individual models for recognition of tumoral regions in spectral images. Images resulting from the proposed pipeline and from reference interactive workflow are very close in terms of image quality metric mean squared error (MSE). The proposed pipeline computational time is of 20 min in average compared to a full working day of the reference interactive workflow. The models achieved 0.94 accuracy on high-resolution spectra and a mean accuracy of 0.59 on lower-resolution spectra.

0.26

Teaching What to Look For: Privileged Perception Questions for Agentic On-Policy Distillation in Video Captioning

cs.CV ·2026-11-17 ·Yifan Xu, Ruopeng Gao, Zhixiao Wang, Jiafu Tang, Yichun Yang +3 more
code/data
Abstract

Detailed video captioning requires comprehensive and faithful descriptions of objects, actions, and temporal events. Fixed-frame sampling, however, can miss brief motions, small objects, and other fine-grained evidence, motivating captioning agents that actively seek relevant moments and inspect local regions. Training such behavior is challenging because, unlike video question answering, generic captioning instructions do not specify what visual evidence should be acquired. We propose Perception-Question On-Policy Distillation (PQ-OPD), which uses training-only Perception Questions (PQs) to guide active video perception. PQs specify what should be inspected without revealing the corresponding answers, encouraging the teacher to explore the video more comprehensively and focus on relevant visual evidence. Along trajectories generated by the student, on-policy distillation transfers this guidance to the student's reasoning, tool calls, and captions; no PQs are required at inference. Across three detailed-captioning and two caption-to-QA benchmarks, PQ-OPD consistently improves the corresponding untrained agents across teacher-student scales and input-frame budgets. Under our evaluation protocol, PQ-OPD achieves the strongest primary-metric results among the compared open-source models on DREAM-1K, CaReBench, ShortVidBench, MotionBench, and VidCapBench-AE. Our ablations, together with analysis of temporal coverage and answer leakage, further indicates that the student internalizes the training-time guidance as a more effective video exploration policy.

0.25

Wear and Go: Calibration Free Sparse Inertial Motion Capture via Online Sensor Setup Estimation

cs.CV ·2026-11-17 ·Yuanshuo Tan, Licheng Zhong
Abstract

Sparse inertial motion capture typically assumes that each sensor's anatomical identity and mounting orientation are known. We present MOSAIC (Motion-based Sensor Assignment and Inertial Calibration), a framework that recovers this setup from ordinary motion without a calibration pose. Our central insight is that body pose changes continuously while the sensor setup persists, allowing movements that are individually ambiguous to provide complementary evidence over time. A setup network processes unordered sensor streams with temporal and cross-sensor attention, then accumulates assignment and rotation evidence into a session-level estimate. The recovered setup canonicalizes measurements for frozen pretrained pose estimators, separating sensor configuration from motion reconstruction. For sensors worn at prescribed body sites, evaluation on real inertial recordings with randomized setups shows near-calibrated pose accuracy for bounded offsets and substantial benefits from accumulation under arbitrary orientations. Transfer across pose networks supports setup recovery as a reusable interface between interchangeable sensors and pretrained pose models.

0.25

MIRAGE: Motion Identity Reduction via Autoregressive Generative Encoding for Privacy-Preserving Skeleton-based Motion Data

cs.CV ·2026-11-17 ·Thomas Carr, Xujun Che, Depeng Xu, Shuhan Yuan, Aidong Lu
Abstract · 14 pages, 3 figures, 12 tables

Skeleton motion identifies people through body proportions and movement style, and skeleton streams are increasingly passed to third-party analytics that run action classifiers trained on raw data. A prior adversarial anonymizer adds a per-frame residual in the raw coordinate space but sees one frame at a time; motion retargeting needs a target skeleton and complete sequences, which rules out online use. We present MIRAGE, a target-free, causal anonymizer that adds an identity-reducing residual to each incoming frame in the raw coordinate space, computed by attention over past frames. One causal, optionally windowed mask governs every attention in a rotary-position encoder-decoder, and key/value caches let the trained function run frame by frame; on all 56,578 NTU 60 sequences, batched and streaming outputs differ by at most 7.2e-6 m. Reported results use an unbounded window, so per-stream state grows with stream length. We evaluate against a passive adversary that applies raw-trained identity models and an adaptive adversary that trains on the anonymizer's output, treating the adaptive adversary as the headline, on NTU RGB+D 60 over five generator seeds, alongside the per-frame residual anonymizer of Moon et al. trained identically. When consumers retrain on the released data, MIRAGE keeps 83.9% action recognition (raw 89.1%; Moon et al. 88.7%) while a retrained identity classifier reaches 43.4% re-identification (RI) over 40 subjects, against 69.4% for Moon et al. and 75.4% on raw data. The strongest adaptive attacker still reaches 49.3% (chance 2.5%), so MIRAGE reduces identity leakage; it does not anonymize. Against raw-trained (frozen) classifiers, MIRAGE lowers RI below the strongest retargeting baseline (10.36% vs. 22.09%), but frozen action recognition falls to 37.04% (vs. 75.82%), so at full residual scale it is not a drop-in replacement for raw input to existing frozen models.

0.25

Strong Fused Performance Can Mask Branch Collapse in a Dual-View Heart-Sound Classifier: A Three-Part Diagnostic Protocol

cs.CV ·2026-11-17 ·Safayet Khan, Hasan Zohirul Islam, Md. Tauhid Bin Iqbal
Abstract

Fusion models are usually judged by their fused score, a practice that assumes that every branch has learned something useful. We show that this assumption can fail silently in dual-view heart-sound classification. In AudioFuse, which fuses a spectrogram transformer with a waveform CNN, the spectrogram branch is unused despite strong fused performance when both branches are trained from scratch. We introduce a three-part diagnostic for branch collapse: whether deleting a branch changes the fused score, whether a linear probe on its features beats an untrained copy of the same branch, and whether the branch can learn its view when trained alone. Deleting the branch has negligible effect, and probing its features shows little improvement over an identically initialised, untrained copy. Yet the same architecture trained alone exceeds 0.9 AUC, so its view is learnable, while the jointly trained model is near chance when it must rely on this branch. Update statistics and an initialisation experiment point to an early optimisation imbalance. Moving a strong artificial label cue from one input to the other can switch which branch the model relies on, even though the branch architectures stay the same. We observe branch collapse on PhysioNet, but the full diagnostic is not met on ICBHI or CirCor. Fused performance alone therefore cannot establish that every branch learned a meaningful representation, and we recommend reporting branch deletion and an untrained-control probe alongside it.

0.25

The shape of individual differences in judgment: shared axes, personal weights, and when their spread crosses the consensus

cs.CV q-bio.NC ·2026-11-17 ·Hiroshi Higashi
Abstract

People differ in how they judge faces, artworks, photographs and jokes. Previous work measured the size of these differences; we ask what shape they take. We wrote three candidate shapes as nested generative models (same-sign reweighting of shared dimensions, a linear transform that allows reversal, and stimulus-specific offsets) and compared how well they predict people unseen in training, using ten public datasets (about 20,000 raters). The models recovered a person's judgment space from 10-20 judgments and predicted judgments of unseen stimuli. For faces, reweighting was sufficient, and weights opposing the population were not statistically detected. For artworks, weights spread continuously along the shared abstract-figurative direction, and 28 percent of raters fell on the opposite side from the population. This tail was much smaller when a different rater sample judged the same artworks. Individual differences thus differ in shape as well as size, and the shape depends on domain and raters.

0.24

CueSplat: Learning to Schedule Sparse Deformation for Dynamic Gaussian Splatting on Edge Devices

cs.CV ·2026-11-17 ·Qingyuan Tang, He Zhu, Shixuan Sun, Anbang Wu, Jingwen Leng +2 more
Abstract · 10 pages, 6 figures. Main paper only

Deformation-based 4D Gaussian splatting (4DGS) enables dynamic scene rendering. However, evaluating deformation networks for all Gaussians at every frame is compute-intensive on edge devices. We find that Gaussian trajectories are often locally predictable, yet prediction reliability varies across Gaussians and time. Thus, we propose CueSplat, a post-training framework that exploits local temporal regularity while keeping the pretrained backbone frozen and retaining its Gaussian set. Specifically, CueSplat predicts Gaussian attributes from recent history and uses a lightweight learned router to select exact deformation updates within a scene-specific budget calibrated offline. Selection accounts for prediction reliability and rendering contribution. Selected indices are reused across nearby frames to reduce routing overhead, while selected Gaussians still receive exact updates each frame. To control accumulated prediction error, we use drift signals to schedule three consecutive full-deformation frames that rebuild the prediction history. Evaluated across four mainstream 4DGS algorithms, CueSplat achieves up to $6.74\times$ and $8.15\times$ speedups on Jetson AGX Orin and Orin NX, respectively, with a negligible quality drop (at most $0.16$ dB in dataset-average PSNR).

0.24

DAP-MR: Degradation-Aware Prompt-Magnitude Routing for Real-World Ultrasound Super-Resolution

cs.CV ·2026-11-17 ·Ngoc-Hoang-Lam Le, Thi-An Le, Dinh-Hung Le, Manh-Hung Nguyen, Kai-Lung Hua +1 more
Abstract

When super-resolution models designed for natural images are used on ultrasound scans, relying on semantic text prompts can lead to unreliable results. The tags these models generate often do not match the actual restoration needs, and specialized medical terms may be misunderstood by general-purpose text encoders. Instead, prompts focused on common image degradations are more directly useful for guiding restoration. However, using a single prompt to describe both the type and severity of image degradation can entangle these factors, making it hard for the model to know how strong the degradation is. To address this limitation, we introduce Degradation-Aware Prompt--Magnitude Routing (DAP-MR). Each degradation type is represented by a dedicated token initialized from the contextualized EOS embedding of a degradation-oriented prompt and subsequently adapted during ultrasound SR training. A separate magnitude router predicts a continuous, input-dependent magnitude that modulates the corresponding token. Thus, language provides a degradation-aware initialization for what should be restored, while magnitude routing supplies an explicit signal for how strongly it should influence restoration. DAP-MR can be added to both one-step and multi-step diffusion super-resolution frameworks without changing their core structures. Our experiments--on synthetic datasets with both in-domain and under anatomical-domain shift, as well as unpaired real-world ultrasound data--show that DAP-MR delivers better restoration quality and generalizes more consistently across different domains.

0.24

Must You See Fakes to Catch Fakes? Language Says What to Look For, Real Images What to Ignore

cs.CV ·2026-11-17 ·Tianchi Xu, Yuxuan Zhou, Yuheng Zhang, Yanan Guan, Tao Dai +1 more
Abstract

Must a detector see generated images in order to catch them? Detection of AI-generated images has long faced two problems. The cues a detector learns from the few generators in its training set carry traces specific to those generators and transfer poorly to new ones, which limits generalization. The real and generated training images also differ in format and content, so the detector takes these differences as evidence of authenticity, and its decisions change with factors that have nothing to do with authenticity, which harms robustness as well as generalization. Both problems stem from training on generated images. A detector never trained on them can largely avoid both at the source. Vision foundation models have already learned the concept of "AI-generated" during pretraining, and text can point to it directly. Building on this, we propose SaySee, a training-free detector: Say builds a generation direction from sets of paired descriptions that differ only in authenticity and scores an image by the projection of its embedding onto it, and See uses real images to correct the score offsets caused by format, image type and content. On 24 evaluation sets that span standard benchmarks, in-the-wild data and the latest generators, SaySee reaches a mean AUC of 98.2% with MoE-ViE, 4.2 points above the best existing detector. When format is decoupled from authenticity, it remains the best of all methods. Code will be released.

0.24

RetiDual-Select: Reliability-Aware Specialist-Generalist Routing for Retinal OCT under Image-Quality Shift

cs.CV ·2026-11-17 ·Zihao Han, Philippe De Wilde
Abstract · 4 figures, 4 tables, 42 references. Submitted to the IEEE for possible publication. Main article only; the separate Supplementary Informatio…

Adaptive fusion can repair a static ensemble's errors, but can also overturn correct decisions. RetiDual-Select combines retinal specialist RETFound with the image encoder of the medical vision-language model MedSigLIP, separating candidate fusion, routing safety and predictive review. Calibrated product-of-experts (PoE) fusion is the static reference, a small gate proposes an alternative for each image, a validation-frozen harm guard decides whether that candidate may replace the reference, and a separate score prioritizes review. Static PoE reached the highest or joint-highest clean accuracy on all three datasets. At a clean 2% accepted-error target the review policy accepted 94.3% of Kermany-4 images at 99.89% selective accuracy, and 92.4% of OCT-C8 at 98.65%. Both branches were correct on 95.0% and 91.8% of clean images; only 4.4% and 6.2% had exactly one correct branch. Across 15 prespecified synthetic conditions, adaptive gains were concentrated in severe contrast reduction, where this routing opportunity widened roughly ninefold; percentile rescaling largely removed the effect, linking it to compressed dynamic range. In post-hoc validation-only analyses, gain was higher when the gate followed the more reliable branch, and a frozen guard cut the worst-condition OCT-C8 loss from 1.39 to 0.36 percentage points. Adaptive routing was useful when image quality created a reliability opportunity and the gate followed it; otherwise the system retained Static PoE or deferred for review. The findings are bounded by the synthetic perturbations evaluated and do not establish robustness to natural acquisition or scanner shift.

0.24

Visual Harness Scales General VLMs to Medical Images

cs.CV ·2026-11-17 ·Shaowen Wan, Hanqi Jiang, Zihao Wu, Junhao Chen, Weihang You +4 more
Abstract

We show that an appropriate visual harness can substantially improve the performance of general vision-language models (VLMs) across diverse medical imaging tasks. We introduce \ours, a medical visual harness that equips general VLMs with interactive visual tools and systematic diagnostic evidence management. With \ours, general VLMs interactively explore medical images and progressively acquire diagnostic evidence, which is preserved in a lossless memory for subsequent retrieval and use. After submitting an initial answer, the VLM receives an assessment of potential omissions and the reliability of identified findings. This assessment guides further image inspection and measurement, followed by answer revision in an iterative agentic loop until sufficient evidence is obtained. Across 11 tasks and four imaging modalities, \ours improves performance on all tasks with Qwen3.8-27B, with notable gains in spleen measurement (63.3\% to 98.3\%), renal search (57.5\% to 71.2\%), and pulmonary nodule detection (0.0\% to 28.2\%). High-capability API-based VLMs achieve even higher performance with \ours. For example, GPT-6.1 Sol improves pulmonary nodule detection from 10.8\% to 69.2\% with \ours, demonstrating the complementary benefits of stronger model capabilities and an effective medical visual harness. Our results highlight the potential of visual harnesses as a new paradigm for scaling general VLMs to medical imaging tasks.

0.24

SlopeGuard: Towards Generalizable AI-Generated Image Detection via Radial Spectral Slope Guidance

cs.CV ·2026-11-17 ·Cheng Fu, Yiheng Li, Yang Yang, Wenzheng Liu
Abstract

AI-generated image detectors often overfit to generator-specific artifacts, hindering generalization to unseen synthesis pipelines. Existing frequency-based methods usually learn high-dimensional spectral patterns that may preserve generator-dependent variations. We explore Radial Spectral Slope, a compact descriptor of global spectral decay that captures transferable spectral trends while suppressing fine-grained variations. Our analysis shows that real and generated images exhibit systematic but generator-dependent slope differences, indicating that slope provides useful forensic evidence but is insufficient as a standalone decision criterion. This motivates combining spectral slope with semantic representations for more robust detection. Based on this insight, we propose SlopeGuard, which couples slope-aware data processing with slope guided feature learning to construct informative patch-based views, integrate Radial Spectral Slope with semantic features, and adapt prior preservation according to spectral atypicality. Extensive experiments demonstrate that SlopeGuard achieves state-of-the-art mean accuracies of 97.5% on AIGCDetectBench, 97.8% on GenImage, and 85.1% on AIGIBench, while maintaining competitive performance under blur, cropping, JPEG compression, and additive noise.

0.24

Segmentation-Guided Direct Capacity Estimation of Rooftop Photovoltaics from Aerial Imagery

cs.CV cs.AI ·2026-11-17 ·Guibin Chen, Emma Fan, Yuzhang Lin
Abstract

Distribution planners need installation-level inventories of rooftop photovoltaic (PV) capacity, but interconnection records may be incomplete or spatially aggregated. Conventional image-based estimators segment PV arrays, convert the mask to physical area, and multiply by an estimated power density, coupling capacity prediction to mask-derived area. We propose Direct Head, a segmentation-guided architecture that retains U-Net segmentation and soft-mask pooling while directly regressing nonnegative capacity from PV-region features and area descriptors. We evaluate three independent runs on 7,590 French rooftop installations under the original split and a department-based holdout. Relative to a parameter-matched area-density neural model, Direct Head reduces mean absolute error by 8.7% and 12.5%, and symmetric mean absolute percentage error by 14.0% and 20.7%, respectively. Comparisons with conventional estimators, controlled architecture variants, auxiliary-loss ablations, and representative cases show that segmentation provides useful spatial evidence without requiring capacity to be determined by an area-density product.

0.23

Multispectral-Polarization-Infrared Multi-Aperture Camouflaged Vehicle Detection

cs.CV ·2026-11-17 ·Shu Wang, Xiong Lin, Yuan Song, Junjie Wang, Feng Huang +1 more
dataset
Abstract · 14 pages, 10 figures

Camouflaged vehicle detection is challenging because camouflage suppresses discriminative cues across sensing modalities. Existing multimodal methods often require precise pixel-level registration and feature-level fusion, creating a trade-off between accuracy and latency. We propose a spectral-polarization-infrared tri-modal detection framework co-designed with a multi-aperture imaging setup to exploit complementary imaging physics. A consensus-guided decision-level fusion strategy combines cross-modal detections within bounded registration tolerances, reducing dependence on precise alignment while maintaining accuracy and real-time performance. We construct a tri-modal camouflaged vehicle dataset containing infrared, polarization, and spectral data. Experiments show that the proposed method achieves 82.7% mean Average Precision (mAP) with an end-to-end latency of 16 ms, outperforming representative feature-level fusion approaches. On the public MID-1K dataset, the framework generalizes to RGB-Depth-Infrared modalities without architectural modification and improves mAP@0.5 by 5.7% over state-of-the-art methods. These results demonstrate the potential of decision-level consensus fusion and multi-aperture sensing for robust and deployable camouflaged vehicle detection.

0.23

BATMoS: Budget-Aware Guidance Allocation for Efficient Video Diffusion

cs.CV ·2026-11-17 ·Zhixu Du, Kuo Yang, Jingwei Sun, Hai "Helen" Li, Yiran Chen
Abstract

The effectiveness of a video diffusion model's inference budget depends on where computation is spent along the denoising trajectory. Across eight equal-budget inference configurations, VBench total scores span 4.80 points, highlighting the importance of explicit allocation design. Our key insight is that guidance evaluations are not equally valuable across diffusion time: retaining the sampling updates while selectively reusing guidance can preserve generation quality with fewer denoiser evaluations. We introduce BATMoS (Budget Aware Temporal Mixture of Strategies), which routes between fresh classifier-free guidance and cached guidance using a single pretrained denoiser. A magnitude-based temporal rule assigns guidance reuse to low-magnitude blocks and reserves fresh guidance for the remaining blocks, with a single integer controlling the computation budget. The resulting schedule requires no additional router training or per-budget threshold search. Evaluated on CogVideoX-2B with 4,720 videos per configuration across all sixteen official VBench dimensions, BATMoS achieves the highest total score among the eight allocation policies and comes within 0.10 points of full classifier-free guidance using 20\% fewer branch evaluations. BATMoS also composes with feature reuse, reducing computation from 80 to 62 effective branch evaluations with a 0.11-point decrease in total score. These findings make temporal guidance allocation a practical mechanism for extracting more generation quality from a fixed inference budget.

0.23

Unified Residual Translation for Paired Image Restoration

cs.CV ·2026-11-17 ·Joyanta Jyoti Mondal, Mridul Banik, Ibne Farabi Shihab, Md. Shifatul Ahsan Apurba
Abstract · 10 pages, 2 figures

Paired image restoration spans changes in imaging contrast, sensor modality, and spatial resolution. These settings differ in their input channels and target grids, raising the question of how much of the prediction pipeline must be specialized to each task. We present the Unified Residual Translator (URT), which separates fixed channel and scale mapping from learned residual prediction. A deterministic reference uses channel averaging, replication, or bicubic interpolation to match the target dimensions, while a NAFNet encoder-decoder predicts an additive correction from the complete input. Separately trained instances share the prediction rule, $\ell_1$ objective, and optimizer across MRI contrast synthesis, visible-to-infrared translation, SAR-to-optical translation, and real-camera super-resolution. Across six task and scale settings, URT achieves the highest PSNR and SSIM among the evaluated baselines on IXI, SEN1-2, and RealSR $\times3$ and $\times4$. It reaches 30.774dB on IXI and 13.478dB on SEN1-2, with luminance PSNR of 33.596, 28.742, and 27.342dB at the three RealSR scales. On LLVIP, it ranks second in PSNR and SSIM. A study of four spectral alternatives identifies complementary benefits of fitted references and bandwise residual gains. Together, these results support explicit dimensional alignment and learned residual correction as a common formulation for heterogeneous paired imaging tasks.

0.22

ARG-Avatar: Animation-Rendering Guidance for Efficient Feedforward Avatar Reconstruction

cs.CV cs.GR ·2026-11-17 ·Jianjin Xu, Soham Dasgupta, Fernando De la Torre
benchmark
Abstract

Reconstructing animatable 3D human avatars from images is important for applications such as gaming and augmented reality. Feedforward methods make avatar reconstruction accessible by reducing inference time to a few seconds, but recent approaches rely heavily on scaling model capacity, training data, and computation. We ask whether such scaling is necessary. To this end, we introduce ARG-Avatar, an efficient feedforward framework with only 68M trainable parameters that achieves SOTA reconstruction quality. Our key insight is that 3D animation and rendering already provide strong geometric correspondences that existing methods largely attempt to learn from data. In light of this, we design the Foreground Avatar Coordinate RoPE attention (FACRoPE) and Intermediate Token Rendering (ITR) pipeline to guide avatar tokens to attend to their corresponding patches. Experiments across multiple datasets show that ARG-Avatar consistently outperforms prior methods when using two or more input images. Compared with LHM++, ARG-Avatar uses approximately 9$\times$ fewer trainable parameters, 4.8$\times$ fewer training sequences, and 14$\times$ less training compute, while running 6$\times$ faster and achieving a 2.7 dB PSNR improvement on the out-of-distribution NeuMan benchmark.

0.22

Motion-Aware Reward Modeling for Instruction-Guided Video Editing

cs.CV cs.AI ·2026-11-17 ·Li Li, Xiangbo Gao, Bo Ni, Samyadeep Basu, Viet Dac Lai +9 more
Abstract

Reward models for instruction-guided video editing can favor temporal consistency at the expense of motion. We study this failure, termed \emph{motion collapse}, using controlled motion-suppression variants of real edits. On VE-Bench, the evaluated composite reward prefers a motion-suppressed candidate in $97\%$ of cases. Learned video rewards and a VLM judge also struggle with mild suppression, while latent optimization and reward-guided fine-tuning exploit this weakness to increase reward as motion declines. A linear probe of our base reward backbone performs near chance on mild suppression, motivating an explicit inter-frame dynamics input. We propose a motion-aware reward model that combines an RGB frame-difference stream with preference training against motion-suppressed and fabricated-motion negatives. A distillation anchor preserves alignment with human quality judgments. On the VE-Bench motion ladder, our model reduces the rate of selecting motion-suppressed candidates to $0\%$ while achieving a Spearman correlation of $0.57$ with human quality ratings. It also detects unseen suppression variants and preserves motion during reward-guided fine-tuning on two editing pipelines. These results support explicitly modeling motion in rewards for video editing.

0.22

CutAgent: Creative-Gene-Driven Planning and Experience-Guided Self-Refinement for High-CTR E-commerce Advertising Video Creation

cs.CV ·2026-11-17 ·Xuyue Huang, Guanbo Huang, Jiasong Feng, Run Ling, Ao Ma +8 more
Abstract

E-commerce advertising video creation requires both high-level creative planning and reliable multi-modal production. However, existing methods often map product attributes directly to scripts or editing decisions without explicitly modeling advertising strategies, narrative structures, and production styles. They also have limited ability to reuse past production failures. To address these limitations, we propose $\textbf{CutAgent}$, a dual-agent framework that formulates advertising video creation as a Plan--Execute--Self-Evolve process. CutAgent comprises two collaborative agents, $\textbf{GeneAgent}$ and $\textbf{VideoAgent}$, connected by a structured creative gene that bridges product understanding and video production. GeneAgent maps multi-modal product inputs---textual attributes together with product images---to creative genes that specify narrative units and production rules. To train GeneAgent, we construct $\textbf{CutAgent-70K}$ with approximately 70K samples of product attributes/images, reverse-extracted creative genes, cleaned clip pools, and real-world CTR, and apply high-CTR supervised fine-tuning followed by CTR-guided GRPO. Guided by the generated gene, VideoAgent retrieves category-then-gene failure--correction experiences from a long-term memory as constraints for shot-script planning, then coordinates material retrieval, narration, subtitles, music, and timeline rendering. Independently, a reflection loop diagnoses quality defects, revises deficient production decisions, and writes verified corrections back to memory for later retrieval. Experiments on $\textbf{CutAgent-Eval}$ show that CutAgent outperforms representative baselines in e-commerce advertising video creation, achieving the highest online CTR of $\textbf{2.21\%}$.

0.22

Fast and Strong Image Protection against Diffusion Personalization

cs.CV cs.CR ·2026-11-17 ·Jiwon Park, Jihye Kang, Namhyuk Ahn
Abstract

Personalized text-to-image diffusion models can reproduce a person from a handful of photographs. Image protection perturbs each image before publication so that fine-tuning on it fails, but it optimizes every perturbation from scratch, so its cost grows with the number of images. We present the protective diffusion model (PDM), which is trained once and samples the perturbation of a new image over many denoising steps. PDM distills an iterative teacher and is then post-trained with the target residual projection (TRP) loss as a reward, a stronger objective than existing ones. PDM outperforms nine baselines on faces and artworks and nearly matches its teacher at a tenth of the cost.

0.21

Attribution Through the Decoder: Prediction-Space Explanations for Structured Outputs

cs.CV cs.AI ·2026-11-17 ·Zaer Bin Zaqy, Brinnae Bent
Abstract · Under review

Networks that predict angles, gaze directions, and rotations do not output these quantities directly: they emit raw coordinates that a fixed decoder, such as normalization, maps to the final prediction. Attribution methods are usually applied to the raw coordinates, so they can credit inputs for changes the decoder discards, such as rescaling a vector when only its direction matters. The resulting explanation can highlight inputs that have no effect on the prediction. We instead attribute through the decoder differential, which for normalization decoders reduces to a scaled tangent projection. We characterize invariance to output rescaling, show that every fixed smooth scalar target on a sphere has points of zero tangential sensitivity, and show that an endpoint-projected variant of Integrated Gradients can still attribute to scale-only inputs, whereas direct angular Integrated Gradients cannot. On synthetic tasks with analytically known relevance, ambient attribution assigns up to 72% of its mass to an irrelevant region. On MPIIGaze validation data, tangent Grad-CAM produces 44% greater prediction displacement under 10% pixel deletion. On PASCAL3D+ with a 6D rotation decoder, the full differential improves mean deletion displacement by 14.1 degrees over the ambient coordinate sum and by 13.8 degrees over a decoder-invariant scalar. Results for a two-polarity tangent LayerCAM variant are mixed: it exceeds ambient-sum LayerCAM on circular regression but no difference is detected on quaternion pose. These results separate two failure sources, radial contamination and scalarization loss. The approach uses existing attribution methods with decoder-aware targets, although some variants also require changes to attribution aggregation. An invariant attribution can still miss inputs that drive the prediction, so invariance and deletion performance should each be checked.

0.21

From Perception to Simulation: Unveiling and Empowering Mental Spatial Dynamics in MLLMs

cs.CV cs.AIcs.CL ·2026-11-17 ·Jiachun Li, Zhuoran Jin, Kang Liu, Jun Zhao
publishedbenchmark
Abstract · Findings of EMNLP 2026

Spatial reasoning is a fundamental component of world understanding for Multimodal Large Language Models (MLLMs). While existing research has advanced the perception of spatial relations in static scenes, the domain of dynamic spatial simulation remains relatively unexplored. To bridge this gap, we introduce MentalGrid, a benchmark focused on spatial reasoning tasks that require models to simulate evolving spatial states without intermediate visual feedback. Extensive evaluation on our benchmark reveals a significant deficiency in the spatial simulation capabilities of existing models. Specifically, we find that performance is constrained by perceptual limitations in simple simulation scenarios and hindered by inadequate multi-step reasoning as complexity increases. To address these challenges, we propose a three-stage curriculum learning framework utilizing 16k specialized spatial reasoning samples. This approach yields SpatialWalker, which achieves a 33.8% average improvement over its base model, surpassing proprietary models like Gemini-3-Flash in spatial simulation. Furthermore, SpatialWalker demonstrates robust generalization, maintaining a 4.9% average gain on out-of-domain benchmarks, demonstrating the pivotal role of spatial simulation as a fundamental building block for broader spatial intelligence.

0.21

Reading Scene States: Uncovering Safety Failures in Vision-Language Models

cs.CV cs.AI ·2026-11-17 ·Tongxi Wu, Yuyang Cheng, Jiacheng Xian, Menglin Yang, Kai Chen +2 more
Abstract · VLM safety, multimodal jailbreaks, and scene-state reasoning

Vision-language models (VLMs) can provide harmful assistance by interpreting visual events, even without an explicit request for the underlying harmful objective. We identify this \textbf{scene-state safety vulnerability} and introduce \textbf{SceneSafetyBench}, comprising 375 image--question pairs spanning Initial, Medium, and Ultimate scene states and five visual presentations. Across eight VLMs, scene-state inputs elicit harmful assistance with attack success rates (ASRs) of 45.6--72.3\%. Comparative analysis across three VLMs shows that scene-state inputs most closely resemble QR-related images in their layer-wise safety-related signal profiles, while exhibiting lower average late-layer readouts and distinct signal dynamics during generation. To turn this vulnerability into a systematic attack, we propose \textbf{S3TA}, which translates harmful objectives into scene states, queries image--question pairs independently, and assembles the responses. On seven MM-SafetyBench categories across eight open-weight and commercial VLMs, S3TA achieves a model-average ASR of \textbf{71.2\%, exceeding the strongest baseline by 6.1 percentage points}. Text-side filters and visual moderation provide limited protection. These findings highlight the need to safeguard the assistance models infer through scene-grounded reasoning, beyond screening input content.

0.21

CoBoVLA: Decoupling Language from Action with Discrete Codes for Robust and Interpretable Robot Manipulation

cs.CV cs.RO ·2026-11-17 ·Mehul Arora, Dhruv Agarwal, Ambuj Agrawal
benchmark
Abstract · 16 pages, 3 figures, 4 tables, and supplemental material included

Vision-Language-Action (VLA) models achieve high success rates on the LIBERO benchmark but remain sensitive to paraphrasing. Existing VLAs ranging from 0.6B to 7.5B parameters experience a 22-52 point performance drop on LIBERO-Para while SmolVLA exhibits an even larger decline, from 83\% to 8\%. Because most VLAs jointly learn language understanding and action generation, it is difficult to determine whether a failure originates from misunderstanding an instruction or executing it incorrectly. To address these challenges, we introduce CoBoVLA, or CodeBookVLA, a 250M-parameter VLA that decouples language understanding, planning, and control through learned discrete interpretable codes. A text encoder maps the instructions to shared step codes, which a visual planner then converts into action code sequences executed by a language-independent actor. This design improves robustness to instruction paraphrasing while allowing failures to be traced to individual stages. It also enables efficient real-time control on resource-constrained hardware by encoding instructions only once and using compact discrete representations to keep each component lightweight. CoBoVLA achieves 91.4\% success on LIBERO-Goal, compared with 82.8\% for SmolVLA, under a matched protocol, with 17 ms per control step latency. On LIBERO-Para, it achieves 81.1\% across 4,092 unseen paraphrases and 78.5\% on matched episodes, compared with 36.5\% for SmolVLA LoRA fine-tuned on paraphrases. Without two-step demonstrations, CoBoVLA completes 48.8\% of 160 free-form two-step tasks. Finally, success reaches 91\% when instructions are correctly encoded but falls to 27.2\% otherwise, identifying language understanding as the main source of remaining failures. Together, these results demonstrate that decoupling language understanding and control can improve instruction robustness while making VLA failures easier to diagnose.

0.21

EgoHandWAM: Probing the Role of Hand Motion in Egocentric World Action Modeling

cs.CV cs.AI ·2026-11-17 ·Kailin Wang, Haoxiang Jie, Yaoyuan Yan, Zhiyou Heng, Zhaosong Li
Abstract

How much does hand motion tell us about what a first-person camera will see next? EgoHandWAM forecasts three-dimensional wrist and fingertip trajectories and a future visual feature from the current image and recent hand states. We train on 751,065 EgoDex windows using a frozen DINOv2 encoder. On held-out development sessions, the action-only model reduces two-second wrist endpoint error from 60.26 to 56.48~mm relative to a stationary forecast. It still underestimates long reaches. Adding future-feature supervision gives nearly identical trajectory accuracy. With the recorded hand history, the visual head reduces smooth L1 error by 1.32\% relative to copying the current feature. Shuffling histories across sessions erases this gain. In a separate four-task probe, adding recorded future displacements to observed context and future tracking-validity masks reduces visual-feature error by 3.93\%. Masks alone give little improvement. A trajectory predicted from the same observations adds little. On two simulated RoboTwin tasks, substituting aligned human examples for an auxiliary batch of robot examples gives no consistent increase in task success across three training seeds.

0.21

Training-Free Building Change Detection with Example-Induced Discriminants

cs.CV ·2026-11-17 ·Zhi Lu, Bin Hong, Zhenming Peng
code/data
Abstract

Building change detection (BCD) relies on dense annotations that are costly to collect across regions. A few annotated examples offer a more economical way to specify the target, but change maps provide incomplete building supervision: they label only changed buildings and leave the building-containing date unspecified. Our key observation is that these annotations nevertheless reveal features that recognize buildings in other images, including unchanged ones. We present Linear INduced Discriminants (LIND), a training-free framework that turns regional feature contrasts into a reusable linear building classifier. Using a single frozen vision transformer, it unifies building recognition and temporal comparison in the same feature space. This approach requires neither text prompts nor an auxiliary segmentation model. Only one parameter is tuned on a single validation dataset and then kept fixed across datasets and shot settings. Extensive experiments demonstrate strong detection accuracy, cross-dataset transfer, and efficient inference, with extensions to other semantic classes and broader land-cover changes. Change annotations thus provide a compact visual specification for detection in new scenes. Code: https://github.com/lzcn/lind

0.20

Transfer Once, Ask Locally: Compact Audio-Visual Memory for On-Device Question Answering

cs.CV ·2026-11-17 ·Qitong Wang, Jihui Jin, Divya Kothandaraman, Vishnu Raj, Christopher Rasmussen
Abstract

Audio-visual question answering on lightweight user devices requires balancing speed, memory usage, privacy, and communication. One approach is to run audio-visual encoders on a server and a lightweight language decoder on the user device, keeping questions local while offloading media encoding. However, transferring and processing the full multimodal representation remains costly. We propose a training-free, query-agnostic token pruning framework that constructs a compact, reusable memory before transmission. We define video spatio-temporal blocks and aligned audio windows as players and estimate their importance through Monte Carlo Shapley valuation. We use an embedding coverage objective that combines query-independent utility with cross-modal redundancy calibration and rewards unimodal diversity through marginal coverage gains. We retain the original token embeddings of high-scoring players under a limited token budget and transfer the resulting memory once for reuse across subsequent local questions. Experiments on Video-MME, Daily-Omni, and Video-Holmes using video-SALMONN 2+ (3B) and video-SALMONN 2 Pro (4B) show that our method achieves the best or tied-best QA accuracy among the compared compression methods across all evaluated settings. On Daily-Omni with video-SALMONN 2+ (3B) at 80\% pruning, our method improves accuracy by 12.86--14.62 percentage points over these methods while achieving the lowest device-side LLM prefill and decoding latency, with up to $1.92\times$ decoding speedup at comparable embedding memory usage.

0.20

GameRealm: Grounding Visual Learning in Interactive Games

cs.CV ·2026-11-17 ·Weiguang Zhao, Yifan Wang, Ni Yang, Yating Wang, Xingyu Miao +7 more
Abstract

Visual foundation models are central to physical intelligence, yet their development remains constrained by fragmented training datasets, inconsistent annotation quality, and limited multimodal supervision. To address these challenges, we introduce GameRealm, a large-scale, richly annotated dataset that grounds visual observations in synchronized, game-native supervision. GameRealm encompasses over 120 million RGB frames and 1,000 hours of gameplay footage across all views. Beyond its unprecedented scale, GameRealm provides rich multimodal supervision, including metric depth, camera parameters, synchronized audio, control inputs, and 3D point trajectories. Its diversity further extends to flexible view configurations, spanning single-view, stereo, multi-view, and panoramic recordings. To enable scalable acquisition across diverse game environments, we develop a unified data collection pipeline that combines human-operated interaction with agentic exploration. Extensive experiments across diverse visual learning tasks demonstrate that GameRealm consistently strengthens visual foundation models in both visual generation and geometric learning. These results establish GameRealm as a versatile and valuable data resource, unlocking new opportunities for advancing visual intelligence.

0.19

Where Demand Meets Supply: Spatial Elastic Consensus for Low-Bandwidth Collaborative Perception

cs.CV ·2026-11-17 ·Linshan Yan
Abstract · 10 pages

Occlusions and limited sensor range make ego-only 3D perception insufficient for safe driving. Collaborative perception addresses this by enabling connected vehicles to share complementary observations, yet it remains constrained by tight vehicular communication budgets. While spatial selection is more flexible than fixed-rate feature compression, existing selectors remain largely sender-driven and waste scarce bandwidth on ego-visible targets. Furthermore, applying convolutions after masking spreads isolated tokens into their neighborhoods, silently violating low-bandwidth constraints. To address this, we propose Spatial Elastic Consensus (SEC), a bandwidth-adaptive framework centered on bilateral demand-supply consensus. Instead of unilateral token selection, the ego vehicle formulates a content-aware demand prior while each collaborator evaluates its local supply. This single-round consensus allocates the token budget to critical ego-missed evidence rather than ego-visible redundancy, without transmitting dense maps. To guarantee exact budget compliance, an anti-dilation refine-before-mask pipeline refines features prior to discrete token selection, preventing receptive-field leakage. A lightweight deformable sampler then aggregates these sparse messages with learned local offsets, compensating for the spatial context lost to token selection and mitigating inter-agent pose drift. Evaluated at a stringent 64KiB budget, SEC consistently achieves the best AP@0.7 on the simulated OPV2V and V2XSet datasets and the real-world DAIR-V2X dataset. These results demonstrate that combining bilateral consensus with budget-exact transmission is key to efficient low-bandwidth collaborative perception.

0.19

Event-Based Dynamic Lip Region Estimation During Speech

cs.CV cs.LG ·2026-11-17 ·Hamze Hammami, Omar Eldardeer, Francesco Rea
Abstract

Event cameras offer high temporal resolution and low power consumption, making them promising for speech-related applications. These include automatic lip reading, audiovisual speech analysis, and human-robot interaction, particularly in challenging environments characterized by low illumination or high dynamic range. Event-based lip reading, speaker recognition, and voice activity detection, however, mostly operate on a cropped region of interest (ROI), by hand or by the dataset, and no event-based method supplies that ROI as a spatial region; existing event-based face analysis gives the mouth as landmarks, which presents a mismatch between application and localization. The present work predicts a dense heatmap over the lip region from the full event "frame". Events are encoded as a polarity-separated voxel grid and processed at every location by a two-pathway continuous-time recurrent network (CT-RNN) with learned, log-space bounded time constants acting on the real elapsed time of each bin: a stateless fast pathway and a stateful slow pathway, utilised for refining and anchoring the prediction. The model reaches 69.5% on lips on four unseen speakers and switches target only 2.1 times per second, against 8.9 and 11.0 for convolutional baselines with more parameters. Adding a spatial encoder improves precision but fails on a new face scale. On a real event sensor, trained only on simulated events, the proposed model reaches 58.0% on lips against 43.2% for the CNN, with fewer than half its target switches.

0.19

GaussianMamba: Geometric and Relational State Space Scanning of 3D Gaussian Primitives

cs.CV ·2026-11-17 ·A. Enes Doruk, Ozgur Erkent
Abstract

3D semantic occupancy prediction requires representing complex spatial structures and modeling contextual relationships across a scene. 3D Gaussian primitives provide a sparse and adaptive representation of scene geometry and semantics, while Mamba offers efficient sequence modeling for capturing dependencies among primitives. However, Gaussian primitives are irregular and unordered, making sequence construction challenging, as conventional spatial ordering may overlook their anisotropic geometry and feature-semantic relationships. In this paper, we present GaussianMamba, a dual-branch Mamba framework that explicitly incorporates Gaussian-specific geometric and feature-semantic relationships into complementary scanning strategies. The first branch refines a locality-preserving traversal using anisotropic Gaussian support derived from primitive scale and orientation. The second branch reorders spatially organized groups according to feature-semantic affinity and adjusts local features using geometry-weighted semantic differences. The resulting sequences are processed by separate Mamba modules, and their features are fused through a support-aware mechanism that incorporates local feature discrepancies arising from the distinct scanning paths. Experiments on nuScenes-SurroundOcc and SSCBench-KITTI-360 demonstrate consistent improvements over GaussianFormer and GaussianFormer-2 in both IoU and mIoU.

0.19

PURE: A Diagnostic Benchmark for Multi-Granularity Multimodal Identity Matching

cs.CV ·2026-11-17 ·Hengyu You, Xintong Wang, Jin Cao, Liang Ding
benchmark
Abstract

Identity matching asks whether two records denote the same identity, yet the answer depends on the target granularity. Two records may match at the base level but not at the variant level: they share an underlying design while differing in attributes such as color or size. Final matching accuracy does not separately characterize the capabilities involved. To study this problem, we use e-commerce as a challenging real-world domain. We introduce PURE, a diagnostic benchmark for multi-granularity multimodal identity matching covering four capability dimensions: Perception, Understanding, Reasoning, and Evidence-grounded Judgment. The benchmark comprises 1,457 multiple-choice questions across 14 tasks, derived from real-world matching pairs and verified by two domain experts. Across ten recent MLLMs, overall scores range from 47.1% to 86.0%. Understanding receives the highest mean score and Judgment the lowest, with weaknesses in visual difference localization and variant-level rule judgment. Modality ablations show that models benefit from multimodal evidence but can incorrectly identify conflicts between matching records. In direct matching, most models recognize matching pairs more reliably than non-matching pairs at the base level, whereas the reverse holds at the variant level. PURE therefore reveals what aggregate matching accuracy hides: not only whether models judge a pair correctly, but where they struggle to interpret evidence and apply matching criteria.

0.18

Keeping Parts Honest: Isolated Part Graphs for Faithful Vision Transformer Explanations

cs.CV ·2026-11-17 ·Bernard Tomczyk, Gail Collyer-Hoar, Bryan Williams
Abstract · 12 pages, 6 figures, 6 tables. Supplementary material will be added in a later version

Part and prototype-based explanations for Vision Transformers (ViTs) present a decision as evidence from individual image regions. However, self-attention mixes information across patches, so part-level features absorb context from the whole image and consequently fail to faithfully reflect the decision. We show that existing interpretable ViT classifiers depend heavily on this leakage: when each explained region is encoded in isolation, their accuracy collapses, and retraining them under isolation recovers only part of the lost accuracy. We propose a classifier that represents each image as a graph of isolated parts. Each part is encoded only from its own patches by a ViT whose last blocks are fine-tuned under isolation, preventing evidence leakage from elsewhere in the image. Parts are discovered without annotation and refined hierarchically through their sub-parts, making them more consistent and stable across images than those of competing methods. Each edge is computed from part pairs, and recovers evidence that attention previously shared across the object without affecting the parts. Matching the image graph to a learned prototype graph per class makes the prediction a direct combination of node and edge similarities, each computed only from the patches it is credited to, so every explanation is faithful by construction. Across 11 fine-grained and general-object benchmarks, our method exceeds the accuracy of existing explainable methods that rely on leakage, while providing explanations that reflect what the model actually uses.

0.18

NexSight: Verifiable Tasks for Fine-Grained Agentic Long-Video Understanding

cs.CV cs.AI ·2026-11-17 ·Tianci Xie, Li Chen, Pinxue Guo, Rui Zheng
Abstract · 25 pages, including references and supplementary material

Fine-grained questions about long videos ask how many times an action occurs, in what order events unfold, or exactly when a state changes. Answering them can require finding brief moments scattered across hours of footage. Agentic models that decide when and where to look promise to find these moments, yet this ability is hard both to measure and to learn. Sparse supporting evidence may establish that an action occurred without establishing its total count; verifying timeline-wide claims requires checking for missed occurrences and double counting. Answer-only supervision, in turn, does not explicitly supervise the search for evidence. We present NexSight, which builds verifiable tasks to both evaluate and train agentic long-video understanding. Its agentic pipeline, SightForge, accepts a task only if a verifier blind to the draft answer can reconstruct it from evidence densely reacquired from the raw video and counterexample checks fail to refute it. This yields SightLine, 310 evidence-traceable tasks from 108 long videos, evaluated through SightLoop, a shared multiscale harness supporting temporal exploration, local inspection, evidence integration, and revisiting while recording returned observations. On disjoint videos, solver trajectories on verified tasks become step-level supervision for observation decisions. The strongest model reaches only 54.52% on SightLine, yet 79.35% when given the annotated evidence. With under 6K examples, Sight-35B nearly doubles its base model's interactive accuracy (21.29% to 42.26%), outperforms the 397B model of its family, begins using explicit spatial inspection after post-training, which its base model never uses, and gains over 8 points on two external long-video benchmarks.

0.18

Leveraging Segmentation Representations for Efficient Classification: A Self-Configuring Framework Based on nnU-Net

cs.CV cs.AI ·2026-11-17 ·Renjun Cai, Antonio Javier Gallego, Félix Fuentes-Hurtado
benchmark
Abstract

We propose a self-configuring framework that extends the nnU-Net workflow to both segmentation and classification in 3D medical imaging. Building on nnU-Net's self-configuring method, the framework reuses representations learned by a segmentation model for image-level classification. Its encoder features and predicted segmentation probabilities are used by a lightweight classifier. Two information-sharing strategies are investigated. In the frozen-backbone branch, features are augmented with region-of-interest information, predicted foreground volume, and segmentation confidence. In the adaptation branch, the final encoder stage is partially fine-tuned for classification, with an auxiliary segmentation loss to preserve segmentation-related information. Additional five-fold experiments showed that more complex configurations did not consistently improve classification performance. On the public validation leaderboard, our selected framework achieved an Overall score of 0.8056. In the official FLARE 2026 Task 2 final evaluation, our framework ranked first overall, achieving the highest DSC, NSD, and AUROC across all four testing tasks, including two hidden testing tasks. Local inference measurements also remained below the challenge limits of $60 \mathrm{s}$ per case and $4 \mathrm{GB}$ GPU memory. These results demonstrate the practical value of reusing segmentation representations for classification while maintaining a simple and efficient framework.

0.18

SpatialCast: Reading the 3D World via Cast Shadows

cs.CV ·2026-11-17 ·Enhan Li, Jinghao Zhou, Jiaze Zhang, Hongyang Du
benchmark
Abstract · 10 pages (main paper), 7 pages (supplementary material), 5 figures and 6 tables in the main paper

As multimodal large language models (MLLMs) advance toward understanding and interacting with the physical world, spatial reasoning becomes essential for understanding scene structure and anticipating the effects of physical changes. Existing spatial benchmarks have made substantial progress in evaluating spatial relations, memory, and viewpoint reasoning, but less systematically assess how models connect physical scene conditions with visual observations. We introduce SpatialCast, a benchmark that uses cast shadows to evaluate this ability. Cast shadows provide observable evidence of a 3D scene. Their image appearance depends on object geometry and pose, illumination, and viewpoint. We examine this relationship through three complementary tasks that assess scene-projection consistency, predict intervention-induced changes, and infer the physical changes behind visual transitions. The benchmark contains 21,000 synthetic questions with separate scene and geometry test partitions, complemented by 200 SpatialCast-Real questions from 20 photographed scenes. Evaluations show that closed-source MLLMs outperform general-purpose open-weight MLLMs and MLLMs designed for spatial reasoning, which remain near chance on the synthetic benchmark, while even strong closed-source models struggle with shadow-length and light-height judgments. We also use SpatialCast training data for GRPO post-training, improving open-weight MLLM performance on the benchmark. Further evaluations show improvements on several external spatial benchmarks, suggesting transfer of the learned capabilities. These experiments support SpatialCast as a controlled resource for diagnosing spatial reasoning capabilities and studying their learning and generalization.

0.18

Hierarchical Diffusion Transformer for End-to-End Image Generation

cs.CV cs.AIcs.LG ·2026-11-17 ·Xiaoyu Wu, Yifei Wang, Weihang Guo, Chen Wei, Lydia E Kavraki +2 more
Abstract · Under Review

Pixel-space diffusion learns visual representations and image generation jointly in a single end-to-end training stage, without a separately pretrained visual encoder. However, plain pixel-space Transformers still trail leading two-stage latent diffusion models in generation quality and training efficiency. We hypothesize that part of this gap arises from retaining one token per patch throughout the backbone, which does not explicitly encourage a visual hierarchy that summarizes related content at progressively coarser levels. To address this, we introduce Hierarchical Diffusion Transformers (HiT), which progressively group related tokens and represent each group with a single token. The main Transformer bottleneck processes these compact group tokens, and ungrouping restores features to their original image positions. A compression budget regulates the average compression ratio across the batch, allowing group memberships and counts to vary across images and diffusion timesteps during both training and generation. The model exhibits emergent semantic structure, learning to group coherent objects and object parts without segmentation labels. In text-to-image generation, HiT further learns to group caption words with image patches depicting the same concepts, without region annotations. HiT achieves 1.48 FID on ImageNet $256\times256$ and reaches comparable generation quality to JiT using approximately 40\% of its cumulative training FLOPs. For text-to-image generation, HiT exceeds UniLiP's average score across five benchmarks with roughly $6\times$ fewer denoiser parameters and $3\times$ fewer forward FLOPs. These results suggest that learning a visual hierarchy within the generator breaks down the barrier to stronger image generation without external visual pretraining.

0.18

VSpark: Speculative Decoding for Autoregressive Vision-Language-Action Models

cs.CV cs.RO ·2026-11-17 ·Suho Yoo, Youngjoon Jang, Hyebin Cho, Taeyeop Lee, Jaewook Jeong +1 more
Abstract · Preprint

Our aim is to accelerate autoregressive vision-language-action models (VLAs), which generate robot actions one token at a time. Speculative decoding makes this possible: a lightweight draft model proposes several tokens at once, and the VLA verifies them in a single pass, accepting only those that match its own predictions, so decoding remains lossless. The speedup, however, is limited in robot control, where many different tokens encode nearly the same motion and a draft close to the correct action is still rejected. To address this, our approach, VSpark, is built upon three design choices: (i) rollouts collected with calibrated action noise, which expose the draft to a wider range of robot states; (ii) feature alignment, which trains the draft to match the VLA's hidden features rather than its output tokens alone; and (iii) continuation reuse, which keeps the predictions the VLA has already computed along rejected drafts as candidates for the next verification step, instead of discarding them. On four VLAs with different action tokenizers and four manipulation benchmarks, VSpark achieves 1.28x to 2.10x speedup while keeping the VLA's decoding lossless.

0.18

Evolve in Panorama, Render in View: Panoramic World State Evolution for Video World Models

cs.CV ·2026-11-17 ·Longtao Jiang, Mingfei Han, Jianmin Bao, Tingjun Dai, Jinjie Wei +5 more
Abstract

Video world models should maintain a world that keeps evolving as the viewpoint changes. Yet when the camera turns away and returns, current models often freeze the unobserved scene, lose objects, or change their appearance, because content outside the view has no representation that keeps changing. We present PanoEvo, which uses a Panoramic Evolver to evolve the world state in the latent space of a panoramic video diffusion model, where every viewing direction is present at every time step, so the scene evolves whether or not it is observed. Existing panoramic video models are not trained to evolve a scene from a single image or the objects within it, so we build PanoEvo-Data, an event-centric dataset from real 360° videos, and train the Panoramic Evolver coarse-to-fine as an image-to-panoramic-video model. Since each view covers only a small part of the low-resolution panorama, directly projected views are blurry and often show low-quality details or structural errors. A Perspective Renderer, built on a perspective video diffusion model, reads this state via latent-space projection and direction-restricted cross-attention to render sharp, high-quality views. PanoEvo achieves the best overall score among 23 publicly evaluated models on WRBench and the highest total score on STEVO-Bench. Code, models, and data will be released.

0.17

MI-Hyper: Generating Multiple Instance Models for Weight Space Transfer Learning in Histopathology via Mixed Tasks Pretrained HyperNetwork

cs.CV ·2026-11-17 ·Xitong Ling, Yousef Kotp, Weiming Chen, Yining Li, Jiawen Li +7 more
Abstract

Zero-shot whole-slide image (WSI) prediction aims to generalize to new pathology tasks without parameter updates at deployment. MI-Zero makes these predictions by pooling patch-text similarities. Conditional weight generation, combined with supervision from diverse pathology tasks, offers an alternative: learning to construct predictive rules from task descriptions. We introduce MI-Hyper, a hypernetwork pretrained across mixed tasks to generate the aggregation and classification weights of multiple instance learning (MIL) models. We first train the generator and a shared MIL basis jointly using source labels. We then freeze the basis and refine the generator. The generated task weights are reused across cases. Across 308 tasks from 52 pathology datasets, MI-Hyper improves mean AUC, balanced accuracy and macro-F1 over MI-Zero by 1.85, 1.17 and 0.57 percentage points, respectively. On some molecular prediction tasks, both AUC and balanced accuracy improve by more than 20 points. These results support task-conditioned weight generation as a promising approach to weight-space transfer for long-sequence modeling.

0.17

Harnessing Image Diffusion Prior for Photo-Realistic Video Restoration

cs.CV ·2026-11-17 ·Fanghua Yu, Hongyu An, Jinfan Hu, Xinqi Lin, Zhiyuan You +4 more
published
Abstract · 40th Conference on Neural Information Processing Systems (NeurIPS 2026)

Recent image generation models produce photo-realistic results with rich details, but transferring their powerful priors to video restoration remains difficult due to temporal inconsistency from stochastic detail synthesis. Existing video restoration methods improve temporal coherence but often sacrifice visual fidelity, leaving a gap between image-level quality and video-level stability. We propose a generative video restoration method that bridges this gap by enabling temporally consistent detail synthesis from image generation priors. Our method combines noise energy rebalancing to suppress structure-disturbing low-frequency stochasticity, motion-aligned noise warping to propagate details along object motion, a self-supervised denoising encoder to mitigate encoder-induced temporal uncertainty, and decoupled temporal attention to model cross-frame dependencies. Together, these components preserve fidelity, enhance fine-grained realism, and maintain temporal consistency. Extensive experiments show that our method substantially outperforms existing video restoration approaches in visual quality, detail richness, and temporal stability.

0.17

EviSlide-MIL: Multi-Query Pooling and Shared Prediction for Class-Incremental Whole-Slide Image Classification

cs.CV ·2026-11-17 ·Zhouhan Qian, Sicheng Chen, Chad Wong, Tianyi Zhang, Enhui Chai +2 more
Abstract

Histopathological analysis of whole-slide images (WSIs) plays a critical role in cancer diagnosis, and multiple instance learning (MIL) is widely adopted for WSI analysis under weak supervision. As new tumor subtypes are introduced, extending existing MIL models requires learning new classes while retaining knowledge of previously learned ones. This presents a class-incremental learning challenge, where models must acquire new classes with limited access to historical training data and recognize all learned classes without task identifiers at inference. A major obstacle is forgetting, as updates to feature aggregation and prediction modules can degrade performance on previously learned classes. To address this challenge, we propose \textbf{EviSlide-MIL}, a class-incremental WSI framework that integrates multiple learnable queries per class with a shared prediction module. The queries aggregate frozen patch features into class-specific representations, while the shared prediction module limits parameter growth across incremental stages. By freezing queries associated with previously learned classes, EviSlide-MIL preserves their attention patterns for identical inputs, while experience replay and knowledge distillation further mitigate forgetting as shared parameters are updated. Experiments on eight tumor subtypes from The Cancer Genome Atlas demonstrate that EviSlide-MIL achieves improved performance and reduced forgetting compared with competing methods across most learning settings, demonstrating its potential for scalable and knowledge-preserving WSI classification. Code and pretrained models will be released.

0.17

Text-Guided Modality Correction for Open-Vocabulary Object Detection

cs.CV ·2026-11-17 ·Yu Quan, Dong Zhang, Ping Gong, Peng Huang, JinHui Tang
Abstract

Open-vocabulary object detection (OVD) is essential for detecting objects beyond the categories seen during training. Establishing vision-language associations is crucial for recognizing unseen categories, while integrating complementary visual features across levels supports detection at different scales. However, existing methods often provide insufficient textual guidance for object regions, rely on limited cross-modal alignment strategies, and underuse complementary features across levels, limiting their ability to detect unseen categories and objects of varying scales. To address these limitations, we propose a Text-Guided Modality Correction Network (TMCNet) that incorporates textual guidance into region feature extraction, cross-modal alignment, and multi-scale decoding to strengthen region-category semantic correspondence and promote complementary feature fusion across levels. First, a text-guided modality encoder incorporates textual semantics into queries to guide regional feature aggregation, reducing semantic ambiguity in object representations. Second, a cross-modal alignment module strengthens the correspondence between region representations and category semantics through interactions between visual and textual features. Finally, a progressive text correction decoder combines a top-down pathway with lateral connections to progressively fuse text-guided deep semantic features with shallow visual features at the corresponding scales, preserving both semantic information and spatial detail for objects of different sizes. Experiments on two public datasets, COCO-OVD and LVIS-OVD, demonstrate that TMCNet achieves strong performance in open-vocabulary object detection using simple text queries as guidance. Extending TMCNet to open-vocabulary segmentation also yields performance gains, demonstrating its cross-task applicability and extensibility.

0.17

BRepA: Graph-Based B-Rep Pretraining for Unified CAD Understanding and Generation

cs.CV cs.GR ·2026-11-17 ·Dekai Zhu, Zhaiyu Chen, Stefan Gavranovic, Slobodan Ilic
Abstract

Boundary representation (B-Rep) of CAD models encodes geometric primitives and their relations as a graph structure, enabling efficient shape modeling within CAD software. Learning on B-Reps for downstream tasks, such as segmentation and novel shape generation, requires specialized neural network designs. Moreover, performance improvements and accelerated training can be achieved with effective pretraining strategies, which is particularly important for tasks with limited labeled B-Rep data. In this paper, we introduce BRepA, a unified self-supervised pretraining framework that explicitly reconstructs both the geometry and topology of B-Rep objects via graph-based modeling. By learning from large-scale unlabeled CAD data only once, BRepA produces representations that generalize across tasks and significantly enhance both discriminative and generative learning. On segmentation tasks, BRepA demonstrates remarkable data efficiency. For instance, with only 1% labeled samples, BRepA pretraining improves mIoU by 32.52% compared to models trained from scratch. On generative tasks, aligning diffusion model representations with pretrained BRepA features substantially accelerates convergence, achieving nearly 2 times faster training while improving generation quality under the same computational budget, despite the diffusion model being roughly 20 times larger than BRepA. Overall, our results show that geometry- and topology-aware B-Rep pretraining provides a powerful foundation for both understanding and generating CAD structures, serving as a versatile representation backbone for diverse downstream applications. We will release the code to facilitate future research.

0.16

Learning Grounded Reasoning from Human Gaze

cs.CV ·2026-11-17 ·Gabriel Sarch, Declan Campbell, Nicholas Budny, Kia Ghods, Katerina Fragkiadaki +3 more
Abstract

Human visual reasoning is driven by systematic exploration and extraction of task-relevant visual evidence. Human gaze provides a window into this process, making gaze a useful source of guidance for visually-grounded reasoning in vision-language models (VLMs). However, human gaze also includes noise and redundancies that may transfer poorly to VLMs. In this work, we investigate how human gaze data can be effectively leveraged to improve visual reasoning in VLMs, while maintaining alignment with human behavior. We collect 7,109 human gaze trials from a diverse and challenging set of visual reasoning tasks, and train grounded reasoning models with different forms of gaze guidance. Across evaluations in chart understanding, visual search, and counting, we find that using human gaze as a prior for grounded reasoning consistently outperforms supervision using the full gaze sequence. Based on these findings, we introduce GGR (Gaze-Guided Reasoner), which achieves greater human alignment and more reliable visual grounding while matching or outperforming leading grounded-reasoning baselines in accuracy. Reinforcement learning further improves task accuracy, gaze alignment, and grounding quality despite using no gaze supervision. Behavioral analyses show that gaze supervised models exhibit more visual exploration before arriving at an answer. Our results suggest that the value of human gaze lies in guiding spatial exploration and shaping reasoning behavior without requiring complete imitation of human behavior.

0.16

Semantic-Guided Geometric Decoding for Token-Efficient Visual Grounding

cs.CV ·2026-11-17 ·Zhaojie Chen, Junyan Yang, Le Yang, Qi Dang, Yucheng Huang +4 more
Abstract · 10 pages, 4 figures, 5 tables

Multimodal large language models (MLLMs) provide a unified interface for language-conditioned visual perception. In pixel-to-sequence grounding, autoregressive coordinate generation requires multiple sequential decoding steps per box, increasing inference cost with the number of instances. Pixel-to-embedding methods reduce this overhead but lack explicit spatial encoding when using a shared special token. We introduce SemGeo, a token-efficient visual grounding framework that represents each instance with a single grid token for coarse localization. A geometric decoder reuses visual features and hidden states to recover precise discrete box coordinates in parallel. To train proposal generation and box localization together, we introduce Gaussian soft labels that encode coordinate distances during supervised fine-tuning (SFT). Reinforcement post-training further combines box-set rewards with coordinate supervision on model-generated proposals. This stage trains the language model and geometric decoder on their own predictions to mitigate the mismatch between teacher-forced training and inference. Across eight benchmarks spanning object detection, referring object detection, and document layout grounding, SemGeo achieves the highest F1@mIou among compared MLLMs on six benchmarks. It also delivers $2.07\times$ and $1.25\times$ the mean box-generation throughput of representative grounding baselines under vLLM and Transformers, respectively, while achieving higher accuracy in both comparisons.

0.16

When AUC Meets Generative Reconstruction: A Unified Principle for Balanced Visual Representation Learning

cs.CV cs.AIcs.LG ·2026-11-17 ·Boyu Han, Qianqian Xu, Shilong Bao, Zhiyong Yang, Qingming Huang
Abstract

Multimodal large language models (MLLMs) require visual encoders that extract fine-grained visual features while constructing semantically disentangled and discriminative representations. Conventional methodologies encounter a fundamental trade-off: contrastive learning enhances inter-class discrimination by suppressing intra-class variability, whereas generative reconstruction emphasizes perceptual fidelity but lacks explicit comparative supervision. The naive aggregation of these objectives via composite loss functions often induces optimization conflicts and instability due to gradient interference within the shared latent space. To resolve this, we introduce a unified representation-learning paradigm that reformulates generative reconstruction as a discriminative ranking problem within the generative output manifold. Our approach maintains perceptual reconstruction as a supervisory signal and integrates a ranking-based discriminative constraint by systematically ordering matched and mismatched reconstruction outputs. This formulation embeds discriminative information seamlessly into the generative process, alleviating the risk of loss competition. Specifically, we propose AUC Generative Reconstruction (AUCGR), which optimizes the area under the ROC curve (AUC) by ensuring higher scores for matched pairs than for mismatched pairs, and Partial AUC Generative Reconstruction (pAUCGR), which addresses supervision bias by focusing on mismatched comparisons within reliable feature subspaces. The proposed objectives are designed to be plug-and-play, requiring no architectural modifications to existing generative reconstruction frameworks. Theoretical analysis establishes rigorous connections to AUC/pAUC optimization and provides generalization guarantees. Extensive experiments validate that our method advances both discriminative accuracy and fine-grained perceptual representation.

0.16

Any2Building: Unifying Point Clouds and Images for Building Reconstruction via Shared Spatial Queries

cs.CV ·2026-11-17 ·Hongye Hou, Zhuoxiao Li, Liu Xiangyu, Tengxi Wang, Jing Ou +4 more
Abstract

Abstracting structured buildings from limited observations is essential for 3D urban modeling and digital twins. Partial point clouds and overhead images of the same building provide incomplete 3D surface measurements and visual cues to its footprint, roof geometry, and rooftop structures, respectively. This motivates a unified framework that aligns geometric and visual cues in 3D and reconstructs buildings from either or both modalities. Building on this connection, we present Any2Building, a unified model for Level of Detail 2 (LoD2) building reconstruction from partial point clouds, overhead images, or both, with optional coarse geometric priors. At its core is an occupancy-grounded shared spatial queries representation that organizes available observations within a common 3D grid. Supervised by complete building surface voxels, these queries connect observational evidence to target geometry, enabling a shared generator to learn structural priors across tasks and providing spatial guidance for recovering missing structure. A trained voxel conditional surface-generation and low-poly remeshing pipeline then converts the predicted structure into compact LoD2 building models. To support this framework, we introduce BuildingWorld-Image, comprising approximately one million paired samples of partial point clouds, overhead images, and LoD2 building models across 18 cities. Experiments on multiple benchmarks demonstrate that Any2Building outperforms existing methods in geometric reconstruction accuracy across unimodal and multimodal point-cloud completion, single-image building generation, and prior-assisted generation, while maintaining detailed and robust recovery when point cloud input modality is degraded.

0.16

SignAligner: Dual-Granularity Span Generation for Sign Language Alignment

cs.CV cs.CL ·2026-11-17 ·Jongbhin Woo, Youngjoon Jang, Joon Son Chung
Abstract · 11 pages, 5 figures

Large sign-language corpora benefit from temporal labels at two scales: subtitle spans for subtitle alignment and word spans for sign spotting. Sign spottings already locate words in time at corpus scale, but training subtitle alignment jointly on subtitle and word spans remains underexplored. We present SignAligner, a prompt-conditioned visual-language span generator that maps a sign-language video and a subtitle-sentence or target-word query to a textual start/end frame interval. It is first trained jointly on subtitle spans and word spans from sign spottings, then refined with overlap-based rewards while retaining auxiliary word examples. On the subtitle test set of the British Broadcasting Corporation-Oxford British Sign Language (BOBSL) corpus, the refined model reaches 66.68 F1@0.50 (F1 with a temporal intersection-over-union threshold of one half), the best among compared systems without external sign-video resources. This overlap-reward refinement outperforms cross-entropy continuation from the same jointly trained model. Joint training with word spans improves subtitle alignment in a pair of our generators matched in code, labels and maximum update budget. The advantage remains after both models of the pair are refined with overlap rewards. For word queries, the refined model follows the queried word, not a fixed position in the video: when the word's sign lies away from the video centre, it is far more accurate than predictions that use position alone. Together, these findings support one shared span generator refined with overlap rewards, aligning subtitles and producing word spans for human review.

0.15

Uncovering the Invisible in Routine Radiology: Cross-Modal Detection of MRI-Verified Soft-Tissue Pathologies from CBCT via Physics-Guided 3D Self-Supervised Learning

eess.IV cs.CVcs.LG ·2026-11-17 ·Evgeny Alves Limarenko
Abstract

Routine X-ray modalities like cone-beam computed tomography (CBCT) are optimized for high-contrast mineralized tissues, leaving soft-tissue abnormalities buried beneath photon noise and below human visual perception. Thus, detecting non-calcified pathologies requires secondary magnetic resonance imaging (MRI). We demonstrate that sub-visual radiographic signatures of soft-tissue pathology are physically preserved within routine CBCT scans and can be extracted for opportunistic screening using a targeted, physics-guided deep learning pipeline, even under extreme paired-data scarcity. Using the temporomandibular joint (TMJ) as a model system (N = 212 paired CBCT-MRI joints), we address the severe overfitting of high-capacity 3D networks through a two-stage framework. First, we employ a frozen full-grid 3D masked autoencoder (BertsWin) pre-trained on 6,550 unlabeled CBCT volumes via the GradientConductor (GCond) optimizer and an anatomy-aware objective (PhysLoss) redirecting gradient flow into the [-300, 300] HU soft-tissue window. Second, automated localization of the condylar zenith drives a deterministic 3D cropping that eliminates 81% of background osseous variance. Evaluated via a patient-isolated 3x10-fold stratified cross-validation protocol, a statistically robust sub-visual signal for MRI-verified synovial effusion (ROC AUC = 0.793) emerges exclusively from representations pre-trained with the PhysLoss and GCond regime. Crucially, only this setup captures true 3D anatomical chirality (p < 0.001) and high intra-patient structural consistency (0.791) over inter-patient noise; conversely, identical architectures pre-trained with standard L2 loss or unmasked variance decomposition (MVC) collapse to random guessing. These findings establish a generalizable paradigm for repurposing routine, low-contrast radiographic scans into opportunistic screening tools for occult soft-tissue disease.

0.14

ReTouchGS: Ultralight Render-free Aesthetic Color Enhancement and Transfer for 3D Gaussian Splatting

cs.CV cs.AIcs.GR ·2026-11-17 ·Chuanzhi Xu, Boyu Wei, Chengkun Yue, Xuanhua Yin, Yiyang Li +3 more
benchmark
Abstract

3D Gaussian Splatting (3DGS) enables photorealistic scene reconstruction. Existing 3DGS color transfer and interactive editing methods are guided by reference images and user-specified color adjustments. However, in many applications, users simply want aesthetically guided color enhancement that automatically makes the scene more visually appealing while maintaining consistent colors and tones across viewpoints. We present ReTouchGS, to our knowledge the first framework for reference-free aesthetic color enhancement of pretrained 3DGS models. It predicts a scene-level color lookup table (LUT) from the model and its associated camera parameters. ReTouchGS also supports reference-guided color transfer by fitting a scene-level LUT using one edited view of the same scene and its corresponding original render. A shared blend-aware color refinement network with only 44.7K parameters refines an analytic color update to realize either color mapping without rendering. It preserves geometry and delivers standard 3DGS assets in seconds. We also design the ReTouch3D benchmark to support the evaluation of aesthetic color enhancement and reference-guided color transfer across viewpoints. Experiments demonstrate high color accuracy and strong generalization to unseen scenes in both aesthetic color enhancement and reference-guided color transfer.

0.13

Refin3R: Geometry-Guided Refinement for Feed-Forward 3D Reconstruction

cs.CV ·2026-11-17 ·Jongmin Lee, Jaeyeong Kim, Seungryong Kim
Abstract

Feed-forward 3D reconstruction models predict dense point maps from multiple images in a single pass. Without explicit geometric constraints, however, their predictions are often inconsistent across views and lack fine-grained precision. Recent methods incorporate geometric priors such as camera pose or depth, but use them merely as auxiliary inputs, process each view independently, or refine the geometry only where the priors are available. We present Refin3R, which refines feed-forward point maps with Structure-from-Motion (SfM) points, propagating SfM geometry over the image grid. In its dual-branch architecture, a 2D branch extends SfM geometry beyond its coverage using geometry-aware backbone features, while a 3D branch reasons over the SfM points of all views, capturing multi-view structure that the image grid alone cannot represent. The two branches exchange information bidirectionally to jointly refine each view, yielding dense and accurate point maps. Across ScanNet++, ETH3D, Tanks\&Temples, and out-of-domain human and surgical scenes, Refin3R consistently outperforms feed-forward and geometry-guided methods in point map accuracy while running faster than the strongest geometry-guided baseline.

0.13

A high-dimensional spatial-physical coupling operator for generalizable super-resolution of three-dimensional turbulent flows

cs.CV physics.flu-dyn ·2026-11-17 ·Zengyuan Zuo, Junjun Jiang, Fei Mu, Xubo Wang, Xianming Liu
Abstract

Accurate representation of turbulent flows requires simultaneously capturing multi-scale spatial structures and the complex, nonlinear coupling between distinct physical quantities (\textit{e.g.}, velocity components and density). In practice, however, turbulent data are often constrained by low spatial resolution due to hardware limitations and computational costs. While current data-driven super-resolution methods have shown promise by learning mappings via convolutional neural networks (CNNs), they typically treat physical variables as independent channels. This simplified approach neglects intrinsic cross-variable correlations, causing models to struggle with high-fidelity reconstruction, particularly under extremely sparse conditions or in out-of-distribution regimes. To bridge this gap, we propose an architecture-agnostic spatial-physical coupling operator that extends the standard convolution into a joint spatial-variable dimension. Unlike conventional linear channel mixing, our framework constructs a unified spatial-physical representation within CNNs to explicitly capture local inter-variable interactions. By lifting features into a higher-dimensional manifold, this operator significantly enhances the model's representation capacity. Experiments demonstrate that this approach enables strong reconstruction performance and zero-shot generalization, offering a physics-aware inductive bias for super-resolution of three-dimensional turbulence. More broadly, this work suggests that high-dimensional operators offer a promising avenue for modeling intricate multi-scale turbulent flows.

0.12

Beyond Co-Generation: Rethinking Jointness in Joint Audio-Video Generation

cs.CV eess.AS ·2026-11-17 ·Jihoo Jung, Ji-Hoon Kim, Suyeon Lee, Gia Thinh Pham, Joon Son Chung
benchmark
Abstract · Preprint

Our aim is to advance the understanding of genuine audio-visual coordination in joint text-to-audio-video (T2AV) generation, with particular emphasis on semantic consistency. Evaluating audio-visual consistency in T2AV generation is nontrivial, as detailed text prompts obscure whether such consistency stems from genuine cross-modal coordination or merely from independent text following. To isolate the former, we introduce intrinsic jointness: the ability of audio and video to produce the same realization for details left underspecified by the prompt. Using semantically underspecified prompts (e.g., "musical instrument playing" rather than "electric guitar playing"), we investigate four recent models and reveal their limited intrinsic jointness. To address this limitation, we propose JointREPA, a simple representation alignment-based post-training method to facilitate cross-modal coordination. Since existing benchmarks use semantically detailed prompts, we introduce UNSAIDBench, a benchmark designed to evaluate intrinsic jointness under semantic underspecification. Experiments across two distinct models and three benchmarks, including UNSAIDBench, demonstrate that JointREPA consistently improves semantic consistency.

0.11

CEVA: Complementary Evidence Verification via Frame-Level Reorganization for Short-Video Fake News Detection

cs.CV ·2026-11-17 ·Wenshen Yang, Junfeng Hao, Jing Li, Decheng Liu, Yangkunfei +2 more
Abstract

Short-video platforms are now a common way for the public to follow unfolding events. Their rapid, visually persuasive format also makes fake news easier to spread. Unlike text-image posts, short-video news interleaves multiple modalities and changing scenes, making it difficult to identify which frames verify a claim. Existing detectors either pool sampled frames homogeneously or select them by text-frame similarity. Both overlook that aligned frames may repeat a misleading claim, whereas less dominant context may challenge it. We present CEVA (Complementary Evidence Verification Approach), to our knowledge the first short-video fake news detection framework to explicitly organize frames by their evidential roles before multimodal fusion. Its Dual-Branch Complementary Reorganization (DBCR) module separates explicit-alignment frames from candidate complementary-context frames, and cross-modal fusion with bidirectional contrastive regularization operates on the reorganized evidence. On FakeSV and FakeTT, CEVA achieves the best overall results among the compared methods. Under fixed frame budgets and backbones, DBCR outperforms uniform, random, and similarity-based selection and improves two structurally different detectors. Together, these results establish pre-fusion evidence organization as an effective and transferable component of short-video fake news detection.

0.09

As Student Rollouts Continue, Visual Backdoors Keep Growing

cs.CV cs.AI ·2026-11-17 ·Zane Fu, Yanzhao Zheng, Jinze Liu, Ziting Qin, Qianshan Wei +5 more
Abstract

Conventional knowledge distillation can facilitate the propagation of backdoors from teacher models to student models. However, how this risk manifests in On-Policy Distillation (OPD) remains insufficiently studied. OPD has recently emerged as an important post-training paradigm for Vision-Language Models (VLMs). Unlike traditional offline distillation, OPD does not require teacher-generated responses. Instead, it uses the teacher's probability distributions to supervise response trajectories generated online by the student. Since the initial student model typically generates benign responses and receives no explicit supervision from malicious target responses, OPD appears to offer some inherent resistance to backdoor propagation. In this paper, we challenge this intuition and reveal a more subtle risk of visual backdoor propagation in OPD. Through experiments across multiple model families and datasets, together with theoretical analysis, we identify three main findings. First, unlike the relatively gradual propagation observed in traditional offline distillation, backdoors rarely emerge in OPD at low poisoning ratios. However, the attack success rate rapidly approaches 100% once a critical threshold is exceeded. Second, backdoor propagation does not depend on explicit malicious responses. Instead, it is driven by the continuous accumulation of anomalous probability mass assigned by the teacher. Finally, based on these findings, we propose a novel trigger design that enables visual backdoor propagation with zero explicit trigger injection. These findings further reveal an overlooked mechanism of backdoor propagation in OPD and highlight its potential security risks.

0.07

CascadeJEPA: Enhancing 3D Geometric Representations via Spatially Coupled Cascaded Prediction

cs.CV ·2026-11-17 ·Kailai Zhou, Ziqi Chen, Xiaoning Xu, Yuefeng Wu, Zixuan Fu +6 more
Abstract

Joint-Embedding Predictive Architectures (JEPAs) have emerged as a promising paradigm for predictive visual representation learning. However, our analysis of I-JEPA reveals a growing discrepancy between semantic and geometric learning: improved semantic recognition during pretraining is accompanied by degraded 3D geometric representations. We hypothesize that this partly stems from predicting highly abstract latent representations with incomplete spatial context, favoring semantic regularities over fine-grained geometric information. To address this issue, we introduce CascadeJEPA, a spatially coupled cascaded prediction framework that promotes richer patch-level interactions through full-grid context encoding and cross-attention between prediction stages, while largely preserving the original I-JEPA architecture. Extensive experiments demonstrate substantial improvements over I-JEPA across multiple 3D geometric tasks while maintaining strong semantic recognition performance. Furthermore, its advantages extend to future feature forecasting and world model planning, highlighting the potential of enhanced geometric representations to support predictive world modeling.

0.06

Security & Privacy 2

TetherMark: Instance-Level Watermark Authentication Against Cross-Instance Transplantation

cs.CR cs.CV ·2026-11-17 ·Jin Wei, Rundong Li, Li Ma, Yikai Wang, Ruihao Yang +9 more
hot direction
Abstract

Robust watermarking is commonly verified by asking whether a hidden payload remains decodable after compression, geometric distortion, or generative editing. Localized and semantic extensions may additionally check whether an expected region or category is still present. These tests establish evidence validity or content consistency, but not whether the evidence still belongs to the particular visual instance registered by its owner. We expose cross-instance transplantation: an attacker extracts or estimates valid watermark evidence from one registered image and transfers it to a different image of the same category. The message can still decode, the expected content can remain present, and the result can look natural, even though the evidence has been reassigned to the wrong instance. To enable instance-level watermark authentication, we introduce TetherMark, a post-hoc watermark that combines two complementary checks. The message check verifies an edit-robust payload, while the relation check uses secret patch-pair selection, a registered response, natural correspondence transport, and enrollment normalization to test consistency with the registered instance. The verifier accepts only when both checks pass. Across 240 test subjects, TetherMark attains 92.08% legitimate true acceptance with 9.17% and 16.25% attack success under WmForger and WMCopier, respectively, yielding the highest instance-binding balanced accuracy under both attacks.

0.38

CaFe: Causally Faithful Conditional Traffic Scenario Generation

cs.CR cs.CVcs.RO ·2026-11-17 ·Steven Yu Tung, Shubham Vaijanath Phoolari, Christoph Lauer, Abhinav Valada, Steven Peters
Abstract

Language-conditioned traffic simulation enables scalable test scenarios steered by natural language. The goal is to control selected agents while surrounding traffic reacts to their observed behaviour. For example, a vehicle instructed to cut in late should cause the following vehicle to brake only when the manoeuvre becomes observable. Yet simulators can leak private instructions across agents, causing surrounding vehicles to react before the instructed behaviour becomes observable. Such premature reactions defuse desired events and alter the ensuing closed-loop scenario. Existing benchmarks measure what agents do, not what they could have known, leaving this failure unmeasured. We formalise it as intent leakage and address it through evaluation, data and modelling. Our paired intervention test holds an instructed agent's executed trajectory fixed while varying its private instruction, isolating leakage-induced changes in surrounding agents. We also construct WOMD-PACT, an automatically generated, verifiable dataset of 531,099 WOMD scenes with timed behaviour tags with reaction episodes identified through a counterfactual ghost test and grounded private instructions. Finally, we introduce CaFe, a dual-stream diffusion-forcing simulator fine-tuned in closed loop that restricts inter-agent interactions to public beliefs, preserving legitimate anticipation without leaking private intent. Our evaluation reveals that existing conditioning designs displace 23-83% of nearby uninstructed agents through private-intent leakage, even when the instructed agent's executed motion is held fixed. CaFe eliminates measurable leakage while maintaining comparable WOSAC realism (0.763), demonstrating that causal faithfulness can be enforced without sacrificing measured simulation realism.

0.21

Robotics, Control & Signals 8

FM-Mamba: An Efficient Native 2D Selective State Space Model with Stability-Aware Discretization

eess.SY cs.AIcs.CV ·2026-11-17 ·Shengyuan Hu, Yiyong Sun, Weihang Shan, Yongzhe Chang
code/data
Abstract

When applied to two-dimensional features such as images, Mamba-based methods either flatten the feature map into 1D sequences and fuse multiple scans, which breaks spatial adjacency and multiplies computation, or recurse directly on the 2D grid but rescale the recurrence with a fixed input-independent factor and aggregate only the predecessors on a single anti-diagonal, leaving the discrete step unconstrained by 2D stability theory. We propose \textbf{FM-Mamba}, a native 2D selective state space model built on the general Fornasini--Marchesini (FM) model. Its coupled recurrence evolves along both spatial axes; cross-axis coupling induces a discrete diagonal-predecessor term $\tilde A_0$, and two independent input-dependent step sizes $Δ_1(i,j)$, $Δ_2(i,j)$ govern propagation along the two axes, so that the coefficients adapt per position and per direction. A normalized ratio $r_k=a_{0,k}/(a_{1,k}a_{2,k})$ yields a closed-form piecewise bound on the two step sizes; when a configuration falls outside the stable interval the steps are projected onto this bound, and the criterion is re-verified per position after training, which suppresses systematic local amplification without global damping. The recurrence is executed by a dual-diagonal wavefront scan: each update gathers information from two anti-diagonals, and the input enters the current state through a single projection, so that neither training nor inference flattens the feature map. Code is available at https://github.com/BitSsm/FM-Mamba-Runtime

0.43

Heave-Ho: Decentralized Long-Horizon Multi-Humanoid Collaboration

cs.RO cs.CVcs.LG ·2026-11-17 ·Ting-Hsuan Chen, Siheng Zhao, Han Jiang, Qi Pan, Louison Lu +5 more
benchmarkcode/data
Abstract · 18 pages, 11 figures, 7 tables. Project page: https://heaveho-humanoid.github.io/

Collaboration promises to extend humanoid capabilities beyond what individual robots can achieve. However, realizing this potential requires coordinated whole-body interactions, coherent long-horizon planning, and precise synchronization across robots. We introduce Heave-Ho, a decentralized framework for long-horizon collaborative multi-humanoid loco-manipulation. At the low level, we develop a multi-agent post-training framework that adapts a humanoid foundation tracking model to physically coupled interactions between humanoids, enabling coordinated decentralized control from local observations alone. At the high level, we train a vision-language planner and design a speech-based leader-follower coordination mechanism to synchronize decentralized execution without network communication. Trained with two identical humanoids, our system scales to teams of up to ten robots and generalizes to heterogeneous teams without retraining. To benchmark these capabilities, we introduce HH-Bench, comprising 12 tasks involving teams of 2 to 10 humanoids, including heterogeneous teams. Systematic evaluations show that Heave-Ho consistently outperforms baseline methods, while real-world experiments demonstrate robust decentralized collaboration across challenging tasks.

0.38

CRA: Continuous Risk Assessment for Safety-critical Visual Navigation

cs.RO cs.CVeess.SY ·2026-11-17 ·Lihao Zheng, Ying Wu, Chen Jiang, Randy Freeman
code/data
Abstract · Project page: https://isaac-restaurant-cra.cra-demos.workers.dev

Traditional safety-critical control methods often rely on predefined constraints that separate states into safe and unsafe regions. While recent studies have extended geometric safety toward semantic safety, real-world safety is rarely purely binary. The risk associated with an object or human can vary continuously with its physical properties, like motion, and interaction context; for example, a person carrying a heavy object or a hot object moving at high speed may require a different safety response from an ordinary static obstacle. In this paper, we propose a vision-based continuous risk assessment framework for safety-critical visual navigation. Our method uses RGB and LiDAR observations to estimate four risk factors (kinetic, potential, contact-severity, and thermal) and combines them into a continuous risk score for each assessed object or person. These estimates are integrated into a Poisson safety function with a Laplace guidance field as a safety filter, enabling online monitoring and intervention in open-world robot navigation. We demonstrate the proposed method in a simulated restaurant scene.

0.33

TCP4D: A Unified Tool-Center-Point Tracking and Generation Model for Robot Manipulation

cs.RO cs.CV ·2026-11-17 ·Haosong Peng, Xiangyang Liu, Zhengshen Zhang, Yunfeng Fan, Tianyu Qi +3 more
code/datacommunity pick
Abstract · Project page: https://livioni.github.io/TCP4D/

Robot manipulation videos provide a rich visual experience, yet recovering robot action and gripper-state annotations from RGB observations still remains challenging. Using the embodiment-agnostic tool-center-point (TCP) pose as the action space, we present TCP4D, a unified model in which a shared 4D scene representation both tracks observed TCP motion and generates future TCP trajectories. Given a monocular video and initial 2D TCP queries, the tracker directly recovers metric TCP poses in SE(3) and gripper states alongside scene geometry and camera parameters. The generator samples local features from the same 4D representation along the recovered TCP tracks, then combines them with robot observations and language instructions to predict future TCP states. Our new cross-embodiment augmentation pipeline converts existing RGB videos into demonstrations for alternative embodiments by retargeting the recovered motion and synthesizing the corresponding observations, without recorded source-robot states. Experiments show that the tracker achieves absolute trajectory errors of 7.60 and 16.07 mm on DROID and RoboLab and adapts to a new embodiment with only ten demonstrations. The generator achieves 69.19% mean success on RoboTwin and outperforms well-established policy models on two real-world tasks, while joint training on source and augmented demonstrations improves target-robot success by 41.5% over augmented-only training.

0.30

Technical Report of Mobile Manipulator Robot for Industrial Environments

cs.RO cs.CVeess.SY ·2026-11-17 ·Aria Dadnavi, Matin Jamshidi, Kian Khatibi, Shahriar Karimi Makhsous, Saina Najafi Shabestari +6 more
Abstract · 25 pages, 22 figures, 4 tables. Technical report of Team Auriga's 2025 mobile manipulator (@Work) platform (Robotics and Intelligent Automat…

This paper presents the design and implementation of Auriga Robotics' mobile manipulator robot, developed at the Robotics and Intelligent Automation Laboratory, Department of Electrical Engineering, Shahid Beheshti University. Hereafter, this platform is referred to as the @Work Robot. The robot targets industrial tasks, with an emphasis on improving efficiency in repetitive or hazardous work environments. It uses a 4-wheel Mecanum drive for omnidirectional movement and a 5-degree-of-freedom manipulator arm with a 3D-printed gripper for object handling and navigation. The electronics consist of custom PCB boards built around ESP32 microcontrollers, together with an NVIDIA Jetson Orin NX that handles high-level processing and real-time system control. On the software side, Hector SLAM provides mapping and localization, A* handles path planning, and YOLO performs real-time object detection. These modules draw on fused sensor data to deliver robust navigation and reliable collision avoidance. Overall, the system shows effective hardware-software integration, with consistent performance in autonomous navigation, object handling, and task execution in industrial-like settings.

0.23

RFW-S: Everyday Deformable Objects with Independent Simulation Verification

cs.RO cs.CV ·2026-11-17 ·Jiarui Guo, Saining Zhang, Jiaying Wan, Jiawei Weng, Junhao Chen +13 more
datasetcode/data
Abstract · 34 pages, 19 figures. Project Page: https://sjijs.github.io/RFW-S_website/

Large-scale robotic simulation requires diverse deformable objects with executable physical properties and realistic interaction behaviors. However, existing 3D asset collections primarily emphasize visual diversity, while scalable physical construction and verification remain challenging. We introduce RFW-S, a large-scale dataset of 83,365 simulation-verified assets spanning 485 everyday categories across five structural families: soft solids, cloth sheets, shells, garments, and bags. We develop an automated pipeline integrating reference-guided 3D generation, structure-aware physical modeling, and LLM-assisted parameter estimation and calibration. To ensure physical reliability, we introduce an independent simulation verification framework that evaluates object behaviors under interactions distinct from those used for parameter calibration. Extensive experiments demonstrate the effectiveness of independent verification and characterize deformable responses under different physical configurations. We further evaluate robotic grasping in terms of both task success and object integrity, and demonstrate the physical realism of our simulated objects through sim-to-real replay of fragile-object manipulation. RFW-S provides a scalable foundation for physically grounded simulation and embodied robotic learning.

0.23

DailyBrick: Language-Guided Synthesis of Brick Assembly Instructions

cs.RO cs.CV ·2026-11-17 ·Hang Zhang, Qijian Tian, Yuan Xie, Xuhong Wang, Xin Tan
Abstract

Generating a LEGO model from a user request is more than a plausible 3D shape generation. LEGO models require physical support per brick, and each brick should be placed unblocked. Current methods address this issue by training an expert large language model. We reduce the problem of generating extremely long context on all brick connections to that of writing a program that builds LEGO models like a human, considering each intermediate step could be executed. We introduce DailyBrick , a multi-stage agentic system that produces a LEGO model together with a brick-by-brick assembly instruction for human building and robot assembly. We design a multi-stage harness with roles to help LLMs generate more effectively. A versioned construction SDK compiles multi-level decisions, from semantic subparts to individual bricks, into canonical operations. An external Meta-Reviewer agent replays and inspects the generated package, verifies the model and its construction process, identifies failures, traces them to the responsible stage, and routes them to bounded repairs or to editable candidates for later skill or tool updates. Extensive experiments demonstrate that DailyBrick generates LEGO models based on simple user requests better than recent agentic systems, and show its utility in downstream applications such as robot manipulations on model assembly from instructions and brick sets (https://SpatialAgent.github.io/DailyBrick/).

0.15

JointGV: Structure-Aware Correction of Visual World Models with Parallel Graph Dynamics for Long-Horizon Modeling

cs.RO cs.AIcs.CV ·2026-11-17 ·Tian Lyu, Yichen Jiang, Ziyuan Cheng, Yunda Yan, Xiangyu Pan +5 more
Abstract

Visual world models predict future visual states conditioned on actions, but recursive errors in entity poses and spatial relations limit their long-horizon prediction for robot planning and control. We ask whether a separately predicted relational state can correct this drift without restructuring the visual model. This is challenging because graph nodes and visual states share neither elements nor a feature space. We propose Joint Graph--Visual Dynamics Correction (JointGV), a plug-in based on a cross-granularity dynamics hypothesis: despite their different representations, graph-node and visual-state changes describe the same physical transition. JointGV rolls out a task-conditioned scene graph alongside the visual world model, couples their predicted changes, and combines the resulting graph-motion context with graph-predicted node positions to compute a bounded residual. Graph-learning and correction gradients do not propagate into the original visual dynamics. We attach JointGV to DINO-WM (spatial tokens) and DreamerV3 (a global latent state), evaluating both on DMC Cheetah Run and Walker Walk across three seeds and extending the evaluation to contact-rich manipulation with DMC Finger Turn Hard. Across 32-step rollouts, JointGV improves structural and visual-prediction accuracy, reducing joint-position error by 11--28\%. For spatial tokens, learned graph--visual change coupling gives lower error than broadcasting the same mean graph-motion context to every token, particularly when non-adjacent parts appear close together in the image. On real-robot inverted-pendulum sequences, JointGV reduces DINO-WM's 32-step pole-angle error from $12.56^\circ$ to $6.69^\circ$ and improves DreamerV3 and LeWorldModel as well. In simulated inverted-pendulum control, an actor trained with JointGV-corrected rollouts completes 46 of 50 episodes, compared with none using the original rollouts.

0.11