Probing Spatial State Representations in Visual Foundation Models
Baiqiao Yin, Zhongang Cai, Ruisi Wang, Qiwei Liang, Lei Yang, Ziwei Liu
AI 速览：该研究通过四个层级的冻结特征探针，检验视觉基础模型在流式自我中心观测下是否具备可维持、可预测更新并可用于目标导向动作的空间状态表征。
An observer sees only part of the surrounding world, yet must keep track of what lies beyond the current view. Spatial perception therefore involves more than recognizing individual images: it calls for a persistent spatial state that is updated under self-motion and used to guide action. Do visual foundation models provide representations that support such a state? We study this question through four levels of spatial capability---spatial structure, object persistence, predictive spatial update, and actionability---using frozen-feature probes on ten representations under streaming egocentric observations of static indoor environments. The resulting capability profiles vary across tasks, network layers, and prefix lengths: strong readout of the visible scene does not ensure accurate object persistence or future-feature retrieval. Matched controls separate readout accuracy from reliance on visual history and motion. Native future prediction retains recoverable spatial structure, but a spatial reader trained on real features does not transfer to predicted features unchanged. Finally, goal-dependent candidate selection does not establish reliable continuous action utility: predicted actions respond to goal changes, yet supplying the correct goal does not reliably improve progress. Across the four levels, object persistence and actionability build on camera-pose readout, whereas feature prediction builds on visible-scene readout. Together, these findings motivate evaluating spatial representations not only by what can be decoded from them, but by whether that information is maintained beyond observation, updated predictively, and used for goal-directed action.
HairDeck: Animation-Ready Stylized Hair Cards from a Single Image
Yuze He, Yixiang Wei, Wang Zhao, Chuhao Chen, Yanning Zhou, Yong-Jin Liu
AI 速览：HairDeck 将风格化发型视为无序的发片集合，用集合扩散 Transformer 从单张图像一次性生成发片及其弹簧骨骼链，得到可直接导出的可动画资产。
The hair of stylized characters in games and avatars is built from hair cards: artist-placed ribbons whose geometry is the visible hair, animated by spring bones. Existing image-based methods produce realistic hair or a single fused mesh, while CHARM, the only generator of stylized hair cards, decodes a hairstyle autoregressively as a long sequence of control points. Observing that a card hairstyle is an unordered set of artist-defined primitives, we propose HairDeck, which generates it as one: each card is a single token, and a set diffusion transformer denoises all cards together. From a single image, HairDeck generates the cards and the spring-bone chains that drive them, yielding an animatable asset that exports to ready-to-use downstream formats. On CHARM's own test set, it lowers Chamfer distance by 44% and card-orientation error by more than half while running 33 times faster, and from single images it outperforms CHARM, image-to-3D pipelines and strand-based methods. Project page: https://hyzcluster.github.io/HairDeck
AI 速览：VitronEdit 提出一个统一框架，用共享序列表示和渐进式多模态训练，把图像与视频编辑知识迁移到数据稀缺的3D场景编辑。
Recent advances in unified visual editing have begun to consolidate image and video editing within a single model, while 3D scene editing remains largely separate and constrained by scarce paired training data. Given the rich image and video editing supervision now available and the common editing semantics shared across modalities, a natural question is whether this knowledge can be transferred to data-scarce 3D editing. We present VitronEdit, a universal framework for image, video, and 3D scene editing. VitronEdit adopts a shared sequence representation and progressive multimodal training to transfer editing knowledge across modalities. We further introduce geometry-aware post-training, which leverages a pretrained geometry reconstruction model to provide depth and camera-pose supervision. Surprisingly, applying geometric supervision only to 3D samples yields the best overall performance, while image and video editing also improve despite receiving no direct geometric supervision, suggesting that geometric regularization transfers through the shared backbone. Our analysis further shows that image and video training mainly transfers editing semantics to 3D, whereas 3D supervision primarily improves multi-view consistency. The project page: https://vitron-llm.github.io/VitronEdit/.
AI 速览：作者构建了迄今规模最大的真实图像婴儿姿态数据集 InfantNet，并建立婴儿 2D 关键点检测与 3D 姿态形状恢复基准。
Infant pose and shape estimation is essential for applications in childcare, developmental monitoring, and medical diagnosis. However, existing methods and datasets are largely designed for adults, and direct transfer to infants fails due to substantial differences in body proportions, articulation limits, and frequent self-occlusion. To address this gap, we introduce InfantNet, the largest real-image infant dataset to date, comprising 108,902 RGB images of infants aged 6-18 months. Each image is annotated with 2D keypoints, and a curated subset of 11,642 images additionally includes 3D pose and shape annotations with full SMIL parameters, verified through multi-view rendering and reprojection-error filtering. We use an iterative annotation pipeline to ensure high fidelity across both 2D and 3D labels. InfantNet establishes a large-scale, comprehensive benchmark for infant 2D keypoint detection and 3D pose-and-shape recovery. Baseline experiments demonstrate that state-of-the-art adult pose estimators do not generalize well to infants, whereas fine-tuning on InfantNet yields a consistent improvement. The gains are even more pronounced for 3D pose and shape estimation, where SMIL-based estimation demonstrates robust performance with accuracy comparable to results reported for SMPL-based models. By releasing the InfantNet dataset and benchmark, we provide a vital resource for advancing infant pose analysis and related healthcare applications.
HiFOR: Hierarchical Foresight with Action-Guided World Grounding for Vision-Language-Action Model
Rongxing Ding, Hongyu Qu, Kehong Wu, Xudong Wang, Fan Zhang, Xinguang Xiang, Xiangbo Shu
AI 速览：HiFOR 提出层级前瞻预测与动作引导的世界检索框架，以增强视觉-语言-动作模型对短期动态和长期目标的联合建模能力。
Vision-Language-Action (VLA) models have demonstrated promising capabilities in robotic manipulation by integrating visual perception, language understanding, and action generation. Recent advances have further incorporated future prediction to enhance the understanding of environmental dynamics and task progression. However, existing methods often model short-term dynamics and long-term task objectives separately, while directly incorporating predicted futures into action policies without sufficiently considering their relevance to current action requirements. To address these limitations, we propose HiFOR, a hierarchical foresight framework with action-guided world retrieval. Specifically, Hierarchical Foresight Prediction jointly predicts short-term state evolution, semantic milestones, and final goal states through a Hierarchical Dependency Mask (HDM), establishing dependencies across temporal horizons. Building upon these predictions, Structure-Aware Latent World Modeling organizes future latent representations along the horizon, view, and spatial dimensions to facilitate fine-grained retrieval. Furthermore, Action-Guided Factorized World Retrieval adaptively retrieves relevant future information from the structured latent world based on evolving action representations during denoising, providing targeted guidance for action generation. Extensive experiments on two simulation benchmarks and six real-world robotic manipulation tasks demonstrate the effectiveness of our approach.
Building general game agents requires both high-level task reasoning and low-latency action control in rapidly changing visual environments. These capabilities favor models with substantially different computational profiles, making them difficult to realize within a single model. In this work, we present TaskFrame, a hierarchical planner-controller framework that connects task-level planning with frame-level keyboard-and-mouse control. A vision-language task planner decomposes a high-level goal into an ordered sequence of atomic instructions and is optimized through supervised learning and dependency-aware GRPO. The hidden states of special goal tokens that terminate each atomic instruction provide continuous goal embeddings that are cached and reused across control steps. A causal Transformer controller combines the embedding of the active goal with recent observations to predict discrete keyboard and mouse-button actions and continuous camera motion at every game step. We train the system in three stages: planner training, controller pretraining, and joint alignment. Experiments in Minecraft show that TaskFrame is competitive with general-purpose vision-language models on instruction decomposition, and that in closed-loop evaluation the agent achieves a higher combat success rate than the evaluated baselines.
Multimodal large language models (MLLMs) are now used as zero-shot judges of image quality, but their color judgments are rarely tested with colorimetric ground truth. We present a five-task benchmark based on the CIE colorimetric framework: AWB pair judgment, skin-tone bias direction, CCT estimation, CIE Delta E 2000 pair ordering, and color naming under changing illuminants. We test eight MLLM endpoints, including four Qwen3-VL variants, GLM-4.5V, GPT-5.4, Claude-Opus-4.7, and Gemini-3.1-Pro. The models often solve coarse binary AWB judgments, but they remain weak on numeric CCT estimation and on some fine color-difference judgments. On ColorChecker-24 stimuli, the three closed-weight frontier models reach CCT MAE of 715-1056 K, compared with 1358-2133 K for the five open-weight models. A non-ColorChecker control shows that this advantage is partly template-dependent. The results suggest that current MLLMs have useful color words and coarse visual priors, but their color judgments are not reliably calibrated to CIE-computable color differences.
Video generation is shifting from end-to-end synthesis toward agentic generation, where planning, tool use, and feedback enable coherent, controllable multi-shot videos. In parallel, video world models provide action-conditioned observations and predicted futures, serving as interactive environments for agents. We study their intersection through the agent harness: an executable runtime that constructs context, exposes tools, maintains state, routes feedback, and manages budgets around video models. This report examines three directions. First, agentic video generation, where harnesses coordinate scripting, shot planning, asset management, production, and editing, alongside the evaluation challenges of these systems. Second, agentic world modeling, where harnesses mediate observation, control, execution, and state. We discuss hybrid designs in which game engines or agent-written programs maintain world state while video models render observations, and how agents and world models can mutually evaluate and improve one another. Third, recursive self-improvement (RSI) in video model development, distinguishing output revisions, updates to harness state, generator parameters, and evaluators, and changes to the improvement procedure itself. We identify how these improvements feed back into subsequent development. Across these directions, we analyze essential agent capabilities, including audiovisual perception, long-horizon planning, memory, code-based interaction, feedback-guided control, skill reuse, and multi-agent coordination. We conclude with challenges in autonomous video production, lifelong memory, programmable hybrid control, autonomous model development, provenance, cost accounting, and evaluation linking video quality and task performance to sustained improvements in models and harnesses.
Minyan Luo, Yuxin Zhang, Meng Zhao, Xincan Wang, Fuzhang Wu, Tong-Yee Lee, Oliver Deussen, Fan Tang 等 9 位作者
An outstanding creative advertising image should both attract attention and communicate a product's positioning. How to choose a persuasive conceptual bridge before rendering begins is challenging for advertising generation. We present AdMuse, a skill-guided multi-agent framework that distills high-quality advertisements into reusable creative skills for cross-domain advertising strategy. Rather than retrieving isolated exemplars, AdMuse abstracts recurring bridge patterns into a skill library that can guide new product briefs. At generation time, AdMuse coordinates skill routing, multi-candidate ideation and selection, and VLM-guided visual refinement: diverse skills propose alternative concepts, relative judgment selects the most promising direction, and vision-language feedback helps realize the selected intent in the final image. This design treats advertising creativity as a structured decision process rather than a single prompt-writing step. Experiments across diverse product and brand positionings show that AdMuse produces visually compelling advertisements that faithfully communicate product positioning.
Zhengdong Hu, Chao Wang, Kaituo Feng, Huanjin Yao, Mingfei Lu, Hehe Fan, Yi Yang
Multimodal large language models (MLLMs) can now process hour-long videos, making extended multi-turn dialogue over a single video a natural mode of use. Sustaining such a dialogue requires more than answering each question in isolation: the model must retrieve and reuse information established earlier in the conversation, both what it has seen and what it has said, an ability we term in-context video memory. Existing video QA benchmarks leave this ability unmeasured: their questions are mutually independent, never reference one another, and each is answerable from the video alone, so nothing ever requires carrying information across turns. We introduce VideoMemBench, in which each sample pairs a single video with a long multi-turn dialogue, whose turns are connected by typed cross-turn dependencies: an Episodic Dependency refers back to a visual episode described in an earlier turn, probing memory of what the model has seen, while an Answer Dependency builds on the model's own earlier answer, probing memory of what the model has said. Every dependency is verified considering two properties: 1) necessity, meaning the question cannot be answered without its earlier turns, and 2) non-leakage, meaning the question does not restate the information from its earlier turns. To enforce these properties, we design a construction pipeline with three LLM roles: a planner, a generator, and a judger, with each dialogue reviewed by human experts. VideoMemBench comprises 326 videos with 2,476 turns, of which 1,387 depend on earlier turns, including multi-parent dependencies and long-range call-back. Among over 20 evaluated open- and closed-source MLLMs, Claude Opus 5 performs the best with an overall score of only 45.7\%, highlighting the challenge of in-context video memory. Our dataset and evaluation are released.
In this paper, we propose Concentrate and Concentrate (CaC), a coarse-to-fine anomaly reward model based on Vision-Language Models. During inference, it first conducts a global temporal scan to anchor anomalous time windows, then performs fine-grained spatial grounding within the localized interval, and finally derives robust judgments via structured spatiotemporal Chain-of-Thought reasoning. To equip the model with these capabilities, we construct the first large-scale generated video anomaly dataset with per-frame bounding-box annotations, temporal anomaly windows, and fine-grained attribution labels. Building on this dataset, we design a three-stage progressive training paradigm. The model initially learns spatial and temporal anchoring through single- and multi-frame supervised fine-tuning, and then is optimized by a reinforcement learning strategy based on two-turn Group Relative Policy Optimization (GRPO). Beyond conventional accuracy rewards, we introduce Temporal and Spatial IoU rewards to supervise the intermediate localization process, effectively guiding the model toward more grounded and interpretable spatiotemporal reasoning. Extensive experiments demonstrate that CaC can stably concentrate on subtle anomalies, achieving a 25.7% accuracy improvement on fine-grained anomaly benchmarks and, when used as a reward signal, CaC reduces generated-video anomalies by 11.7% while improving overall video quality.
Accurate multi-organ segmentation in computed tomography (CT) is essential for computer-aided diagnosis, treatment planning, and disease monitoring. However, substantial variations in organ shape and size, together with low tissue contrast and ambiguous boundaries, complicate the modeling of heterogeneous anatomical structures. To address these challenges, we propose a Morphology-Aware Channel-Decoupled Mamba Network (MCDNet), which integrates adaptive local structural enhancement with channel-decoupled long-range modeling. Specifically, we introduce a Morphology-Aware Channel-Decoupled State-Space (MCSS) block that divides feature channels into separate processing groups, allowing differentiated structural representations to be learned within an efficient state-space framework. Within each group, a Morphology-Aware Enhancement Module (MAEM) combines deformable convolutions with channel attention to adapt local feature extraction to anatomical shape variations. Group-wise state-space modeling then captures long-range spatial dependencies, and channel gating adaptively integrates the resulting features. Experiments on the Synapse abdominal CT dataset demonstrate the effectiveness of MCDNet, achieving an average Dice Similarity Coefficient of 81.61% for multi-organ segmentation.
Understanding the internal routing of visual concepts is crucial for the safe and controllable adaptation of generative models. While concept localization has been widely studied in diffusion models, the emerging paradigm of next-scale visual autoregressive models remains largely unexplored. In this paper, we introduce LoCo, the first model-agnostic framework to localize where and when specific concepts emerge within autoregressive models. Driven by the native coarse-to-fine nature of next-scale generation, our method precisely maps conceptual knowledge across three distinct dimensions: Layer, Scale, and Position. To systematically localize and evaluate concept routing without context bias, we propose LoCoBench, a comprehensive dataset spanning 10 diverse categories. Extensive probing on image and video autoregressive models like Infinity, HunyuanImage-3.0, and InfinityStar shows that the localized positions are both interpretable and causally related to concept emergence. Building on these insights, we apply our localization method to three key applications: concept erasure, model personalization, and adversarial concept injection. Experiments demonstrate that our targeted intervention achieves state-of-the-art performance, substantially reducing computational overhead while preserving benign utility. Overall, our findings offer insights into how conceptual knowledge is routed during autoregressive generation, introducing a practical pathway for more interpretable, efficient, and secure adaptation.
Medical image segmentation is crucial for clinical diagnosis and treatment planning, requiring both global contextual understanding and fine-grained structural precision. While the Segment Anything Model (SAM) demonstrates strong generalization capabilities, its direct application to medical images often lacks domain-specific adaptability. Meanwhile, State Space Models (SSMs)-based methods, such as Mamba, efficiently capture long-range dependencies but may struggle with preserving fine anatomical details. To address these challenges, we propose a segment anything model based context guided mamba unet, SAMBA-UNet, a novel framework that integrates SAM's context knowledge with Mamba's hierarchical feature learning to enhance both global representation and local detail refinement. Specifically, we introduce the Hierarchical Context Enhancement Module (HCEM), an attention-based fusion module that deeply incorporates the multi-level features of SAM and Mamba to improve detail perception and segmentation precision. Additionally, we employ LoRA and Adapter-based fine-tuning to adapt SAM to medical imaging tasks, further enhancing its domain-specific segmentation capability and robustness. We conducted a comprehensive study on several medical segmentation datasets, including Synapse, ISIC17, and ISIC18, evaluating our method across multiple metrics. Experimental results demonstrate that SAMBA-UNet achieves state-of-the-art performance, effectively balancing generalization, segmentation accuracy, and computational efficiency. Our code will be released upon acceptance.
Xiangfei Sheng, Xiaofeng Pan, Leida Li, Qian Zhang, Biao Wang, Johan Wagemans, Weisi Lin
Aesthetic image cropping (AIC) aims to improve image composition and aesthetic quality by removing extraneous or distracting content. Over the past decade, AIC methods have evolved from specialized vision models to multimodal large language models(MLLMs). Reliable evaluation that reflects human preferences is essential for benchmarking AIC methods and guiding model optimization. However, the current AIC evaluation relies largely on geometry-based metrics such as intersection over union (IoU), as well as image aesthetic assessment(IAA) models. Although inconsistencies with human preferences have long been observed, the robustness and reliability of these evaluation criteria remain underexplored. To bridge this gap, we first construct CropBench, a large-scale benchmark comprising 106,995 human preference pairs for 1,019 images, with each candidate set containing the original image and candidate crops produced by 14 AIC methods. A systematic study on CropBench reveals that the existing criteria, ranging from IoU to off-the-shelf IAA assessment models, only exhibit limited alignment with human preferences. Building upon CropBench, we further propose CropReward, an MLLM-based reward model that evaluates a candidate crop by comparing with the original uncropped image and outputs a score indicating explicit aesthetic improvement or degradation relative to the original. In addition, we employ CropReward to construct CropReward-300k, the largest AIC dataset to date, and further develop CropAlign, a preference-aligned AIC model trained solely with CropReward supervision, without human-annotated crop boxes. Experimental results demonstrate that CropReward aligns with human preferences significantly better than AIC existing criteria. CropAlign also surpasses the state-of-the-art AIC methods, demonstrating the advantage of CropReward in guiding AIC model optimization.
Yuqiang Lin, Kehua Chen, Sam Lockyer, Arjun Yadav, Mingxuan Sui, Shucheng Zhang, Yan Shi, Bingzhang Wang 等 15 位作者
Traffic Anomaly Understanding (TAU) is important for traffic safety in Intelligent Transportation Systems. Recent vision-language models (VLMs) have shown strong capabilities in video understanding. However, progress on TAU remains limited due to the lack of benchmarks and task-specific methodologies. To address this limitation, we introduce Roundabout-TAU, a dataset constructed from real-world roundabout videos collected in collaboration with the City of Carmel, Indiana. The dataset contains 342 clips and is annotated with more than 2,000 question-answer pairs covering multiple aspects of traffic anomaly understanding. Building on this benchmark, we propose TAU-R1, a two-layer vision-language framework for TAU. The first layer is a lightweight anomaly classifier that performs coarse anomaly categorisation, while the second layer is a larger anomaly reasoner that generates detailed event summaries. To improve task-specific reasoning, we introduce a two-stage training strategy consisting of decomposed-QA-enhanced supervised fine-tuning followed by TAU-GRPO, a GRPO-based post-training method with TAU-specific reward functions. Experimental results show that TAU-R1 achieves strong performance on both anomaly classification and reasoning tasks while maintaining deployment efficiency. The dataset and code are available at: https://github.com/siri-rouser/TAU-R1
Novel view synthesis (NVS) in degraded scenes aims to reconstruct 3D scenes from multi-view observations affected by various degradations, and to generate clear and photorealistic images at unseen viewpoints. Despite recent advances, existing methods often rely on degradation-specific models, limiting their generalization across diverse extreme environments. Generative models provide rich visual priors, but their outputs often suffer from pixel misalignment and color deviations, which can propagate into the reconstructed 3D scene and compromise reconstruction quality. Furthermore, the high computational and memory costs of 3D reconstruction hinder the practical application of NVS in extreme environments. To this end, we propose MLLM-GS, a multimodal large language model (MLLM) prior-guided framework for fast NVS in extreme environments. MLLM-GS consists of two branches: MLLM prior-guided view enhancement and fast 3D scene reconstruction. The former leverages the rich visual priors of a frozen MLLM to enhance views under different severe degradations and introduces a lightweight correction network to reduce pixel shifts and color distortions in MLLM outputs. With supervision from only a few paired clean views, this branch produces reliable enhanced views for all degraded observations, providing pseudo-GT views for subsequent 3D reconstruction. The latter improves 3DGS efficiency by optimizing rasterization, Gaussian memory layout, and parameter updates, reducing redundant computation, memory-access overhead, and optimization costs. Extensive experiments on multiple degraded scenes demonstrate that MLLM-GS effectively handles various severe degradations and achieves high-quality NVS with significantly lower reconstruction time and memory consumption. Project Page: https://wzlwlm13.github.io/MLLM-GS
We present an inference-time approach that enhances the visual diversity of image sets generated by text-to-image models, without compromising the average image quality. Our method leverages the intrinsic geometry of hidden representations in pretrained models to diversify images generated from the same prompt. First, we introduce a token repulsion mechanism that increases the angle between feature tokens across samples. It takes a negative Riemannian gradient step on a von Mises-Fisher mixture distribution of tokens on the unit hypersphere. Second, to preserve the quality of each output image, we propose a token density preservation mechanism that projects the repulsion-induced prediction change onto a half-space aligned with the model's score direction. This encourages the updates to remain within the high-density regions learned by the generative model. We further extend this mechanism to the output of each intermediate layer, enabling finer-grained iterative guidance. We evaluate our method on three types of pretrained text-to-image models. Experiments show substantial gains in visual diversity under identical prompts, while maintaining a comparable level of image fidelity.
Shuijin Lin, Shaozhen Yan, Tuo Liu, Haifeng Wang, Chunfeng Lian, Jie Lu, Jianhua Ma, Lin Xie
The biological definition of Alzheimer's disease (AD) relies on multi-modal neuroimaging, yet the clinical utility of positron emission tomography (PET) is limited by cost and radiation exposure, hindering early screening at preclinical or prodromal stages. While generative models offer a promising alternative by synthesizing PET from magnetic resonance imaging (MRI), achieving subject-specific precision remains a primary challenge. Here, we introduce DIReCT$++$, a Domain-Informed ReCTified flow model for synthesizing multi-tracer PET from MRI combined with fundamental clinical information. Our approach integrates a 3D rectified flow architecture to capture complex cross-modal and cross-tracer relationships with a domain-adapted vision-language model (BiomedCLIP) that provides text-guided, personalized generation using clinical scores and imaging knowledge. Extensive evaluations on multi-center datasets demonstrate that DIReCT$++$ not only produces synthetic PET images ($^{18}$F-AV-45 and $^{18}$F-FDG) of superior fidelity and generalizability but also accurately recapitulates disease-specific patterns. Crucially, combining these synthesized PET images with MRI enables precise personalized stratification of mild cognitive impairment (MCI), advancing a scalable, data-efficient tool for the early diagnosis and prognostic prediction of AD. The source code will be released on https://github.com/ladderlab-xjtu/DIReCT-PLUS.
Yiqi Nie, Fei Wang, Junjie Chen, Kun Li, Yudi Cai, Dan Guo, Chenglong Li, Meng Wang
Multimodal Emotional Reappraisal (MER) aims to shift affective interpretation by rewriting and editing a situation without altering its underlying semantic context. Despite advances in controllable editing and sentiment transfer, MER remains challenging under coupled multimodal constraints, especially in internet memes where images, embedded text, and layout jointly encode affective meaning. To bridge this gap, we present MER-Bench, a systematic benchmark for MER on image-text memes, containing over 3,000 real-world memes with fine-grained annotations for emotion transformation and structure-preserving multimodal editing. MER-Bench spans diverse visual, textual, and structural settings and is paired with a structured MLLM-as-a-Judge evaluation suite, comprising eight fine-grained metrics and three aggregated objectives for emotional alignment, content fidelity, and transformation strength. Extensive evaluation of 14 advanced multimodal models reveals that: (1) current models struggle to jointly balance affect, semantics, structure, and text; (2) strong visual editing does not imply effective MER, with poor embedded text rendering and affect-aware rewriting remaining key bottlenecks. Our results identify MER as a critical capability gap in current multimodal generation systems and highlight affective image-text reasoning as a key challenge for future work. Project Page: https://github.com/one-seven17/MER-Bench
Normalizing flows (NFs) provide exact likelihoods and deterministic invertible sampling, but have historically lagged behind diffusion models for large-scale image generation. We identify a key obstacle: NFs are required to learn a single invertible transport over the full ambient space, making them highly sensitive to high-dimensional representations. This leads to a semantic-capacity mismatch in modern visual representation spaces, where semantic information is compact but encoded in overcomplete features. We propose SRC-Flow, which introduces a Semantic Representation Compressor (SRC) to compact high-dimensional RAE features into a low-dimensional semantic space before flow modeling and preserve reconstruction through the frozen RAE decoder. This compact space reduces the modeling burden of NFs and enables effective likelihood-based generation in semantic representation space. We further adopt constant noise regularization tailored to the fixed unconditional bijection learned by flows. On ImageNet 256x256 and 512x512, SRC-Flow achieves state-of-the-art generation quality among normalizing flow methods, with gFID scores of 1.65 and 2.07 under classifier-free guidance, while retaining exact likelihood computation in the compact semantic representation space and deterministic invertible sampling at the flow level. Codes and models will be released in the future.
Yunpeng Gong, Can Yang, Qingyuan Zeng, Dejun Xu, Zhiming Luo, Zhenzhong Wang, Min Jiang
Cross-modal person re-identification (ReID) adversarial attacks face a transferability challenge: existing methods may exhibit source-pair overfitting, where learned perturbations become overly dependent on the feature statistics of the source modality pair, thereby limiting their transferability under heterogeneous domain shifts. We argue that a key challenge is reducing source-pair-specific bias while maintaining attack effectiveness. To address this, we propose Multiform Attack (MA). The framework first learns a universal perturbation via Mahalanobis-guided gradient optimization. Since this perturbation may still retain source-pair-specific bias, the second stage performs sparse refinement through attention-shift-filtered multi-objective evolutionary search, leveraging heterogeneous source distributions in transfer settings. By separating universal perturbation learning from sparse evolutionary refinement, MA aims to reduce reliance on source-pair-specific patterns and improve transferability to unseen domains. Extensive experiments demonstrate strong attack effectiveness on cross-modal ReID benchmarks and improved transferability in the evaluated held-out-domain settings, with additional evaluations across model architectures.
Qifan Wang, Qihang Fang, Boao Shi, Yushi Du, Pei Zhou, Siyan Dong, Yanchao Yang
Modeling human-object interaction (HOI) in real environments requires jointly representing whole-body motion, manipulated objects, and scene geometry, yet existing datasets trade off scalable capture against structured, scene-aligned supervision. We introduce SituHOI, a large-scale RGB-D dataset containing 18,429 annotated clips across 16 indoor rooms and approximately 86 hours of quality-screened activity. It pairs recordings from a single moving RGB-D camera with static scene scans, object geometry, language descriptions, and estimated human and object trajectories. Using the scans as persistent references, our pipeline combines camera localization and depth-based metric alignment with human mesh recovery, object tracking, and temporal refinement to register all geometric annotations in a shared floor-aligned frame. These data support FlexHOI-Bench, a goal-conditioned joint motion prediction benchmark with variable observation and prediction durations. Given observed histories, scene and object geometry, and a terminal object pose, models predict future human and object motion. We provide SharedHOI as a reference predictor and evaluate world-space accuracy and human-object coordination. On a MoCap subset, our pipeline reduces human G-MPJPE from 195.4 to 68.3 mm and object translation error from 43.0 to 28.7 mm. SharedHOI leads four adapted baselines on all three aggregate FlexHOI-Bench metrics. Data will be released upon acceptance. An anonymized code repository is available at https://anonymous.4open.science/r/iclr2027-anonymous-code-1DFB.
As vision language models are increasingly deployed in clinical diagnosis, understanding how they internally resolve competing visual and textual signals becomes a safety imperative. Existing mechanistic analyses remain confined to unimodal text and offer no explanation for why a single misleading sentence can override a correct image based diagnosis, or why a model commits to a confident answer despite insufficient visual evidence. We find that these two safety risks, arbitration failure where textual context overrides visual grounding and brake failure where the model commits without adequate evidence, are mediated by spatially disjoint attention head populations: arbitration heads form a mid-to-deep wideband reflecting cross-layer evidence competition, while brake heads concentrate in a narrow middle-to-late layer band that regulates evidence sufficiency and abstention behavior. To ground these observations in causal circuitry, we introduce CRAFT, which localizes each failure mode to a minimal causal head set via dual criteria and verifies necessity and sufficiency through temporal probes and Tuned Lens trajectory analysis. Excising arbitration heads sharply reduces conflict following with negligible degradation on clean inputs, while excising brake heads restores appropriate abstention under degraded visual evidence. The two interventions target spatially disjoint head sets and produce distinct corrective effects, underscoring the mechanistic separability of the failure modes. Experiments across multiple medical VQA benchmarks and VLM architectures validate both the localization and interventions, demonstrating that the identified heads causally drive each failure mode and that targeted modulation generalises without retraining. The code is available at https://github.com/zhcz328/CRAFT.
Wenjiang Zhang, Sifeng Wang, Yuwei Su, Xinyu Li, Chen Zhang, Suyu Zhong
Visual neural decoding seeks to reconstruct or infer perceived visual stimuli from brain activity patterns, providing critical insights into human cognition and enabling transformative applications in brain-computer interfaces and artificial intelligence. Current approaches, however, remain constrained by the scarcity of high-quality stimulus-brain response pairs and the inherent semantic mismatch between neural representations and visual content. Inspired by perceptual variability and co-adaptive strategy of the biological systems, we propose a novel self-supervised architecture, named NeuroBridge, which integrates Cognitive Prior Augmentation (CPA) with Shared Semantic Projector (SSP) to promote effective cross-modality alignment. Specifically, CPA simulates perceptual variability by applying asymmetric, modality-specific transformations to both EEG signals and images, enhancing semantic diversity. Unlike previous approaches, SSP establishes a bidirectional alignment process through a co-adaptive strategy, which mutually aligns features from two modalities into a shared semantic space for effective cross-modal learning. NeuroBridge surpasses previous state-of-the-art methods under both intra-subject and inter-subject settings. In the intra-subject scenario, it achieves the improvements of 12.3% in top-1 accuracy and 10.2% in top-5 accuracy, reaching 63.2% and 89.9% respectively on a 200-way zero-shot retrieval task. Extensive experiments demonstrate the effectiveness, robustness, and scalability of the proposed framework for neural visual decoding.
Currently, there is a gap in the field of ultra-high-definition (UHD) video dehazing due to the lack of a benchmark for evaluation. Furthermore, existing video dehazing methods cannot run on consumer-grade GPUs when processing continuous UHD sequences of 3--5 frames at a time. In this paper, we address both issues with a new benchmark and an efficient method. Our key observation is that atmospheric dehazing reduces to a per-pixel affine transform governed by the low-frequency depth field, which can be compactly encoded in bilateral grids whose prediction cost is decoupled from the output resolution. Building on this, we propose LiBrA-Net, which factorizes the spatiotemporal affine field into a spatial--color and a temporal bilateral sub-grid predicted at a fixed low resolution, fuses their coefficients in the $\mathfrak{gl}(3)$ Lie algebra under group-theoretic regularization, maps the result to invertible $GL(3)$ transforms via a Cayley parameterization, and restores high-frequency detail through a lightweight input-guided branch. We further release UHV-4K, the first paired 4K video dehazing benchmark with depth, transmission, and optical-flow annotations on every frame. Across UHV-4K, REVIDE, and HazeWorld, LiBrA-Net sets a new state of the art among compared video dehazing methods while running native 4K at 25FPS on a single GPU with only 6.12M parameters. Code is available at https://github.com/ycwang31/LiBrA-Net.
Kaisong Ji, Ziyang Xu, Chang Bian, Yu Zhou, Wenyu Liu, Xinggang Wang
Zero-shot anomaly detection aims to transfer anomaly discrimination from auxiliary training domains to unseen categories while localizing subtle defects. A key challenge is to construct discriminative references that combine contextual representations with complementary local evidence. We present Cross-Encoder Anchor Refinement (CEAR), which connects two visual encoding paths for the same image at the anchor stage, before normal and anomalous prototypes are constructed. An independent convolutional encoder pretrained for industrial product recognition builds an image-specific memory shared across selected vision transformer (ViT) layers. Spatial anchors read from ViT patch features query this memory through cross-attention and undergo residual refinement. Image-conditioned normal and anomalous queries then gate anchor aggregation to construct prototypes for anomaly scoring. Across 13 industrial and medical datasets, CEAR achieves the best performance in 57 of 68 evaluations spanning image-level detection and pixel-level localization. Notably, CEAR surpasses all other state-of-the-art methods in 41 of 44 pixel-level comparisons. Across 11 localization datasets, mean gains in pixel-level maximum F1 score and average precision are 4.45 and 3.79 percentage points, respectively. This study shows how cross-encoder interaction at the anchor stage can integrate complementary visual representations for fine-grained anomaly localization across domains.
Minqi Wang, Jie Peng, Hejun Dong, Ruchang Yao, Puyi Wang, Siyuan Huang, Jihai Zhang, Yafu Li 等 11 位作者
Video world models generate future visual observations from camera or action inputs, with the goal of synthesizing persistent 3D worlds that can be explored across time and space. Existing benchmarks typically evaluate short interactions or rely on camera trajectories fixed before generation. The performance of video world models under long-horizon exploration with extensive camera motion remains under-explored. We introduce WorldProbe, a diagnostic benchmark that uses a goal-conditioned multimodal agent to adapt camera actions online during generation. WorldProbe comprises 164 test cases across four probing families. Across 10 state-of-the-art models, we generate more than 1,600 rollouts totaling approximately 800 minutes, with horizons up to 180 seconds. Our evaluation across four dimensions reveals both the degradation of local measures and the emergence of long-range inconsistencies. We find that recent models sustain visual and content quality for longer, while long-range memory and evolution remain weak. These results expose a gap between locally stable generation and persistent world consistency that is largely missed by existing benchmarks. Code and benchmark resources are available at https://github.com/Untitled-W/WorldProbe.
Kiljoon Han, Kyumin Hwang, Wonheok Choi, Jeonghoon Kim, Sunghoom Im
Iterative refinement models have significantly improved stereo matching accuracy, but their rigid, fixed update schedules incur heavy computational costs and a strict trade-off between speed and accuracy. This challenge is magnified in video streams, where traditional online methods resort to reducing feature channels or relying on expensive temporal warping. To address this, we propose \textbf{DEQ-Stereo}, a novel video stereo matching framework based on the Deep Equilibrium (DEQ) paradigm. Instead of unrolling a fixed number of iterations, DEQ-Stereo reformulates the refinement as an implicit fixed-point problem, adaptively terminating once convergence is met. To ensure fast and stable intra-frame convergence, we introduce the \textbf{Convolutional Simplified Gated Unit (ConvSGU)}, which eliminates unnecessary non-linearities by modeling the latent state as a gated, element-wise sum of inputs. Furthermore, to exploit cross-frame temporal coherence, we propose \textbf{Temporal State Transfer (TST)}, a warm-start strategy that propagates the converged hidden state to the next frame without requiring explicit pose or warping priors. Extensive evaluations demonstrate that DEQ-Stereo delivers state-of-the-art performance with a superior accuracy-to-latency trade-off.
Agentic image editors decompose complex requests into steps, but when a step fails, they mostly retry with the same narrow set of tools and a judge that scores each output by eye directly. We argue that such agents are bounded by two spaces: the action space of what the agent can do, and the observation space of what its judge can see. On our expert-annotated benchmark, a general-purpose editor yields no usable result in over 40% of the cases, and no judge is reliable across edit types. We present EditAgent, which expands both spaces. It routes each step to general editors and specialized methods, and attaches to each candidate a verification record. With these records, off-the-shelf VLM judges outperform finetuned reward models. Candidates are generated in parallel and judged in one listwise call: experts prefer this to sequential refinement in 78.5% of decided comparisons, at no more than half the judge calls. EditAgent improves two open-weight editors on three benchmarks and serves as a data engine for training stronger editors.