CVPR 2026 论文列表

Video temporal grounding, the task of localizing the start and end times of a natural language query in untrimmed video, requires capturing both global context and fine-grained temporal detail. This challenge is particularly pronounced in long videos, where existing methods often compromise temporal fidelity by over-downsampling or using fixed windows. We present HieraMamba, a hierarchical architecture that preserves temporal structure and semantic richness across scales. At its core are Anchor-MambaPooling (AMP) blocks, which utilize Mamba's selective scanning to produce compact anchor tokens summarizing video content across scales. We further introduce anchor-conditioned and segment-pooled contrastive losses-two complementary objectives that encourage anchors to retain local detail while remaining globally discriminative. HieraMamba sets a new state-of-the-art on Ego4D-NLQ, MAD, and TACoS, demonstrating precise, temporally faithful localization in long, untrimmed videos.

Recent Vision-Language-Action (VLA) models reformulate vision-language models by tuning them with millions of robotic demonstrations. While they perform well when fine-tuned for a single embodiment or task family, extending them to multi-skill settings remains challenging: directly merging VLA experts trained on different tasks results in near-zero success rates. This raises a fundamental question: what prevents VLAs from mastering multiple skills within one model? With an empirical decomposition of learnable parameters during VLA fine-tuning, we identify two key sources of non-mergeability:(1) Finetuning drives LoRA adapters in the VLM backbone toward divergent, task-specific directions beyond the capacity of existing merging methods to unify.(2) Action experts develop inter-block dependencies through self-attention feedback, causing task information to spread across layers and preventing modular recombination.To address these challenges, we present MergeVLA, a merging-oriented VLA architecture that preserves mergeability by design.MergeVLA introduces sparsely activated LoRA adapters via task masks to retain consistent parameters and reduce irreconcilable conflicts in the VLM.Its action expert replaces self-attention with cross-attention-only blocks to keep specialization localized and composable.When the task is unknown, it uses a test-time task router to adaptively select the appropriate task mask and expert head from the initial observation, enabling unsupervised task inference.Across LIBERO, LIBERO-Plus, RoboTwin, and multi-task experiments on the real SO101 robotic arm, MergeVLA achieves performance comparable to or even exceeding individually finetuned experts, demonstrating robust generalization across tasks, embodiments, and environments. Project page: https://mergevla.github.io/

Incremental Object Detection (IOD) aims to continuously learn new object categories without forgetting previously learned ones. Recently, prompt-based methods have gained popularity for their replay-free design and parameter efficiency. However, due to prompt coupling and prompt drift, these methods often suffer from prompt degradation during continual adaptation. To address these issues, we propose a novel prompt-decoupled framework called PDP. PDP innovatively designs a dual-pool prompt decoupling paradigm, which consists of a shared pool used to capture task-general knowledge for forward transfer, and a private pool used to learn task-specific discriminative features. This paradigm explicitly separates task-general and task-specific prompts, preventing interference between prompts and mitigating prompt coupling. In addition, to counteract prompt drift resulting from inconsistent supervision where old foreground objects are treated as background in subsequent tasks, PDP introduces a Prototypical Pseudo-Label Generation (PPG) module. PPG can dynamically update the class prototype space during training and use the class prototypes to further filter valuable pseudo-labels, maintaining supervisory signal consistency throughout the incremental process. PDP achieves state-of-the-art performance on MS-COCO (with a 9.2% AP improvement) and PASCAL VOC (with a 3.3% AP improvement) benchmarks, highlighting its potential in balancing stability and plasticity. The code and dataset are released at: https://github.com/zyt95579/PDP_IOD.

Referring Video Object Segmentation (RVOS) aims to segment target objects in videos based on natural language descriptions. However, fixed keyframe-based approaches that couple a vision language model with a separate propagation module often fail to capture rapidly changing spatiotemporal dynamics and to handle queries requiring multi-step reasoning, leading to sharp performance drops on motion-intensive and reasoning-oriented videos beyond static RVOS benchmarks. To address these limitations, we propose VIRST (Video-Instructed Reasoning Assistant for Spatio-Temporal Segmentation), an end-to-end framework that unifies global video reasoning and pixel-level mask prediction within a single model. VIRST bridges semantic and segmentation representations through the Spatio-Temporal Fusion (STF), which fuses segmentation-aware video features into the vision-language backbone, and employs the Temporal Dynamic Anchor Updater to maintain temporally adjacent anchor frames that provide stable temporal cues under large motion, occlusion, and reappearance. This unified design achieves state-of-the-art results across diverse RVOS benchmarks under realistic and challenging conditions, demonstrating strong generalization to both referring and reasoning oriented settings.

Video reasoning has advanced with large multimodal models (LMMs), yet their inference is often a single pass that returns an answer without verifying whether the reasoning is evidence-aligned. We introduce **Reinforce to Learn, Elect to Reason (RLER)**, a dual paradigm that decouples learning to produce evidence from obtaining a reliable answer. In **RLER-Training**, we optimize the policy with group-relative reinforcement learning (RL) and 3 novel task-driven rewards: Frame-sensitive reward grounds reasoning on explicit key frames, Think-transparency reward shapes readable and parsable reasoning traces, and Anti-repetition reward boosts information density. These signals teach the model to emit structured, machine-checkable evidence and potentiate reasoning capabilities. In **RLER-Inference**, we apply a train-free orchestrator that generates a small set of diverse candidates, parses their answers and cited frames, scores them by evidence consistency, confidence, transparency, and non-redundancy, and then performs a robust evidence-weighted election. This closes the loop between producing and using evidence, improving reliability and interpretability without enlarging the model. We comprehensively evaluate RLER against various open-source and RL-based LMMs on 8 representative benchmarks. RLER achieves state of the art across all benchmarks and delivers an average improvement of 6.3 % over base models, while using on average 3.1 candidates per question, indicating a favorable balance between compute and quality. The results support a simple thesis: making evidence explicit during learning and electing by evidence during inference is a robust path to trustworthy video reasoning.

The Abstraction and Reasoning Corpus (ARC) is designed to promote research on abstract reasoning, a fundamental aspect of human intelligence. Common approaches to ARC treat it as a language-oriented problem, addressed by large language models (LLMs) or recurrent reasoning models. However, although the puzzle-like tasks in ARC are inherently visual, existing research has rarely approached the problem from a vision-centric perspective. In this work, we formulate ARC within a vision paradigm, framing it as an image-to-image translation problem. To incorporate visual priors, we represent the inputs on a "canvas" that can be processed like natural images. It is then straightforward for us to apply standard vision architectures, such as a vanilla Vision Transformer (ViT), to perform image-to-image mapping. Our model is trained from scratch solely on ARC data and generalizes to unseen tasks through test-time training. Our framework, termed Vision ARC (VARC), achieves 60.4% accuracy on the ARC-1 benchmark, substantially outperforming existing methods that are also trained from scratch. Our results are competitive with those of leading LLMs and close the gap to average human performance.

Multimodal Emotion Analysis (MEA) is crucial for human-centric AI, yet current methods struggle with two core challenges: the sparse nature of emotional cues across modalities and their inherent temporal asynchrony. Existing approaches, which often rely on implicit fusion, consequently suffer from diluted salient features and entangled representations. To address this issue, we propose EmoThinker, a new framework that advances MEA through explicit, structured reasoning. Our method introduces a structural token selection mechanism to concentrate on pivotal facial regions while refining background context, enhancing visual saliency and efficiency. For audio, an audio evidence extractor aggregates critical paralinguistic features into compact, emotion-rich tokens. More importantly, we enable step-by-step reasoning by constructing a Chain-of-Emotion-Thought dataset, which provides fine-grained annotations for disentangling asynchronous cues and resolving inter-modal conflicts. By decoupling evidence acquisition from reasoning, EmoThinker achieves a more interpretable and robust emotion analysis. Extensive experiments on multiple benchmarks demonstrate that our framework achieves new state-of-the-art performance.

Large Multimodal Models (LMMs) have attained impressive achievements in multimodal processing tasks, yet their massive memory demands pose major obstacles to deployment on resource-limited devices. Singular Value Decomposition (SVD) has emerged as a promising compression technique for LMMs, delivering substantial reductions in memory overhead. However, existing SVD-based methods often struggle to effectively alleviate the errors caused by SVD truncation, resulting in a noticeable performance gap when compared to the original models. Moreover, adopting a uniform compression ratio across all transformer layers fails to consider the varying importance of different layers. To tackle these challenges, we propose AdaSVD, an adaptive SVD-based LMM compression approach. Specifically, AdaSVD introduces adaComp, which adaptively compensates for SVD truncation errors by alternately updating the singular matrices. Additionally, AdaSVD introduces adaCR, which adaptively assigns layer-specific compression ratios according to the relative importance of each layer. Comprehensive experiments across multiple LMM families show the effectiveness of AdaSVD, achieving better performance while significantly reducing memory requirements. We will make all the code and models of AdaSVD publicly available.

Pose stylization, which aims to synthesize stylized content aligning with target poses, serves as a fundamental task across 2D, 3D, and video domains. In the 3D realm, prevailing approaches typically rely on a cascade pipeline: first manipulating the image pose via 2D foundation models and subsequently lifting it into 3D representations. However, this paradigm limits the precision and diversity of the 3d pose stylization. To this end, we propose a novel paradigm for 3D pose stylization that unifies pose stylization and 3D generation within a cohesive framework. This integration minimizes the risk of cumulative errors and enhances the model's efficiency and effectiveness. In addition, diverging from previous works that typically utilize 2D skeleton images as guidance, we directly utilize the 3D skeleton because it can provide a more accurate representation of 3D spatial and topological relationships, which significantly enhances the model's capacity to achieve richer and more precise pose stylization. Moreover, we develop a scalable data engine to construct a large-scale dataset of "Image-Skeleton-Mesh" triplets, enabling the model to jointly learn identity preservation and geometric alignment. Extensive experiments demonstrate that PoseMaster significantly outperforms state-of-the-art methods in both qualitative and quantitative metrics. Owing to the strict spatial alignment between the generated 3D meshes and the conditioning skeletons, PoseMaster enables the direct creation of animatable assets when coupled with automated skinning models, highlighting its compelling potential for automated character rigging.

We present Scalable Pixel-anchored End-to-end Diffusion (SpeeDiff), a latent diffusion method that jointly trains the VAE and the diffusion model from scratch. In principle, joint training allows the diffusion loss gradient to directly guide the VAE encoder, encouraging the formation of a generation-friendly latent space and potentially yielding faster convergence than the conventional two-stage approach with a pretrained frozen VAE. However, a naive end-to-end implementation severely degrades performance, as unrestricted backpropagation of the diffusion loss leads to latent space collapse. Our main technical contribution is a simple yet effective Tweedie Pixel Reconstruction (TPR) loss, which provides additional pixel-level feedback by decoding a predicted clean latent from an intermediate noisy state using Tweedie's formula, thereby alleviating collapse. Furthermore, our method enables jointly scaling a fully transformer-based architecture and enhances representation alignment within the end-to-end framework. Our SpeeDiff-XL model achieves over 140x and 61x faster training compared to Vanilla SiT and REPA, respectively, while attaining an FID of 1.50 without guidance on ImageNet 256x256 generation. With a more efficient 32x compressed VAE, our model further reaches an FID of 1.53 without guidance on ImageNet 512x512 generation.

Collaborative fairness in federated learning ensures that clients are rewarded according to their contributions, thereby fostering long-term participation among clients. However, existing methods often under-reward low-contributing clients in the early training stage and neglect critical issues (consistency across local models or unequal neuron training frequencies in the global model), leading to degraded performance. To address these issues, we propose FedRAC, a novel Federated learning framework employing Rolling submodel Allocation for Collaborative fairness, without compromising the global model performance. First, we design a dynamic reputation calculation module with a theoretical fairness guarantee to generate reputations matching clients' contributions. It adjusts their reputations dynamically during training, ensuring low-contribution clients access better models in the early stages for adequate training. Second, we propose a rolling submodel allocation module that assigns high-performance submodels to clients with high reputations. This module prioritizes low-frequency neurons during allocation and is supported by theoretical convergence guarantees, ensuring that all neurons in the global model are fully trained. Extensive experiments are conducted on four public datasets to confirm the advantages of our method in terms of fairness and model accuracy. The source code is available at https://github.com/ZiHuiWangpcl1/FedRAC.

Text-to-video (T2V) generation has recently attracted considerable attention, resulting in the development of numerous high-quality datasets that have propelled progress in this area. However, existing public datasets are primarily composed of isolated text-video (T-V) pairs and thus fail to model inter-clip relationships. To address this limitation, we introduce CI-VID, a dataset that moves beyond isolated T2V generation toward text-and-video-to-video (T&V2V) generation. CI-VID contains over 340,000 samples, each comprising a semantically coherent video sequence with interleaved text captions that capture both clip-level content and inter-clip relationships. To validate its effectiveness, we design a comprehensive, multi-dimensional benchmark incorporating human evaluation, VLM-based assessment, and similarity-based metrics. Experimental results demonstrate that models trained on CI-VID significantly improve both accuracy and content consistency in multi-clip video generation. This enables the creation of story-driven content with smooth transitions and strong semantic coherence.

Recent years have seen impressive advances in text-to-image generation, with image generative or unified models, generating high-quality images from text. Yet these models still struggle with fine-grained color control, often failing to accurately match colors specified in text prompts. While existing benchmarks evaluate compositional reasoning and prompt adherence, none systematically assess the color precision. Color is fundamental to human visual perception and communication, and critical for applications from art to design workflows requiring brand consistency. However, current benchmarks either neglect color or rely on coarse assessments, missing key capabilities like interpreting RGB values or aligning with human expectations. To this end, we propose GenColorBench, the first comprehensive benchmark for T2I color generation, grounded in color systems like ISCC-NBS and CSS3/X11, including numerical colors which are absent elsewhere. With 44K color-focused prompts covering 400+ colors, it reveals models' true capabilities via perceptual and automated assessments. Evaluations of popular T2I models on GenColorBench reveal significant performance variation, indicating which color conventions models understand the best and exposing their failure modes. Furthermore, GenColorBench provides insights to guide future improvements in precise color generation. The benchmark is available at https://moatifbutt.github.io/gencolorbench/.

Achieving human-like spatial intelligence for vision-language models (VLMs) requires inferring 3D structures from 2D observations, recognizing object properties and relations in 3D space, and performing high-level spatial reasoning. In this paper, we propose a principled hierarchical framework that decomposes the learning of 3D spatial understanding in VLMs into four progressively complex levels, from geometric perception to abstract spatial reasoning. Guided by this framework, we construct an automated pipeline that processes approximately 5M images with over 45M objects to generate 3D spatial VQA pairs across diverse tasks and scenes for VLM supervised fine-tuning. We also develop an RGB-D VLM incorporating metric-scale point maps as auxiliary inputs to further enhance spatial understanding. Extensive experiments demonstrate that our approach achieves state-of-the-art performance on multiple spatial understanding and reasoning benchmarks, surpassing specialized spatial models and large proprietary systems such as Gemini-2.5-pro and GPT-5. Moreover, our analysis reveals clear dependencies among hierarchical task levels, offering new insights into how multi-level task design facilitates the emergence of 3D spatial intelligence.

Visual Autoregressive (VAR) modeling introduces a new paradigm for image generation by extending autoregressive mechanisms from next-token prediction to next-scale prediction, achieving remarkable performance. However, as the number of tokens increases rapidly with scale, processing full token maps at high resolution becomes computationally expensive. In addition, the inherently sequential nature of autoregressive modeling prevents parallel inference across scales, which further increases latency.To address these challenges, we propose LazyVAR, a training-free and plug-and-play acceleration method for VAR models. Our key observation is that the similarity of aggregated latent features between adjacent scales progressively increases with the scale index, reaching particularly higher values at larger scales. We treat this similarity as a Scale-wise Update Index, which serves as the pruning criterion. Consequently, more tokens can be pruned at larger scales to improve efficiency. Furthermore, we propose Parallel Group Decoding, which leverages this high similarity at larger scales to decode tokens from different scales in parallel, further accelerating inference. Experimental results show that the proposed LazyVAR achieves up to a 2.94x speedup of recent VAR models with negligible performance degradation, allowing the Infinity-2B text-to-image model to generate 1024x1024 resolution images within 0.5 seconds on a single RTX 4090 GPU.

Visible-thermal (RGB-T) object detection is a crucial technology for applications such as autonomous driving, where multimodal fusion enhances performance in challenging conditions like low light. However, the security of RGB-T detectors, particularly in the physical world, has been largely overlooked. This paper proposes a novel approach to RGB-T physical attacks using adversarial clothing with a non-overlapping RGB-T pattern (NORP). To simulate full-view (0^ \circ -360^ \circ ) RGB-T attacks, we construct 3D RGB-T models for human and adversarial clothing. NORP is a new adversarial pattern design using distinct visible and thermal materials without overlap, avoiding the light reduction in overlapping RGB-T patterns (ORP). To optimize the NORP on adversarial clothing, we propose a spatial discrete-continuous optimization (SDCO) method. We systematically evaluated our method on RGB-T detectors with different fusion architectures, demonstrating high attack success rates both in the digital and physical worlds. Additionally, we introduce a fusion-stage ensemble method that enhances the transferability of adversarial attacks across unseen RGB-T detectors with different fusion architectures.

Existing image SR and generic diffusion models transfer poorly to fluid SR: they are sampling-intensive, ignore physical constraints, and often yield spectral mismatch and spurious divergence. We address fluid super-resolution (SR) with **ReMD** (**Re**sidual-**M**ultigrid **D**iffusion), a physics-consistent diffusion framework. At each reverse step, ReMD performs a **multigrid residual correction**: the update direction is obtained by coupling data consistency with lightweight physics cues and then correcting the residual across scales; the multiscale hierarchy is instantiated with a **multi-wavelet** basis to capture both large structures and fine vortical details. This coarse-to-fine design accelerates convergence and preserves fine structures while remaining equation-free. Across atmospheric and oceanic benchmarks, ReMD improves accuracy and spectral fidelity, reduces divergence, and reaches comparable quality with markedly fewer sampling steps than diffusion baselines. Our results show that enforcing physics consistency **inside** the diffusion process via multigrid residual correction and multi-wavelet multiscale modeling is an effective route to efficient fluid SR.

With the rapid deployment of multimodal large language models (MLLMs), disputes regarding model ownership have become increasingly frequent, raising significant concerns about intellectual property protection. In this paper, we propose a framework for generating copyright triggers for MLLMs, enabling model publishers to embed verifiable ownership information into the model. The goal is to construct trigger images that elicit ownership-related textual responses exclusively in fine-tuned derivatives, while remaining inert in other non-derivative models. Our method constructs a tracking trigger image by treating the image as a learnable tensor, performing adversarial optimization with dual-injection of ownership-relevant semantic information. The first injection is achieved by enforcing textual consistency between the output of an auxiliary MLLM and a predefined ownership-relevant target text; the consistency loss is backpropagated to inject this ownership-related information into the image. The second injection is performed at the semantic-level by minimizing the distance between the CLIP features of the image and those of the target text. Furthermore, we introduce an additional adversarial training stage involving the auxiliary model. It is specifically trained to resist generating ownership-relevant target text, thereby enhancing robustness in heavily fine-tuned derivative models. Extensive experiments demonstrate the effectiveness of our dual-injection approach in tracking model lineage under various fine-tuning and domain-shift scenarios.

LiDAR-based perception is critical for autonomous driving due to its robustness to poor lighting and visibility conditions. Yet, current models operate under the closed-set assumption and often fail to recognize unexpected out-of-distribution (OOD) objects in the open world. Existing OOD scoring functions exhibit limited performance because they ignore the pronounced class imbalance inherent in LiDAR OOD detection and assume a uniform class distribution. To address this limitation, we propose the Neural Distribution Prior (NDP), a framework that models the distributional structure of network predictions and adaptively reweights OOD scores based on alignment with a learned distribution prior. NDP dynamically captures the logit distribution patterns of training data and corrects class-dependent confidence bias through an attention-based module. We further introduce a Perlin noise-based OOD synthesis strategy that generates diverse auxiliary OOD samples from input scans, enabling robust OOD training without external datasets. Extensive experiments on the SemanticKITTI and STU benchmarks demonstrate that NDP substantially improves OOD detection performance, achieving a point-level AP of 61.31% on the STU test set, which is more than 10xhigher than the previous best result. Our framework is compatible with various existing OOD scoring formulations, providing an effective solution for open-world LiDAR perception.