cs.CV（2026-06-12）

📊 共 72 篇论文 | 🔗 6 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (25) 支柱九：具身大模型 (Embodied Foundation Models) (23 🔗4) 支柱三：空间感知与语义 (Perception & Semantics) (11) 支柱一：机器人控制 (Robot Control) (8 🔗2) 支柱七：动作重定向 (Motion Retargeting) (3) 支柱八：物理动画 (Physics-based Animation) (1) 支柱六：视频提取与匹配 (Video Extraction) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (25 篇)

#题目一句话要点标签🔗⭐
1 From Seeing to Experiencing: Scaling Navigation Foundation Models with Reinforcement Learning 提出S2E框架以提升导航基础模型的交互能力 reinforcement learning 3D gaussian splatting gaussian splatting
2 Masked and Predictive Self-Supervised Foundation Models for 3D Brain MRI 提出自监督基础模型以提升3D脑MRI疾病检测效果 JEPA Joint-Embedding Predictive Architecture joint-embedding predictive architecture
3 Radar-Guided Polynomial Fitting for Metric Depth Estimation 提出POLAR以解决单目深度估计的度量深度转换问题 MAE depth estimation monocular depth
4 PaLMR: Towards Faithful Visual Reasoning via Multimodal Process Alignment 提出PaLMR以解决多模态推理过程不一致问题 reinforcement learning reward design large language model
5 GeoWorld-VLM: Geometry from World Models for Vision-Language Models 提出GeoWorld-VLM以解决视觉语言模型的空间关系识别问题 world model world models teacher-student
6 V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning 提出V-JEPA 2.1以解决视频自监督学习中的密集特征学习问题 world model world models JEPA
7 Periodic-MAE: Periodic Video Masked Autoencoder for rPPG Estimation 提出Periodic-MAE以解决rPPG估计中的视频信号表示问题 representation learning masked autoencoder MAE
8 AudioX-Turbo: A Unified Framework for Efficient Anything-to-Audio Generation 提出AudioX-Turbo以解决多模态音频生成效率问题 flow matching teacher-student distillation
9 Iterative Tool Usage Exploration for Multimodal Agents via Step-wise Preference Tuning 提出SPORT方法以解决多模态智能体工具使用探索问题 reinforcement learning multimodal
10 MoVerse: Real-Time Video World Modeling with Panoramic Gaussian Scaffold 提出MoVerse以解决实时视频世界建模问题 world model world models
11 VISA: VLM-Guided Instance Semantic Auditing for 3D Occupancy World Models 提出VISA以解决3D占用世界模型中的语义审计问题 world model world models
12 Analyzing and Improving Fine-grained Preference Optimization in Medical LVLMs 提出细粒度偏好优化方法以解决医疗LVLM中的对齐问题 DPO direct preference optimization visual grounding
13 ReFree: Towards Realistic Co-Speech Video Generation via Reward-Free RL and Multilevel Speech Guidance 提出ReFree-S2V以解决真实感对话视频生成问题 reinforcement learning flow matching character animation
14 Visual Place Recognition in Forests with Depth-Aware Distillation 提出深度感知蒸馏框架以解决森林环境中的视觉地点识别问题 distillation
15 BSViT: A Burst Spiking Vision Transformer for Expressive and Efficient Visual Representation Learning 提出BSViT以解决现有脉冲视觉变换器的效率与表现问题 representation learning
16 Objects Before Words: Object-First Inductive Biases for Grounding Language in Child-View Video 提出BabyMind以解决儿童视角视频中的语言基础问题 contrastive learning egocentric
17 OR-Action: Multi-Role Video Understanding with Fine-Grained Actions 提出OR-Action以解决手术室活动细粒度理解问题 distillation egocentric
18 MaskWAM: Unifying Mask Prompting and Prediction for World-Action Models 提出MaskWAM以解决WAM中空间瓶颈问题 world action model world action models
19 ReFoCUS: Reinforcement-guided Frame Optimization for Contextual Understanding 提出ReFoCUS以解决视频理解中的帧选择问题 reinforcement learning policy learning
20 RT-VLA: Real-Time Vision-Language-Action Models via Knowledge Distillation 提出RT-VLA以解决实时视觉-语言-动作模型的延迟问题 distillation vision-language-action VLA
21 S$^2$COPE: Self-Supervised Concept Discovery via Preference Learning 提出S$^2$COPE以解决自监督概念发现问题 preference learning representation learning
22 HumP-KD: A Hybrid Uncertainty-Aware Multi-Stage Progressive Knowledge Distillation Framework for Efficient Fire Classification 提出HumP-KD框架以提高火灾分类的效率与准确性 distillation
23 SED:Lightweight Saliency prediction for Event-based data via Distillation 提出SED以解决事件数据的轻量级显著性预测问题 distillation
24 ViT-Up: Faithful Feature Upsampling for Vision Transformers 提出ViT-Up以解决视觉变换器特征上采样问题 representation learning depth estimation
25 FLaRA: Predicting Future Latent Representations for Accident Anticipation 提出FLaRA以解决交通事故预测问题 Joint-Embedding Predictive Architecture joint-embedding predictive architecture

🔬 支柱九：具身大模型 (Embodied Foundation Models) (23 篇)

#题目一句话要点标签🔗⭐
26 VLADriveBench: Evaluating CoT-Action Relationship in VLA for Autonomous Driving 提出VLADriveBench以评估VLA模型中的CoT-Action关系 vision-language-action VLA chain-of-thought
27 GRIP: Feedback-Guided Prompt Retrieval for Large Multimodal Models 提出GRIP以解决多模态模型的示例检索问题 large language model multimodal
28 LaME: Learning to Think in Latent Space for Multimodal Embedding via Information Bottleneck 提出LaME以解决多模态嵌入中的推理效率问题 multimodal chain-of-thought
29 Ex-Omni: Enabling 3D Facial Animation Generation for Omni-modal Large Language Models 提出Ex-Omni以解决多模态语言模型生成3D面部动画问题 large language model multimodal
30 Navigating Gigapixel Pathology Images with Large Multimodal Models 提出GIANT以解决病理全切片图像导航问题 multimodal
31 MultiToP: Learning to Patch Visual Tokens to Mitigate Hallucinations in Video Large Multimodal Models 提出MultiToP以解决视频多模态模型的幻觉问题 multimodal
32 Cross-Modal Masked Compositional Concept Modeling for Enhancing Visio-Linguistic Compositionality 提出MACCO以解决视觉语言模型的组合理解问题 large language model multimodal
33 GAE: Unleashing Physical Potential of VLM with Generalizable Action Expert 提出通用动作专家GAE以解决视觉语言模型的机器人动作生成问题 vision-language-action
34 Goal2Pixel: Grounding Goals to Pixels for Vision-Language Navigation 提出Goal2Pixel以解决视觉语言导航中的低效问题 VLN
35 Perceive, Interact, Reason: Building Tool-Augmented Visual Agents for Spatial Reasoning 提出PERIA以解决空间推理任务中的工具增强问题 multimodal
36 LatentLens: Revealing Highly Interpretable Visual Tokens in LLMs 提出LatentLens以提升视觉标记的可解释性 large language model
37 OccAny: Generalized Unconstrained Urban 3D Occupancy 提出OccAny以解决城市3D占用预测的泛化问题 foundation model
38 MinhwaNet: Faithful but Insufficient Object Grounding in Korean Folk Painting 提出MinhwaNet以解决韩国民画中的对象定位不足问题 multimodal
39 Encoder Winners Do Not Reliably Transfer Across VLA Backbone Scale: A Frozen-Backbone Grafting Diagnostic 提出冻结骨干移植诊断以解决VLA编码器迁移问题 vision-language-action VLA
40 RepFusion: Leveraging Multimodal Priors for Denoising in Representation Space 提出RepFusion以解决视觉表示去噪问题 large language model multimodal
41 ClinHallu: A Benchmark for Diagnosing Stage-Wise Hallucinations in Medical MLLM Reasoning 提出ClinHallu以解决医疗多模态大语言模型推理中的幻觉诊断问题 large language model multimodal ✅
42 IndustryBench-MIPU: Benchmarking Multi-Image Attribute Value Extraction for Industrial Products 提出IndustryBench-MIPU以解决工业产品多图像属性值提取问题 large language model multimodal
43 GarmentSketch: Large-scale Sketch-to-Fashion Benchmark 提出GarmentSketch数据集以解决时尚草图生成问题 large language model multimodal ✅
44 Gaze Heads: How VLMs Look at What They Describe 提出凝视头机制以提升视觉语言模型的描述能力 multimodal
45 One Layer's Trash is Another Layer's Treasure: Adaptive Layer-wise Visual Token Selection in LVLMs 提出自适应层级视觉标记选择以解决LVLM计算负担问题 multimodal
46 MUSE: Agentic 3D Scene Authoring via Memory-Grounded Incremental Requirement Satisfaction 提出MUSE以解决3D场景生成的可编辑性问题 embodied AI
47 FEMOT: Multi-Object Tracking using Frame and Event Cameras 提出FEMOT以解决RGB-事件多目标跟踪数据集不足问题 multimodal ✅
48 FoleyGenEx: Unified Video-to-Audio Generation with Multi-Modal Control, Temporal Alignment, and Semantic Precision 提出FoleyGenEx以解决视频到音频生成中的多模态控制与时间对齐问题 large language model ✅

🔬 支柱三：空间感知与语义 (Perception & Semantics) (11 篇)

#题目一句话要点标签🔗⭐
49 Wild3R: Feed-Forward 3D Gaussian Splatting from Unconstrained Sparse Photo Collection 提出Wild3R以解决稀疏照片集合中的3D高斯点云问题 3D gaussian splatting 3DGS gaussian splatting
50 SemanticXR: Low Power and Real-time Queryable Semantic Mapping with an Object-Level Device-Cloud Architecture 提出SemanticXR以解决移动XR设备的语义映射问题 semantic mapping semantic map open-vocabulary
51 DepthMaster: Unified Monocular Depth Estimation for Perspective and Panoramic Images 提出DepthMaster以解决单目深度估计在透视与全景图像中的通用性问题 depth estimation monocular depth metric depth
52 Triangle Splatting SLAM 提出Triangle Splatting SLAM以实现高效的3D地图构建与跟踪 3D gaussian splatting gaussian splatting splatting
53 Flex4DHuman: Flexible Multi-view Video Diffusion for 4D Human Reconstruction 提出Flex4DHuman以解决动态人类重建问题 gaussian splatting splatting 4DHumans
54 Visual enhancement and 3D representation for underwater scenes: a review 综述水下视觉增强与三维重建技术以应对复杂成像挑战 3D gaussian splatting 3D reconstruction gaussian splatting
55 Modality-Aware Feature Matching in Visual and Vision-Language Applications: A Comprehensive Survey 综述多模态特征匹配以解决计算机视觉中的挑战 3D reconstruction feature matching
56 MooMIns -- Monocular 3D Reconstruction and Object Pose Estimation from Multiple Instances 提出MooMIns以解决单目图像中的3D重建与物体姿态估计问题 depth estimation monocular depth 3D reconstruction
57 NEST3D: A High-Resolution Multimodal Dataset of Sociable Weaver Tree Nests 提出NEST3D数据集以解决生态结构3D重建问题 3D reconstruction multimodal
58 Pano3D: Unified 3D Reconstruction and Panoptic Segmentation 提出Pano3D以解决3D重建与全景分割的统一问题 3D reconstruction
59 What Drives Test-Time Adaptation for CLIP? A Controlled Empirical Study from an Update Perspective 提出TTABC基准以系统研究CLIP的测试时适应性问题 open-vocabulary open vocabulary

🔬 支柱一：机器人控制 (Robot Control) (8 篇)

#题目一句话要点标签🔗⭐
60 RepWAM: World Action Modeling with Representation Visual-Action Tokenizers 提出RepWAM以解决现有世界动作模型的局限性 manipulation world action model world action models
61 InterleaveThinker: Reinforcing Agentic Interleaved Generation 提出InterleaveThinker以解决图像生成中的交错生成问题 manipulation multimodal instruction following
62 CPAM: Context-Preserving Adaptive Manipulation for Zero-Shot Real Image Editing 提出CPAM以解决复杂非刚性真实图像编辑问题 manipulation
63 Measurement-Calibrated Multi-Camera Fusion for Vision-Based Indoor Localization 提出测量校准的多摄像头融合方法以解决室内定位问题 motion tracking
64 WAM4D: Fast 4D World Action Model via Spatial Register Tokens 提出WAM4D以解决3D空间约束不足的问题 manipulation world action model world action models
65 Scratched Lenses, Shifted Depth: Passive Camera-Side Optical Attacks 提出SLASH以解决相机镜头划痕引发的深度偏差问题 manipulation depth estimation monocular depth
66 Conditioning Matters: Stabilizing Inversion and Attention in Diffusion Image Editing 提出SimEdit以解决图像编辑中的稳定性与注意力问题 manipulation ✅
67 Toward 360-Degree Indoor Panorama Editing via Tuning-Free Diffusion Model with Refocusing Cross-Attention 提出FocusDiff以解决360度室内全景编辑中的精确性问题 manipulation ✅

🔬 支柱七：动作重定向 (Motion Retargeting) (3 篇)

#题目一句话要点标签🔗⭐
68 Dual-Domain Equivariant Generative Adversarial Network for Multimodal CT-PET Synthesis 提出双域等变生成对抗网络以解决多模态CT-PET合成问题 geometric consistency multimodal
69 OmniDirector: General Multi-Shot Camera Cloning without Cross-Paired Data 提出OmniDirector以解决多镜头视频生成中的相机运动克隆问题 motion representation multimodal
70 VideoWeave: Unlocking Geometric Consistency in Video Generation via Joint Geometry-Video Modeling 提出VideoWeave以解决视频生成中的几何一致性问题 geometric consistency

🔬 支柱八：物理动画 (Physics-based Animation) (1 篇)

#题目一句话要点标签🔗⭐
71 HYDRA-X: Native Unified Multimodal Models with Holistic Visual Tokenizers 提出HYDRA-X以解决多模态模型中的图像与视频统一表示问题 spatiotemporal multimodal

🔬 支柱六：视频提取与匹配 (Video Extraction) (1 篇)

#题目一句话要点标签🔗⭐
72 Appearance-Invariant Detection of Suggestive Motion via Laban Movement Descriptors 提出基于Laban运动描述符的运动检测方法以解决内容审核问题 SMPL

⬅️ 返回 cs.CV 首页 · 🏠 返回主页