CVPR 2026 论文列表

Large vision-language models (LVLMs) have demonstrated remarkable capabilities by integrating pre-trained vision encoders with large language models (LLMs). Similar to single-modal LLMs, chain-of-thought (CoT) prompting has been adapted for LVLMs to enhance multi-modal reasoning by generating intermediate rationales based on visual and textual inputs. While CoT is assumed to improve grounding and accuracy in LVLMs, our experiments reveal a key challenge: existing LVLMs often ignore the contents of generated rationales in CoT reasoning. To address this, we re-formulate multi-modal CoT reasoning as a KL-constrained reward maximization focused on rationale-conditional log-likelihood. As the optimal solution, we propose rationale-enhanced decoding (RED), a novel plug-and-play inference-time decoding strategy. RED harmonizes visual and rationale information by multiplying distinct image-conditional and rationale-conditional next token distributions. Extensive experiments show that RED consistently and significantly improves reasoning over standard CoT and other decoding methods across multiple benchmarks and LVLMs. Our work offers a practical and effective approach to improve both the faithfulness and accuracy of CoT reasoning in LVLMs, paving the way for more reliable rationale-grounded multi-modal systems.

End-to-end driving frameworks, which directly map raw sensor data to vehicle control commands, have shown remarkable potential. However, their performance often deteriorates in sparse-critical scenarios, where rare but safety-sensitive events occur. To address this problem, we propose Explorer-Expert Reinforcement Learning (EE-RL), a novel end-to-end framework that integrates an RL-based explorer, a fine-tuned vision-language model (VLM)-based expert, and a dual replay buffer. EE-RL adopts a collaborative learning strategy in which an explorer and two experts jointly generate experiences from regular driving scenarios to guide policy learning. As training progresses, one of the VLMs focuses on reasoning about sparse-critical scenarios, thereby enhancing learning efficiency and policy optimization in both scenarios. Additionally, the StateHash algorithm is designed to separately measure RGB image similarity and vehicle kinematic similarity, thereby skipping unnecessary VLM reasoning and enabling denser, more effective expert experience generation. Experiments on the CARLA Leaderboard demonstrate that EE-RL significantly outperforms state-of-the-art (SOTA) baselines, achieving +19.82% and +20.98% improvements in driving and infraction scores on Town03, respectively. The EE-RL further achieves 0% accident probability for red-light running and an average driving score of 80.09 in Town05-06, demonstrating robustness, generalizability, and the capability to handle sparse-critical scenarios.

The Aerial Object Goal Navigation, a challenging frontier in Embodied AI, requires an Unmanned Aerial Vehicle (UAV) agent to autonomously explore, reason, and identify a specific target using only visual perception and language description. However, existing methods struggle with the memorization of complex spatial representations in aerial environments, reliable and interpretable action decision-making, and inefficient exploration and information gathering. To address these challenges, we introduce **APEX** (Aerial Parallel Explorer), a novel hierarchical agent designed for efficient exploration and target acquisition in complex aerial settings. APEX is built upon a modular, three-part architecture: 1) Dynamic Spatio-Semantic Mapping Memory, which leverages the zero-shot capability of a Vision-Language Model (VLM) to dynamically construct high-resolution 3D Attraction, Exploration, and Obstacle maps, serving as an interpretable memory mechanism. 2) Action Decision Module, trained with reinforcement learning, which translates this rich spatial understanding into a fine-grained and robust control policy. 3) Target Grounding Module, which employs an open-vocabulary detector to achieve definitive and generalizable target identification. All these components are integrated into a hierarchical, asynchronous, and parallel framework, effectively bypassing the VLM's inference latency and boosting the agent's proactivity in exploration. Extensive experiments show that APEX outperforms the previous state of the art by +4.2% SR and +2.8% SPL on challenging UAV-ON benchmarks, demonstrating its superior efficiency and the effectiveness of its hierarchical asynchronous design.

Humans can robustly localize visual evidence and provide grounded answers even in noisy environments by identifying critical cues and then relating them to the full context in a bottom-up manner. Inspired by this, we propose DeepScan, a training-free framework that combines Hierarchical Scanning, Refocusing, and Evidence-Enhanced Reasoning for visually grounded reasoning in Large Vision-Language Models (LVLMs). Unlike existing methods that pursue one-shot localization of complete evidence, Hierarchical Scanning performs local cue exploration and multi-scale evidence extraction to recover evidence in a bottom-up manner, effectively mitigating the impacts of distractive context. Refocusing then optimizes the localized evidence view through collaboration of LVLMs and visual experts. Finally, Evidence-Enhanced Reasoning aggregates multi-granular views via a hybrid evidence memory and yields accurate and interpretable answers. Experimental results demonstrate that DeepScan significantly boosts LVLMs in diverse visual tasks, especially in fine-grained visual understanding. It achieves 90.6% overall accuracy on V* when integrated with Qwen2.5-VL-7B. Moreover, DeepScan provides consistent improvements for LVLMs across various architectures and model scales without additional adaptation cost. The code will be open-source soon.

Reinforcement learning (RL) provides a principled framework for improving vision-language models (VLMs) on complex reasoning tasks. However, existing RL approaches often depend on human-annotated labels or task-specific heuristics to define verifiable rewards--both costly and limited in scalability. We introduce VisPlay, a self-evolving RL framework that enables VLMs to autonomously improve their reasoning capabilities from massive unlabeled image data. Starting from a single base VLM, VisPlay assigns the model into two interacting roles: an Image-Conditioned Questioner that formulates challenging yet answerable visual questions, and a Multimodal Reasoner that generates silver responses. These roles are jointly trained using Group Relative Policy Optimization (GRPO), which uses diversity and difficulty rewards to balance the difficulty of generated questions with the quality of silver answers. VisPlay scales efficiently across two model families. Trained on Qwen2.5-VL and MiMo-VL, VisPlay achieves consistent improvements in visual reasoning, compositional generalization, and hallucination reduction across eight benchmarks including MM-Vet and MMMU, and establishes a scalable path toward self-evolving multimodal intelligence.

Finetuning Large Vision-Language Models with reinforcement learning has emerged as a promising approach to enhance their capability in object-level grounding. However, existing methods, mainly based on GRPO, assign rewards at the response level. Such sparse reward leads to minimal learning signals when all candidate responses are failed in challenging scenarios.In this work, we propose a group-revision optimisation paradigm that enhances learning on hard cases. It begins with a sampled initial response and generates a set of revised candidates to explore improved grounding outcomes. Inspired by reward shaping, we introduce a consolidation process that quantifies each candidate's improvement over the initial attempt and converts it into informative shaping signals.These signals are used to both refine the reward and modulate the advantage, amplifying the influence of high-quality revisions.Our method achieves consistent gains across referring and reasoning segmentation, REC, and counting benchmarks compared with prior GRPO-based models. Code is available at https://github.com/yyliu01/GroupRevision.

Multimodal Large Language Models have achieved impressive performance on a variety of vision-language tasks, yet their fine-grained visual perception and precise spatial reasoning remain limited. In this work, we introduce DiG (Differential Grounding), a novel proxy task framework where MLLMs learn fine-grained perception by identifying and localizing all differences between similar image pairs without prior knowledge of their number. To support scalable training, we develop an automated 3D rendering-based data generation pipeline that produces high-quality paired images with fully controllable discrepancies. To address the sparsity of difference signals, we further employ curriculum learning that progressively increases complexity from single to multiple differences, enabling stable optimization. Extensive experiments demonstrate that DiG significantly improves model performance across a variety of visual perception benchmarks and that the learned fine-grained perception skills transfer effectively to standard downstream tasks, including RefCOCO, RefCOCO+, RefCOCOg, and general multimodal perception benchmarks. Our results highlight differential grounding as a scalable and robust approach for advancing fine-grained visual reasoning in MLLMs.

Interactive Text-to-Image Retrieval (I-TIR) aims to refine image retrieval results through natural language dialogues, which allows users to progressively supplement or correct their search intention across multiple rounds, enabling a more precise and user-aligned visual search experience.However, existing methods perform cross-modal retrieval within a fixed multimodal feature space, mapping all dialogue text and images onto the same static embedding manifold.Such static formulation easily causes semantic vagueness, making it difficult to capture subtle embedding shifts in the user's updated intention for fine-grained retrieval.To address this limitation, we propose Context-Aware Latent Space Transformation (CAST), a lightweight framework that dynamically transforms the common latent space of both textual and visual representations according to the specific evolving user's search intention, enabling fine-grained and adaptive semantic alignment.The core of CAST is the Context-Aware Space Regulator (CASR), a crucial space transformation module composed of two key components: (1) the Context-Aware Low-Rank Projector (CLP), which learns to predict the projection direction of embedding space based on the intent's context;and (2) the Context-Guided Modulator (CGM), which adaptively determines appropriate projection strength. CASR is highly lightweight, adding negligible parameters and computational overhead, and can be seamlessly integrated into diverse I-TIR frameworks.Extensive experiments demonstrate the effectiveness of our proposed framework, indicating that it can serve as a general, plug-and-play solution for efficient and scalable interactive text-to-image retrieval. Our source code is available at https://github.com/HuiGuanLab/CAST.

Open-Vocabulary Object Detection (OVOD) aims to enable detectors to generalize across categories by leveraging semantic information. Although existing methods are pretrained on large vision-language datasets, their inference is still limited to fixed category names, creating a gap between multimodal training and unimodal inference. Previous work has shown that improving textual representation can significantly enhance OVOD performance, indicating that the textual space is still underexplored. To this end, we propose OVOD-Agent, which transforms passive category matching into proactive visual reasoning and self-evolving detection. Inspired by the Chain-of-Thought (CoT) paradigm, OVOD-Agent extends the textual optimization process into an interpretable Visual Chain of Thought with explicit actions. OVOD's lightweight nature makes LLM-based management unsuitable; instead, we model visual context transitions as a Weakly Markovian Decision Process (w-MDP) over eight state spaces, which naturally represents the agent's state, memory, and interaction dynamics. A Bandit module generates exploration signals under limited supervision, helping the agent focus on uncertain regions and adapt its detection policy. We further integrate Markov transition matrices with Bandit trajectories for self-supervised Reward Model (RM) optimization, forming a closed loop from Bandit exploration to RM learning. Experiments on COCO and LVIS show that OVOD-Agent improves existing OVOD baselines and outperforms prior methods on novel categories, demonstrating strong generalization and scalability.

With the rapid rise of Large Vision Language Models (LVLMs) for image understanding, the objective of image compression is gradually shifting from human visual perception to machine-oriented semantic understanding. However, conventional learned compression techniques are optimized for pixel-level fidelity and typically operate at fixed or rigid bitrate points, misaligned with the semantic consistency and flexible bitrate control. This gap becomes critical in ultra-low-bitrate regimes, where latent representations often ignore semantic relevance and struggle to disentangle meaningful content from redundant visual details as the bitrate varies. To address these challenges, we develop a token-based flexible compression framework, Token Flow Guided Compression (TFGC), which unifies human- and machine-oriented objectives. TFGC supports variable bitrate control in ultra-low bitrate regimes and enables LVLMs to directly process compressed token without image reconstruction. Specifically, we explore token flow phenomenon in 1D token sequences and exploit it to design token flow propagation, which predicts missing tokens by propagating contextual information from unmasked tokens. Moreover, token semantic guidance aligns compressed representations with LVLM semantic space, while a progressive semantic alignment training strategy further bridges the gap between perceptual reconstruction and semantic reasoning. Experiments show that our framework achieves state-of-the-art LVLMs understanding at comparable bitrates while maintaining satisfactory perceptual quality.

Multiple Instance Learning (MIL) has emerged as a promising paradigm for Whole Slide Image (WSI) diagnosis, offering effective learning with limited annotations. However, existing MIL frameworks overlook diagnostic priorities and fail to differentiate the severity of misclassifications in multiclass, leaving clinically critical errors unaddressed. We propose a mistake-severity-aware training strategy that organizes diagnostic classes into a hierarchical structure, with each level optimized using a severity-weighted cross-entropy loss that penalizes high-severity misclassifications more strongly. Additionally, hierarchical consistency is enforced through probabilistic alignment, a semantic feature remix applied to the instance bag to robustly train class priority and accommodate clinical cases involving multiple symptoms. An asymmetric Mikel's Wheel-based metric is also introduced to quantify the severity of errors specific to medical fields. Experiments on challenging public and real-world in-house datasets demonstrate that our approach significantly mitigates critical errors in MIL diagnosis compared to existing methods. We present additional experimental results on natural domain data to demonstrate the generalizability of our proposed method beyond medical contexts.

Vision Transformers (ViTs) have recently achieved state-of-the-art performance in 2D human pose estimation due to their strong global modeling capability. However, existing ViT-based pose estimators are designed for static images and process each frame independently, thereby ignoring the temporal coherence that exists in video sequences. This limitation often results in unstable predictions, especially in challenging scenes involving motion blur, occlusion, or defocus. In this paper, we propose TAR-ViTPose, a novel Temporal Aggregate-and-Restore Vision Transformer tailored for video-based 2D human pose estimation. TAR-ViTPose enhances static ViT representations by aggregating temporal cues across frames in a plug-and-play manner, leading to more robust and accurate pose estimation. To effectively aggregate joint-specific features that are temporally aligned across frames, we introduce a joint-centric temporal aggregation (JTA) that assigns each joint a learnable query token to selectively attend to its corresponding regions from neighboring frames. Furthermore, we develop a global restoring attention (GRA) to restore the aggregated temporal features back into the token sequence of the current frame, enriching its pose representation while fully preserving global context for precise keypoint localization. Extensive experiments demonstrate that TAR-ViTPose substantially improves upon the single-frame baseline ViTPose, achieving a +2.3 mAP gain on the PoseTrack2017 benchmark. Moreover, our approach outperforms existing state-of-the-art video-based methods, while also achieving a noticeably higher runtime frame rate in real-world applications. Project page: https://github.com/zgspose/TARViTPose.

The dominant paradigm of monolithic scaling in Vision-Language Models (VLMs) is failing for understanding and reasoning in documents, yielding diminishing returns as it struggles with the inherent need of this domain for document-based procedural reasoning, cognitive complexity, and factual accuracy. To this end, we introduce MACT, a Multi-Agent Collaboration framework with agent-wise adaptive Test-time scaling that pioneers a paradigm shift to procedural scaling, adapting dynamically to the functional entities of visual documents understanding and reasoning. MACT decomposes the visual document processing flow into four specialized agents, i.e., planning, execution, judgment, and answer, to resolve cognitive overload and introduce a critical self-correction loop for factual grounding. This collaborative architecture is amplified by an agent-wise adaptive test-time scaling strategy that intelligently allocates computational resources based on the complexity and redundancy of each functionality. Evaluated on multiple visual document understanding benchmarks, MACT achieves superior performance with a smaller parameter scale, adapting effectively to various document scenarios without compromising its general or mathematical reasoning capabilities. The three variants consistently attain top-three average performance rankings, with average performance enhancements of 9.9-11.5% over the base models.

Video LLMs suffer from temporal inconsistency: small shifts in frame timing can flip attention and suppress relevant frames. We trace this instability to the common extension of Rotary Position Embeddings to video through multimodal RoPE. The induced inverse Fourier time kernel exhibits frame-scale ripples that multiply adjacent frames by different factors, which perturbs attention that should otherwise be governed by the raw query key inner product. We present Phase Aggregated Smoothing (PAS), a simple, training-free mechanism that applies small opposed phase offsets across heads and then aggregates their outputs. PAS preserves the per-head spectrum magnitude, while the aggregation effectively smooths the temporal kernel and reduces phase sensitivity without changing the positional encoding structure. Our analysis shows that the RoPE rotated logit can be approximated as a content dot product scaled by a time kernel; smoothing this kernel yields Lipschitz stability of attention to small temporal shifts; multi phase averaging attenuates high frequency ripples while preserving per-head spectra under Nyquist-valid sampling. Experiments on multiple video understanding benchmarks under matched token budgets show consistent improvements with negligible computational overhead. PAS provides a plug and play upgrade for robust temporal encoding in Video LLMs.

We propose Context-aware Video-text Alignment (CVA), a novel framework to address a significant challenge in video temporal grounding: achieving temporally sensitive video-text alignment that remains robust to irrelevant background context. %CVA integrates complementary data-centric and architectural innovations to enhance contextual robustness. Our framework is built on three key components. First, we propose Query-aware Context Diversification (QCD), a new data augmentation strategy that ensures only semantically unrelated content is mixed in. It builds a video-text similarity-based pool of replacement clips to simulate diverse contexts while preventing the false negative caused by query-agnostic mixing. Second, we introduce the Context-invariant Boundary Discrimination (CBD) loss, a contrastive loss that enforces semantic consistency at challenging temporal boundaries, making their representations robust to contextual shifts and hard negatives. Third, we introduce the Context-enhanced Transformer Encoder (CTE), a hierarchical architecture that combines windowed self-attention and bidirectional cross-attention with learnable queries to capture multi-scale temporal context. Through the synergy of these data-centric and architectural enhancements, CVA achieves state-of-the-art performance on major VTG benchmarks, including QVHighlights and Charades-STA. Notably, our method achieves a significant improvement of approximately 5 points in Recall@1 (R1) scores over state-of-the-art methods, highlighting its effectiveness in mitigating false negatives.

With the emergence of 3D foundation models, there is growing interest in fine-tuning them for downstream tasks, where LoRA is the dominant fine-tuning paradigm. As 3D datasets exhibit distinct variations in texture, geometry, camera motion, and lighting, there are interesting fundamental questions: 1) Are there LoRA subspaces associated with each type of variation? 2) Are these subspaces disentangled (i.e., orthogonal to each other)? 3) How do we compute them effectively? This paper provides answers to all these questions. We introduce a robust approach that generates synthetic datasets with controlled variations, fine-tunes a LoRA adapter on each dataset, and extracts a LoRA subspace associated with each type of variation. We show that these subspaces are approximately disentangled. Integrating them leads to a reduced LoRA subspace that enables efficient LoRA fine-tuning with improved prediction accuracy for downstream tasks. In particular, we show that such a reduced LoRA subspace, despite being derived entirely from synthetic data, generalizes to real datasets. An ablation study validates the effectiveness of the choices in our approach.

Long-video understanding remains challenging due to extreme temporal redundancy, sparse yet decisive events, and the instability of long-horizon reasoning in visual-language models (VLMs). Existing agent-based methods invoke external micro-tools but remain static, repeatedly rebuilding long chains of fine-grained operations for each task without acquiring reusable multi-step perceptual skills.We propose META, the first training-free agent capable of self-evolving its tool-augmented reasoning. META operates through dual Solving and Evolving loops: it analyzes its own tool trajectories, abstracts recurring multi-step patterns into reusable macro-tools, and distills failed executions into structured failure priors that refine tool usage. Through symbolic consolidation and pruning, META progressively shortens reasoning paths and acquires more general perceptual and temporal abilities--without any parameter updates. META achieves state-of-the-art performance on long-video benchmarks, demonstrating a scalable, model-agnostic paradigm for long-video understanding that can continually evolve without additional training.

Computer vision and robotics applications ranging from augmented reality to robot autonomy in large-scale environments require spatio-temporal memory frameworks that capture both geometric structure for accurate language-grounding and semantic detail. Existing methods face a tradeoff, where producing rich open-vocabulary descriptions comes at the expense of real-time performance when these descriptions have to be grounded in 3D. To address these challenges, we propose Describe Anything, Anywhere, at Any Moment (DAAAM), a novel spatio-temporal memory framework for large-scale and real-time 4D scene understanding. DAAAM introduces a novel optimization-based frontend to infer detailed semantic descriptions from localized captioning models, such as the Describe Anything Model (DAM), using batch processing to speed up inference by an order of magnitude for online deployment. It leverages such semantic understanding to build a hierarchical 4D scene graph (SG), which acts as an effective globally spatially and temporally consistent memory representation. DAAAM constructs 4D SGs with detailed, geometrically grounded descriptions while maintaining real-time performance. We show that DAAAM's 4D SG interfaces well with a tool-calling agent for inference and reasoning. We thoroughly evaluate DAAAM in the complex task of spatio-temporal question answering (SQA) on the NaVQA benchmark and show its generalization capabilities for sequential task grounding on the SG3D benchmark. We further curate an extended OC-NaVQA benchmark for large-scale and long-time evaluations. DAAAM achieves state-of-the-art results in both tasks, improving OC-NaVQA question accuracy by 53.6%, reducing position errors by 21.9% and temporal errors by 21.6%, and improving SG3D task grounding accuracy by 27.8% over the most competitive baselines. We release our data and code open-source.

Active search and tracking of arbitrary targets by Unmanned Aerial Vehicles (UAVs) in cluttered environments remains a highly challenging problem. Existing methods either construct complex modular pipelines, leading to substantial computational costs, or adopt end-to-end controllers that often fail to generalize across different targets and scenes. Moreover, search and tracking are typically treated separately despite their strong interdependence. In this paper, we present UAST, a simple yet effective mapping-free framework that unifies active search and persistent tracking using only RGB-D observations. The proposed system couples a dual-branch perception module with a Rule-Based Point Search Policy that adaptively switches between tracking and search-based recovery. A lightweight control network generates dynamically feasible trajectories directly from fused perception and UAV states. Furthermore, we introduce a training strategy with an elaborated tracking-aware visibility loss and a tailored data construction. Extensive experiments in both simulated and real-world environments show that our approach achieves higher success rates, more stable long-term tracking, and faster target search compared with existing methods, while maintaining high efficiency. The code is available at https://github.com/qinliangql/UAST.