CVPR 2026 论文列表

CLIP achieves strong zero-shot image-text retrieval by aligning global vision and text representations, yet it falls behind on fine-grained tasks even when fine-tuned on long, detailed captions. In this work, we propose b-CLIP, a multi-granular text-conditioned contrastive learning framework designed to achieve hierarchical alignment between multiple textual granularities--from full captions to sentences and phrases--and their corresponding visual regions. For each level of granularity, b-CLIP uses cross-attention to dynamically pool image patches into contextualized visual embeddings. To address the semantic overlap inherent in this hierarchy, we introduce the b-Contextualized Contrastive Alignment Loss (b-CAL), which parameterizes the trade-off between query-specific matching and intra-image contextualization under both Cross-Entropy and Binary Cross-Entropy formulations. Through extensive experiments, we demonstrate that b-CLIP significantly improves dense alignment, achieving 91.8% T2I 92.3% I2T at R@1 on Urban1K and 30.9% on FG-OVD (Hard), setting state-of-the-art among methods trained without hard negatives. b-CLIP establishes a robust, adaptive baseline for dense vision-language correspondence. The code and models are released at \href https://github.com/fzohra/B-CLIP https://github.com/fzohra/B-CLIP .

Vision-language models (VLMs) demonstrate powerful capabilities in multimodal tasks. However, the large number of visual tokens imposes a significant computational cost. In this paper, we propose QuietPrune, a QUery-guIded Early Token Pruning method to remove redundant visual tokens within VLMs, thereby enhancing computational efficiency. Unlike previous late pruning methods, we recognize that implementing early pruning within the vision transformer (ViT) can achieve benefits in both latency reduction and accuracy maintenance. To address the semantic loss problem in early pruning, we design a lightweight adapter by performing an inverse transformation of the projector in VLMs. The proposed adapter converts the contextual query into a visual domain [Q-CLS] (Query [CLS]) token, providing textual guidance for ViT pruning. During pruning, we further introduce a semi-structured pruning scheme based on visual-textual relevance. Specifically, we group spatially adjacent 2 x2 tokens to accommodate the visual token merging operation prevalent in mainstream VLMs. We use the mean attention scores between the [Q-CLS] token and the visual tokens as the relevance metric for each group, avoiding additional computation. Pruning is then applied at the group level based on the relevance score, preserving positional continuity. After pruning, we aggregate the redundant tokens into a single token to maintain context cues. Our method achieves up to 19.0% reduction in prefill latency while outperforming 4.2% in accuracy on the recent Qwen3-VL and InternVL3 series compared to existing late pruning methods.

Vision-language-action (VLA) models achieve strong in-distribution performance but degrade sharply under novel camera viewpoints and visual perturbations. We show that this brittleness primarily arises from misalignment in Spatial Modeling, rather than Physical Modeling. To address this, we propose a one-shot adaptation framework that recalibrates visual representations through lightweight, learnable updates. Our first method, Feature Token Modulation (FTM), applies a global affine transformation to visual tokens and improves Libero viewpoint accuracy from 48.5% to 87.1% with only 4K parameters. Building on this, Feature Linear Adaptation (FLA) introduces low-rank updates to the ViT encoder, achieving 90.8% success with 4.7M parameters--matching LoRA-scale finetuning at far lower cost. Together, these results reveal substantial untapped robustness in pretrained VLA models and demonstrate that targeted, minimal visual adaptation is sufficient to restore viewpoint generalization.

Point-level weakly-supervised temporal sentiment localization (P-WTSL) aims to detect sentiment-relevant segments in untrimmed multimodal videos using timestamp sentiment annotations, which greatly reduces the costly frame-level labeling. To tackle the intrinsic challenges of imprecise sentiment boundaries in P-WTSL, we propose the Face-guided Sentiment Boundary Enhancement Network FSENet, a unified framework that leverages fine-grained facial features to guide sentiment localization. Specifically, our approach first introduces the Face-guided Sentiment Discovery (FSD) module, which integrates facial features into multimodal interaction via dual-branch modeling for effective sentiment stimuli clues; We then propose the Point-aware Sentiment Semantics Contrast (PSSC) strategy to discriminate sentiment semantics of candidate points (frame-level) near annotation points via contrastive learning, thereby enhancing the model's ability to recognize sentiment boundaries. At last, we design the Boundary-aware Sentiment Pseudo-label Generation (BSPG) approach to convert sparse point annotations into temporally smooth supervisory pseudo-labels. Extensive experiments and visualizations on the benchmark demonstrate the effectiveness of our framework, achieving state-of-the-art performance under full supervision, video-level, and point-level weak supervision, thereby showcasing the strong generalization ability of our FSENet across different annotation settings.

Multimodal Large Language Models (MLLMs) have recently demonstrated remarkable perceptual and reasoning abilities. However, they struggle to perceive fine-grained geometric structures, constraining their ability of geometric understanding and visual reasoning. To address this, we propose GeoTikzBridge, a framework that enhances local geometric perception and visual reasoning through tikz-based code generation. Within this framework, we build two models supported by two complementary datasets. The GeoTikzBridge-Base model is trained on GeoTikz-Base dataset, the largest image-to-tikz dataset to date with 2.5M pairs (16 xlarger than existing open-sourced datasets). This process is achieved via iterative data expansion and a localized geometric transformation strategy. Subsequently, GeoTikzBridge-Instruct is fine-tuned on GeoTikz-Instruct dataset which is the first instruction-augmented tikz dataset supporting visual reasoning. Extensive experimental results demonstrate that our models achieve state-of-the-art performance among open-sourced MLLMs. Furthermore, GeoTikzBridge models can serve as plug-and-play reasoning modules for any MLLM(LLM), enhancing reasoning performance in geometric problem-solving.

Humor, as both a creative human activity and a social binding mechanism, has long posed a major challenge for AI generation. Although producing humor requires complex cognitive reasoning and social understanding, theories of humor suggest that it follows learnable patterns and structures, making it theoretically possible for generative models to acquire them implicitly. In recent years, multimodal humor has become a prevalent form of online communication, especially among Gen Z, highlighting the need for AI systems capable of integrating visual understanding with humorous language generation. However, existing data-driven approaches lack explicit modeling or theoretical grounding of humor, often producing literal descriptions that fail to capture its underlying cognitive mechanisms, resulting in the generated image descriptions that are fluent but lack genuine humor or cognitive depth. To address this limitation, we propose HUMORCHAIN (HUmor-guided Multi-step Orchestrated Reasoning Chain for Image Captioning), a theory-guided multi-stage reasoning framework. It integrates visual semantic parsing, humor- and psychology-based reasoning, and a fine-tuned discriminator for humor evaluation, forming an interpretable and controllable cognitive reasoning chain. To the best of our knowledge, this is the first work to explicitly embed cognitive structures from humor theories into multimodal humor generation, enabling a structured reasoning process from visual understanding to humor creation. Experiments on Meme-Image-No-Text, Oogiri-GO, and OxfordTVG-HIC datasets show that HUMORCHAIN outperforms state-of-the-art baselines in human humor preference, Elo/BT scores, and semantic diversity, demonstrating that theory-driven structured reasoning enables large language models to generate humor aligned with human perception.

Recent Vision-Language Models (VLMs) have made remarkable progress in multimodal understanding tasks, yet their evaluation on long video understanding remains unreliable. Due to limited frame inputs, key frames necessary for answering the question may be missing from the model's input. However, models that truthfully refuse to answer under such uncertainty are marked as incorrect, while those that guess may coincidentally produce the correct answer and thus obtain deceptively higher accuracy, leading to misleading evaluation results and encouraging models to guess rather than respond honestly. To address this issue, we introduce VirtueBench, a benchmark explicitly designed to assess model trustworthiness under uncertainty. VirtueBench constructs multiple frame-sampling levels for each video and provides ground truths that distinguish between answerable and unanswerable cases. Evaluations on 25 open-source and commercial VLMs reveal distinct refusal behaviors across different model families, with refusal accuracy ranging from over 70% in the best models to nearly 0% in the worst. Moreover, most models exhibit a substantial drop in refusal when the prompt does not explicitly require them to do so. These findings highlight the need for developing trustworthy VLMs for multimodal understanding, guided by benchmarks and leaderboards that emphasize reliability and trustworthiness.

Unlike offline processing, streaming video vision-language models face two fundamental constraints: causality and accumulation. Causality prevents access to future frames that offline methods exploit, while accumulation causes tokens to grow unbounded, creating efficiency bottlenecks. However, existing approaches only regulate post-LLM kv-cache, leaving costly pre-LLM prefill unchanged. We introduce StreamingTOM, a training-free, plug-and-play two-stage framework that addresses both pre-LLM and post-LLM bottlenecks. Causal Temporal Reduction imposes a fixed per-frame budget and selects tokens based on adjacent-frame changes and token saliency, drastically reducing per-frame prefill cost by processing only a compact subset of visual tokens, ensuring predictable latency. Online Quantized Memory stores tokens in 4-bit format, retrieves relevant groups on demand, and dequantizes them, keeping the active kv-cache bounded regardless of stream length. Experiments demonstrate our method achieves 15.7x kv-cache compression ratio; compared to prior SOTA (LiveVLM), it delivers 1.2x lower peak memory and 2x faster TTFT. StreamingTOM achieves state-of-the-art accuracy among training-free methods with an average of 63.8% on offline benchmarks and 55.8% accuracy and 3.7 score on RVS. These results demonstrate that real-time streaming video understanding with bounded active memory is achievable without model retraining.

We present SWIM (See What I Mean), a novel training strategy that aligns vision and language representations to enable fine-grained object understanding solely from textual prompts. Unlike existing approaches that require explicit visual prompts, such as masks or points, SWIM leverages mask supervision only during training to guide cross-modal attention, allowing the model to automatically attend to the user-specified object at inference.Our cross-attention analysis of pretrained multimodal large language models (MLLMs) reveals a systematic discrepancy: Attribute words produce sharp, localized activations in the visual modality, whereas object nouns yield diffuse and scattered patterns due to semantic reference bias and distributed high-level representations. To address this misalignment, we construct NL-Refer, an enriched dataset, in which each object mask is paired with a precise natural language referring expression. SWIM extracts multi-layer cross-attention maps from object nouns and enforces spatial consistency with ground-truth masks. Experimental results demonstrate that SWIM substantially improves text-visual alignment and achieves superior performance over visual-prompt-based methods on fine-grained object understanding benchmarks. The code and data are available at https://github.com/HumanMLLM/SWIM.

Synthesis of diverse driving scenes serves as a crucial data augmentation technique for validating the robustness and generalizability of autonomous driving systems. Current methods aggregate high-definition (HD) maps and 3D bounding boxes as geometric conditions in diffusion models for conditional scene generation. However, implicit inter-condition dependency causes generation failures when control conditions change independently. Additionally, these methods suffer from insufficient details in both semantic and structural aspects. Specifically, brief and view-invariant captions restrict semantic contexts, resulting in weak background modeling. Meanwhile, the standard denoising loss with uniform spatial weighting neglects foreground structural details, causing visual distortions and blurriness. To address these challenges, we propose DrivePTS, which incorporates three key innovations. Firstly, our framework adopts a progressive learning strategy to mitigate inter-dependency between geometric conditions, reinforced by an explicit mutual information constraint. Secondly, a Vision-Language Model is utilized to generate multi-view hierarchical descriptions across six semantic aspects, providing fine-grained textual guidance. Thirdly, a frequency-guided structure loss is introduced to strengthen the model's sensitivity to high-frequency elements, improving foreground structural fidelity. Extensive experiments demonstrate that our DrivePTS achieves state-of-the-art fidelity and controllability in generating diverse driving scenes. Notably, DrivePTS successfully generates rare scenes where prior methods fail, highlighting its strong generalization ability.

Monocular metric depth estimation (MMDE) is a core challenge in computer vision, playing a pivotal role in real-world applications that demand accurate spatial understanding. Although prior works have shown promising zero-shot performance in MMDE, they often struggle with generalization across diverse camera types, such as fisheye and 360^\circ cameras. Recent advances have addressed this through unified camera representations or canonical representation spaces, but they require either including large FoV camera data during training or separately trained models for different domains. We propose UniDAC, an MMDE framework that presents universal robustness in all domains and generalizes across diverse cameras using a single model. We achieve this by decoupling metric depth estimation into relative depth prediction and spatially varying scale estimation, enabling robust performance across different domains. We propose a lightweight Depth-Guided Scale Estimation module that upsamples a coarse scale map to high resolution using the relative depth map as guidance to account for local scale variations. Furthermore, we introduce RoPE-\phi, a distortion-aware positional embedding that respects the spatial warping in Equi-Rectangular Projections (ERP) via latitude-aware weighting. UniDAC achieves state of the art (SoTA) in cross-camera generalization by consistently outperforming prior methods across all datasets.

Articulated 3D objects are critical for embodied AI, robotics, and scene understanding, yet creating simulation-ready assets remains labor-intensive and requires expert modeling of part hierarchies and motion structures. We introduce SPARK, a framework for reconstructing physically consistent, kinematic part-level articulated objects from a single RGB image. Given the image, we first leverage VLMs to extract coarse URDF parameters and generate part-level reference images. We then integrate the part-image guidance and the inferred structure graph into a diffusion transformer to synthesize consistent part and complete shapes of articulated objects. To further refine the URDF parameters, we explore a VLM-based reprediction strategy for discrete attribute refinement and a differentiable forward kinematics module for continuous parameter optimization under VLM-generated open-state supervision. Extensive experiments show that SPARK produces high-quality, simulation-ready articulated assets across diverse categories, enabling downstream applications such as robotic manipulation and interaction modeling.

Vision-and-Language Navigation (VLN) requires agents to accurately perceive complex visual environments and reason over navigation instructions and histories. However, existing methods passively process redundant visual inputs and treat all historical contexts indiscriminately, resulting in inefficient perception and unfocused reasoning. To address these challenges, we propose ProFocus, a training-free progressive framework that unifies Proactive Perception and Focused Reasoning through collaboration between large language models (LLMs) and vision-language models (VLMs). For proactive perception, ProFocus transforms panoramic observations into structured ego-centric semantic maps, enabling the orchestration agent to identify missing visual information needed for reliable decision-making, and to generate targeted visual queries with corresponding focus regions that guide the perception agent to acquire the required observations. For focused reasoning, we propose Branch-Diverse Monte Carlo Tree Search (BD-MCTS) to identify top-k high-value waypoints from extensive historical candidates. The decision agent focuses reasoning on the historical contexts associated with these waypoints, rather than considering all historical waypoints equally. Extensive experiments validate the effectiveness of ProFocus, achieving state-of-the-art performance among zero-shot methods on R2R and REVERIE benchmarks.

Vision-Language-Action (VLA) models are emerging as a promising paradigm for end-to-end autonomous driving, valued for their potential to leverage world knowledge and reason about complex driving scenes. However, existing methods suffer from two critical limitations: a persistent misalignment between language instructions and action outputs, and the inherent inefficiency of typical auto-regressive action generation. In this paper, we introduce LinkVLA, a novel architecture that directly addresses these challenges to enhance both alignment and efficiency. First, we establish a structural link by unifying language and action tokens into a shared discrete codebook, processed within a single multi-modal model. This structurally enforces cross-modal consistency from the ground up. Second, to create a deep semantic link, we introduce an auxiliary action understanding objective that trains the model to generate descriptive captions from trajectories, fostering a bidirectional language-action mapping. Finally, we replace the slow, step-by-step generation with a two-step coarse-to-fine generation method (C2F) that efficiently decodes the action sequence, saving 86% inference time. Experiments on closed-loop driving benchmarks show consistent gains in instruction following accuracy and driving performance, alongside reduced inference latency.

Dexterous manipulation remains challenging due to the cost of collecting real-robot teleoperation data, the heterogeneity of hand embodiments, and the high dimensionality of control. We present UniDex, a robot foundation suite that couples a large-scale robot-centric dataset with a unified vision-language-action (VLA) policy and a practical human-data capture setup for universal dexterous hand control. First, we construct UniDex-Dataset, a robot-centric dataset over 50K trajectories across eight dexterous hands (6-24 DoFs), derived from egocentric human video datasets. To transform human data into robot-executable trajectories, we employ a human-in-the-loop retargeting procedure to align fingertip trajectories while preserving plausible hand-object contacts, and we operate on explicit 3D pointclouds with human hands masked to narrow kinematic and visual gaps. Second, we introduce the Function-Actuator-Aligned Space (FAAS), a unified action space that maps functionally similar actuators to shared coordinates, enabling cross-hand transfer. Leveraging FAAS as the action parameterization, we train UniDex-VLA, a 3D VLA policy pretrained on UniDex-Dataset and finetuned with task demonstrations. In addition, we build UniDex-Cap, a simple portable capture setup that records synchronized RGB-D streams and human hand poses and converts them into robot-executable trajectories to enable human-robot data co-training that reduces reliance on costly robot demonstrations. On challenging tool-use tasks across two different hands, UniDex-VLA achieves 81% average task progress and outperforms prior VLA baselines by a large margin, while exhibiting strong spatial, object, and zero-shot cross-hand generalization. Together, UniDex-Dataset, UniDex-VLA, and UniDex-Cap provide a scalable foundation suite for universal dexterous manipulation.

Active 3D reconstruction enables an agent to autonomously select viewpoints to build accurate and complete scene geometry efficiently, rather than passively reconstructing scenes from pre-collected images. Existing active reconstruction methods often rely on geometric heuristics, which may result in redundant observations without improving reconstruction quality. To address this, we propose AREA3D, an active reconstruction agent for 3D reconstruction by leveraging feed-forward 3D models and vision-language guidance. The framework decouples view uncertainty modeling from feed-forward reconstruction, enabling precise uncertainty estimation without online optimization. Moreover, the integrated Vision-Language Model provides high-level semantic guidance that guides exploration beyond purely geometric cues. Extensive experiments on both scene-level and object-level benchmarks (Replica and OmniObject3D) demonstrate that AREA3D achieves state-of-the-art reconstruction accuracy, especially in sparse views.

Vision-Language-Action models (VLAs) achieve remarkable performance in sequential decision-making but remain fragile to subtle environmental shifts, such as small changes in object pose. We attribute this brittleness to trajectory overfitting, where VLAs over-attend to the spurious correlation between actions and entities, then reproduce memorized action patterns. We propose Perturbation learning with Delayed Feedback (PDF), a verifier-free test-time adaptation framework that improves decision performance without fine-tuning the base model. PDF mitigates the spurious correlation through uncertainty-based data augmentation and action voting, while an adaptive scheduler allocates augmentation budgets to balance performance and efficiency. To further improve stability, PDF learns a lightweight perturbation module that retrospectively adjusts action logits guided by delayed feedback, correcting overconfidence issue. Experiments on LIBERO (+7.4% success rate) and Atari (+10.3 human normalized score) demonstrate consistent gains of PDF in task success over vanilla VLA and VLA with test-time adaptation, establishing a practical path toward reliable test-time adaptation in multimodal decision-making agents.

Multimodal Continual Learning (MMCL) aims to enable models to continuously accumulate knowledge across multiple tasks and modalities without forgetting prior information. MMCL presents more challenges than single-modal continual learning, as it requires effective cooperation and complementarity between modalities. Existing methods often treat modality alignment as a static process, assuming once alignment is established, it remains fixed. However, we argue that modality alignment is inherently dynamic, evolving with task learning and feature propagation across layers. To address this, we introduce Dynamic Alignment Graph Regularization (DAGR), a novel approach that explicitly models the evolving alignment across layers. By incorporating multi-level graph regularization, our method stabilizes the alignment process and mitigates catastrophic forgetting. Extensive experiments on benchmarks, such as MTIL, show that DAGR outperforms static alignment-based methods and other continual learning techniques, achieving superior stability.

Visual Geometry Grounded Transformer (VGGT) has advanced 3D vision, yet its global attention layers suffer from quadratic computational costs that hinder scalability. Several sparsification-based acceleration techniques have been proposed to alleviate this issue, but they often suffer from substantial accuracy degradation. We hypothesize that the accuracy degradation stems from the heterogeneity in head-wise sparsification sensitivity, as the existing methods apply a uniform sparsity pattern across all heads. Motivated by this hypothesis, we present a two-stage sparsification pipeline that effectively quantifies and exploits headwise sparsification sensitivity. In the first stage, we measure head-wise sparsification sensitivity using a novel metric, the Head Sensitivity Score (HeSS), which approximates the Hessian with respect to two distinct error terms on a small calibration set. In the inference stage, we perform HeSS-Guided Sparsification, leveraging the pre-computed HeSS to reallocate the total attention budget-assigning denser attention to sensitive heads and sparser attention to more robust ones. We demonstrate that HeSS effectively captures head-wise sparsification sensitivity and empirically confirm that attention heads in the global attention layers exhibit heterogeneous sensitivity characteristics. Extensive experiments further show that our method effectively mitigates performance degradation under high sparsity, demonstrating strong robustness across varying sparsification levels. Code is available at https://github.com/libary753/HeSS.