CVPR 2026 Papers

Large Vision-Language Models (LVLMs) have achieved remarkable success across diverse multimodal tasks, yet their practical deployment remains constrained by the computational burden arising from lengthy visual tokens. While visual token pruning has emerged as a promising solution, existing methods suffer from a fundamental limitation: once tokens are pruned at a specific layer, they become inaccessible to all subsequent layers, leading to premature information loss that can compromise model performance. Through empirical studies, we observe that different layers exhibit distinct visual region focus, indicating a varying optimal token subset across layers. Motivated by this insight, we propose Adaptive Layer-wise Visual Token Selection (ALVTS), a novel framework that breaks away from the conventional static token pruning paradigm. ALVTS incorporates a lightweight token selector to identify and route important tokens for further processing, while allowing less important tokens to skip the layer, thus minimizing computational redundancy. These two streams of tokens are seamlessly reintegrated before being fed into subsequent layers, facilitating adaptive compression across the entire model. Grounded in our importance consistency constrained low-rank approximation, the proposed token selection module closely emulates the full attention mechanism, effectively capturing its essential patterns without requiring model retraining. Extensive experiments on LLaVA-1.5, LLaVA-NeXT, and Qwen2.5-VL validate the effectiveness of our method. With an 89% token compression ratio, ALVTS retains 96.7% of the original model's accuracy, achieving a superior efficiency-accuracy trade-off for LVLM inference.

Interpretable-by-design models are gaining traction in computer vision because they provide faithful explanations for their predictions. In image classification, these models typically recover human-interpretable concepts from an image and use them for classification. Sparse concept recovery methods leverage the latent space of vision-language models to represent image embeddings as sparse combinations of concept embeddings. However, by ignoring the hierarchical structure of semantic concepts, these methods may produce correct predictions with explanations that are inconsistent with the hierarchy. In this work, we propose Hierarchical Concept Embedding & Pursuit (HCEP), a framework that induces a hierarchy of concept embeddings in the latent space and performs hierarchical sparse coding to recover the concepts present in an image. Given a hierarchy of semantic concepts, we introduce a geometric construction for the corresponding hierarchy of embeddings. Under the assumption that the true concepts form a rooted path in the hierarchy, we derive sufficient conditions for their recovery in the embedding space. We further show that hierarchical sparse coding reliably recovers hierarchical concept embeddings, whereas standard sparse coding fails. Experiments on real-world datasets show that HCEP improves concept precision and recall compared to existing methods while maintaining competitive classification accuracy. Moreover, when the number of samples available for concept estimation and classifier training is limited, HCEP achieves superior classification accuracy and concept recovery. Our results demonstrate that incorporating hierarchical structure into sparse concept recovery leads to more faithful and interpretable image classification models.

Self-reflection mechanisms that rely on purely text-based rethinking processes perform well in most multimodal tasks. However, when directly applied to long-form video understanding scenarios, they exhibit clear limitations. The fundamental reasons for this lie in two points: (1) long-form video understanding involves richer and more dynamic visual input, meaning rethinking only the text information is insufficient and necessitates a further rethinking process specifically targeting visual information; (2) purely text-based reflection mechanisms lack cross-modal interaction capabilities, preventing them from fully integrating visual information during reflection. Motivated by these insights, we propose REVISOR (REflective VIsual Segment Oriented Reasoning), a novel framework for tool-augmented multimodal reflection. REVISOR enables MLLMs to collaboratively construct introspective reflection processes across textual and visual modalities, significantly enhancing their reasoning capability for long-form video understanding. To ensure that REVISOR can learn to accurately review video segments highly relevant to the question during reinforcement learning, we designed the Dual Attribution Decoupled Reward (DADR) mechanism. Integrated into the GRPO training strategy, this mechanism enforces causal alignment between the model's reasoning and the selected video evidence. Notably, the REVISOR framework significantly enhances long-form video understanding capability of MLLMs without requiring supplementary supervised fine-tuning or external models, achieving impressive results on four benchmarks including VideoMME, LongVideoBench, MLVU, and LVBench.

Recent advances in robot manipulation have leveraged pre-trained vision-language models (VLMs) and explored integrating 3D spatial signals into these models for effective action prediction, giving rise to the promising vision-language-action (VLA) paradigm. However, most existing approaches overlook the importance of active perception: they typically rely on static, wrist-mounted cameras that provide an end-effector-centric viewpoint. As a result, these models are unable to adaptively select optimal viewpoints or resolutions during task execution, which significantly limits their performance in long-horizon tasks and fine-grained manipulation scenarios. To address these limitations, we propose ActiveVLA, a novel vision-language-action framework that empowers robots with active perception capabilities for high-precision, fine-grained manipulation. ActiveVLA adopts a coarse-to-fine paradigm, dividing the process into two stages:(1) Critical region localization. ActiveVLA projects 3D inputs onto multi-view 2D projections, identifies critical 3D regions, and supports dynamic spatial awareness.(2) Active perception optimization. Drawing on the localized critical regions, ActiveVLA uses an active view selection strategy to choose optimal viewpoints. These viewpoints aim to maximize amodal relevance and diversity while minimizing occlusions. Additionally, ActiveVLA applies a 3D zoom-in to improve resolution in key areas. Together, these steps enable finer-grained active perception for precise manipulation.Extensive experiments demonstrate that ActiveVLA achieves precise 3D manipulation and outperforms state-of-the-art baselines on three simulation benchmarks. Moreover, ActiveVLA transfers seamlessly to real-world scenarios, enabling robots to learn high-precision tasks in complex environments.

Visual Place Recognition (VPR) aims to match query images against a database using visual cues. State-of-the-art methods aggregate features from deep backbones to form global descriptors. Optimal transport-based aggregation methods reformulate feature-to-cluster assignment as a transport problem, but the standard Sinkhorn algorithm symmetrically treats source and target marginals, limiting effectiveness when image features and cluster centers exhibit substantially different distributions. We propose an asymmetric aggregation VPR method with geometric constraints for locally aggregated descriptors, called A2GC-VPR. Our method employs row-column normalization averaging with separate marginal calibration, enabling asymmetric matching that adapts to distributional discrepancies in visual place recognition. Geometric constraints are incorporated through learnable coordinate embeddings, computing compatibility scores fused with feature similarities, thereby promoting spatially proximal features to the same cluster and enhancing spatial awareness. Experimental results on MSLS, NordLand, and Pittsburgh datasets demonstrate superior performance, validating the effectiveness of our approach in improving matching accuracy and robustness. Our code is available at https://github.com/CV4RA/A2GC.

Foundation models have achieved success in computational pathology, demonstrating generalization across histopathology tasks. However, existing models overlook the heterogeneous and non-uniform organization of regions of interest (ROIs) because they rely on natural image backbones not tailored for tissue morphology. Consequently, they fail to capture the coherent tissue architecture beyond patches, limiting interpretability and clinical relevance. To address these challenges, we present Cross-modal Adaptive Region Encoder (CARE), a foundation model for pathology that partitions WSIs into several morphologically relevant regions. Specifically, CARE employs a two-stage pretraining strategy: (1) a self-supervised unimodal pretraining stage that learns morphological representations from 34,277 whole-slide images (WSIs) without segmentation annotations, and (2) a cross-modal alignment stage that leverages RNA and protein profiles to refine the construction and representation of adaptive regions. This molecular guidance enables CARE to identify biologically relevant patterns and generate irregular yet coherent tissue regions, selecting the representative area as ROI. CARE supports a broad range of pathology-related tasks, using either the ROI feature or the slide-level feature obtained by aggregating adaptive regions. Based on only one-tenth of the pretraining data typically used by mainstream foundation models, CARE achieves superior average performance across 33 downstream benchmarks, including morphological classification, molecular prediction, and survival analysis, and outperforms other foundation model baselines overall.

Large Vision-Language Models (LVLMs) can reason from image-text inputs and perform well in various multimodal tasks. Despite this success, they are affected by language priors and often produce hallucinations. Hallucinations denote generated content that is grammatically and syntactically coherent, yet bears no match or direct relevance to visual input. To address this problem, we propose Residual Decoding (ResDec). It is a novel training-free method that uses historical information to aid decoding. The method relies on the internal implicit reasoning mechanism and token logits evolution mechanism of LVLMs to correct biases. Extensive experiments demonstrate that ResDec effectively suppresses hallucinations induced by language priors, significantly improves visual grounding, and reduces object hallucinations. In addition to mitigating hallucinations, ResDec also performs exceptionally well on comprehensive LVLM benchmarks, highlighting its broad applicability.

Weakly Supervised Video Anomaly Detection (WSVAD) aims to localize abnormal segments using only video-level labels during training.Although the paradigm significantly reduces annotation costs, the coarse-grained labels fail to precisely describe the full videos, resulting in the introduction of substantial Weakly Labeled Information (WLI) during training. The presence of WLI makes it difficult for the model to accurately learn the boundary between normal and abnormal behaviors, leading to misclassifications and compromising the precision of anomaly localization.To tackle the challenges posed by WLI, we propose a triplet learning strategy that selects hard segments from normal videos as anchors. By combining contrastive learning with Multiple Instance Learning (MIL) strategy, we increase the projection distance between abnormal segments and anchor samples, to reduce the interference of WLI in anomaly detection.Moreover, considering that anomalies typically occur in dynamic foreground regions, we further design a motion-aware feature enhancement module that extracts dynamic areas within each video segment to emphasize the representation of critical features.This not only improves the accuracy of anchors in triplets, but also enhances the discriminative power of instance features in MIL. Extensive experiments on UCF-Crime, XD-Violence, and MSAD datasets demonstrate the effectiveness of our approach.

Electronic Navigational Charts (ENCs) are the safety-critical backbone of modern maritime navigation, yet it remains unclear whether multimodal large language models (MLLMs) can reliably interpret them. Unlike natural images or conventional charts, ENCs encode regulations, bathymetry, and route constraints via standardized vector symbols, scale-dependent rendering, and precise geometric structure---requiring specialized maritime expertise for interpretation. We introduce ENC-Bench, the first benchmark dedicated to professional ENC understanding. ENC-Bench contains 20,490 expert-validated samples from 840 authentic National Oceanic and Atmospheric Administration (NOAA) ENCs, organized into a three-level hierarchy: Perception (symbol and feature recognition), Spatial Reasoning (coordinate localization, bearing, distance), and Maritime Decision-Making (route legality, safety assessment, emergency planning under multiple constraints). All samples are generated from raw S-57 data through a calibrated vector-to-image pipeline with automated consistency checks and expert review. We evaluate 10 state-of-the-art MLLMs such as GPT-4o, Gemini 2.5, Qwen3-VL, InternVL-3, and GLM-4.5V, under a unified zero-shot protocol. The best model achieves only 47.88% accuracy, with systematic challenges in symbolic grounding, spatial computation, multi-constraint reasoning, and robustness to lighting and scale variations. By establishing the first rigorous ENC benchmark, we open a new research frontier at the intersection of specialized symbolic reasoning and safety-critical AI, providing essential infrastructure for advancing MLLMs toward professional maritime applications.

Video relighting with background replacement is a challenging task critical for applications in film production and creative media. Existing methods struggle to balance temporal consistency, spatial fidelity, and illumination naturalness. To address these issues, we introduce FlowPortal, a novel training-free flow-based video relighting framework. Our core innovation is a Residual-Corrected Flow mechanism that transforms a standard flow-based model into an editing model, guaranteeing perfect reconstruction when input conditions are identical and enabling faithful relighting when they differ, resulting in high structural consistency. This is further enhanced by a Decoupled Condition Design for precise lighting control and a High-Frequency Transfer mechanism for detail preservation. Additionally, a masking strategy isolates foreground relighting from background pure generation process. Experiments demonstrate that FlowPortal achieves superior performance in temporal coherence, structural preservation, and lighting realism, while maintaining high efficiency.

Transformers have achieved widespread and remarkable success, while the computational complexity of their attention modules remains a major bottleneck for vision tasks. Existing methods mainly employ 8-bit or 4-bit quantization to balance efficiency and accuracy. In this paper, with theoretical justification, we indicate that binarization of attention preserves the essential similarity relationships, and propose BinaryAttention, an effective method for fast and accurate 1-bit qk-attention. Specifically, we retain only the sign of queries and keys in computing the attention, and replace the floating dot products with bit-wise operations, significantly reducing the computational cost. We mitigate the inherent information loss under 1-bit quantization by incorporating a learnable bias, and enable end-to-end acceleration. To maintain the accuracy of attention, we adopt quantization-aware training and self-distillation techniques, mitigating quantization errors while ensuring sign-aligned similarity. BinaryAttention is more than 2xfaster than FlashAttention2 on A100 GPUs. Extensive experiments on vision transformer and diffusion transformer benchmarks demonstrate that BinaryAttention matches or even exceeds full-precision attention, validating its effectiveness. Our work provides a highly efficient and effective alternative to full-precision attention, pushing the frontier of low-bit transformers for vision tasks. The codes and models can be found at https://github.com/EdwardChasel/BinaryAttention.

Intricate correlations among atomic actions and inherent visual confounders in long-term action recognition (LTAR) contribute to the persistent challenges in this domain. While methods based on vision-language models that employ label text for supervision offer potential for handling visual confounders, their reliance on statistical correlations rather than causal mechanisms introduces two vulnerabilities: (1) spurious alignments with non-causal co-occurring visual features during cross-modal interaction, and (2) misinterpretation of codependencies among actions. To address these limitations, this paper introduces Progressive Cross-Modal Causal Intervention (PCMCI). PCMCI first mitigates co-occurrence hallucination via causal intervention grounded in optimal transport theory. Subsequently, an action relation-aware mechanism counters the backdoor path induced by codependency illusion, enabling the derivation of deconfounded text embeddings. Finally, these deconfounded embeddings serve as mediator to implement front-door adjustment to remove visual confounders. This progressive causal intervention framework facilitates learning robust representations for LTAR. Experiments on three long-term action benchmarks demonstrate the effectiveness of the proposed model.

Scalable Vector Graphics (SVG) are central to digital design due to their inherent scalability and editability. Despite significant advancements in content generation enabled by Visual Language Models (VLMs), existing text-to-SVG generation methods are limited by a core challenge: the autoregressive training process does not incorporate visual perception of the final rendered image, which fundamentally constrains generation quality. To address this limitation, we propose an Introspective SVG Generation Framework (IntroSVG). At its core, the framework instantiates a unified VLM that operates in a closed loop, assuming dual roles of both generator and critic. Specifically, through Supervised Fine-Tuning (SFT), the model learns to draft SVGs and to provide feedback on their rendered outputs; moreover, we systematically convert early-stage failures into high-quality error-correction training data, thereby enhancing model robustness. Subsequently, we leverage a high-capacity teacher VLM to construct a preference dataset and further align the generator's policy through Direct Preference Optimization (DPO). During inference, the optimized generator and critic operate collaboratively in an iterative "generate-review-refine" cycle, starting from imperfect intermediate drafts to autonomously improve output quality. Experimental results demonstrate that our method achieves state-of-the-art performance across several key evaluation metrics, generating SVGs with more complex structures, stronger semantic alignment, and greater editability. These results corroborate the effectiveness of incorporating explicit visual feedback into the generation loop.

All-in-One Image Restoration (AiOIR) has emerged as a promising yet challenging research direction. To address the core challenges of diverse degradation modeling and detail preservation, we propose UniLDiff, a unified framework enhanced with degradation- and detail-aware mechanisms, unlocking the power of diffusion priors for robust image restoration. Specifically, we introduce a Degradation-Aware Feature Fusion (DAFF) to dynamically inject low-quality features into each denoising step via decoupled fusion and adaptive modulation, enabling implicit modeling of diverse and compound degradations. Furthermore, we design a Detail-Aware Expert Module (DAEM) in the decoder to enhance texture and fine-structure recovery through expert routing. Extensive experiments across multi-task and mixed degradation settings demonstrate that our method consistently achieves state-of-the-art performance, highlighting the practical potential of diffusion priors for unified image restoration.

Multimodal image registration is a fundamental task and a prerequisite for downstream cross-modal analysis. Despite recent progress in shared feature extraction and multi-scale architectures, two key limitations remain. First, some methods use disentanglement to learn shared features but mainly regularize the shared part, allowing modality-private cues to leak into the shared space. Second, most multi-scale frameworks support only a single transformation type, limiting their applicability when global misalignment and local deformation coexist. To address these issues, we formulate hybrid multimodal registration as jointly learning a stable shared feature space and a unified hybrid transformation. Based on this view, we propose HRNet, a Hybrid Registration Network that couples representation disentanglement with hybrid parameter prediction. A shared backbone with Modality-Specific Batch Normalization (MSBN) extracts multi-scale features, while a Cross-scale Disentanglement and Adaptive Projection (CDAP) module suppresses modality-private cues and projects shared features into a stable subspace for matching. Built on this shared space, a Hybrid Parameter Prediction Module (HPPM) performs non-iterative coarse-to-fine estimation of global rigid parameters and deformation fields, which are fused into a coherent deformation field. Extensive experiments on four multimodal datasets demonstrate state-of-the-art performance on rigid and non-rigid registration tasks. The code is available at the project website.

We present Common Inpainted Objects In-N-Out of Context (COinCO), a novel dataset addressing the scarcity of out-of-context examples in existing vision datasets. By systematically replacing objects in COCO images through diffusion-based inpainting, we create 97,722 unique images featuring both contextually coherent and inconsistent scenes, enabling effective context learning. Each inpainted object is meticulously verified and categorized as in- or out-of-context through Large Vision Language Model assessments. We demonstrate three key tasks enabled by COinCO: (1) a fine-grained context reasoning approach that classifies objects as in- or out-of-context based on three criteria; (2) a novel Objects-from-Context prediction task that determines which new objects naturally belong in given scenes at both instance and clique level semantics, and (3) context-enhanced fake detection on state-of-the-art methods without fine-tuning. COinCO provides a controlled testbed with contextual variations, establishing a foundation for advancing context-aware visual understanding in computer vision, including image forensics. Code and dataset are available at https://co-in-co.github.io/.

Multimodal Large Language Models (MLLMs) have achieved remarkable progress in visual reasoning tasks, serving as the core perception engines for emerging AI agents like OpenClaw. While recent studies have introduced several effective image-based jailbreak methods, the vulnerabilities inherent in the video modality remain a largely unexplored frontier. As a pioneering effort to bridge this critical safety gap, we demonstrate that video-driven jailbreak attacks are significantly more effective and robust against pre-defined system prompts than their image-based counterparts. Specifically, we find that simply repeating a harmful image across multiple frames to construct a video can bypass the safety mechanisms of MLLMs. Our analysis reveals that unsafe videos are embedded more similarly to safe videos in the model's representation space than individual harmful images, making them harder to detect. Moreover, videos composed of identical frames are processed more like static images and are more likely to trigger safety defenses compared to videos with diverse frames. Motivated by these findings, we propose an algorithm that injects harmful content into typographic videos by interleaving it with diverse, safety-proximal frames, thereby evading MLLM safety alignment. Extensive experiments demonstrate that our approach achieves state-of-the-art jailbreak performance on several widely-used MLLMs (e.g., VideoLLaMA-2, Qwen2.5-VL, GPT-4.1, and Gemini-2.5) under 16 different safety policies.

Recent advances in generative modeling can create remarkably realistic synthetic videos, making it increasingly difficult for humans to distinguish them from real ones and necessitating reliable detection methods. However, two key limitations hinder the development of this field.**From the dataset perspective**, existing datasets are often limited in scale and constructed using outdated or narrowly scoped generative models, making it difficult to capture the diversity and rapid evolution of modern generative techniques. Moreover, the dataset construction process frequently prioritizes quantity over quality, neglecting essential aspects such as semantic diversity, scenario coverage, and technological representativeness. **From the benchmark perspective**, current benchmarks largely remain at the stage of dataset creation, leaving many fundamental issues and in-depth analysis yet to be systematically explored.Addressing this gap, we propose AIGVDBench, a benchmark designed to be comprehensive and representative, covering **31** state-of-the-art generation models and over **440,000** videos. By executing more than **1,500** evaluations on **33** existing detectors belonging to four distinct categories. This work presents **8 in-depth analyses** from multiple perspectives and identifying **4 novel findings** that offer valuable insights for the field. We hope this work provides a solid foundation for advancing the field of AI-generated video detection.

Large Vision-Language Models (LVLMs) have shown strong performance across various multimodal tasks by leveraging the reasoning capabilities of Large Language Models (LLMs). However, processing visually complex and information-rich images, such as infographics or document layouts, requires these models to generate a large number of visual tokens, leading to significant computational overhead. To address this, we propose PinPoint, a novel two-stage framework that first identifies instruction-relevant image regions and then refines them to extract fine-grained visual features for improved reasoning and efficiency. Central to our approach is the Instruction-Region Alignment, which localizes relevant regions using both visual input and textual instructions. We further introduce new annotations that provide richer ground-truth supervision for instruction-relevant regions across challenging VQA benchmarks: InfographicVQA, MultiPageDocVQA, and SinglePageDocVQA. Experimental results show that PinPoint not only achieves superior accuracy compared to existing methods but also reduces computational overhead by minimizing irrelevant visual tokens.