cs.CV（2026-10-22）

📊 共 50 篇论文 | 🔗 12 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (18 🔗6) 支柱三：空间感知与语义 (Perception & Semantics) (15 🔗2) 支柱九：具身大模型 (Embodied Foundation Models) (6 🔗1) 支柱一：机器人控制 (Robot Control) (6 🔗2) 支柱六：视频提取与匹配 (Video Extraction) (2) 支柱八：物理动画 (Physics-based Animation) (2 🔗1) 支柱七：动作重定向 (Motion Retargeting) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (18 篇)

#题目一句话要点标签🔗⭐
1 Visual Grounding First, Multimodal In-context Learning Follows 提出ProCoRe以解决多模态ICL中的视觉基础问题 reinforcement learning multimodal visual grounding
2 I$^2$-KAN: Interpretable Intra- and Inter-Modal Multi-Pathway Representation Learning for Multimodal Sentiment Analysis 提出I$^2$-KAN以解决多模态情感分析中的可解释性问题 representation learning multimodal
3 Sensorimotor Alignment: A Platonic Proxy for Embodied Policy Success 提出传感运动对齐方法以评估机器人模型能力 world action model world action models vision-language-action
4 HiSTEVE 1.0: Harnessing Human-Interface VLA Agents for Open-World Games 提出HiSTEVE以解决开放世界游戏中的自主控制问题 reinforcement learning imitation learning VLA
5 Learning from Next-Frame Prediction: Towards Unified and Scalable Visual Representation Learning 提出VANE框架以解决视频预测与视觉表示学习的统一性问题 world model world models JEPA
6 GM-GReFEL: Geometry-Guided Spatiotemporal State-Space Learning for Robust Dynamic Facial Expression Recognition 提出GM-GReFEL以解决动态人脸表情识别中的噪声标签和计算复杂度问题 Mamba spatiotemporal ✅
7 Where Do Long Captions Fail? Position-Aware Diagnosis and Reinforcement Learning for Detailed Image Captioning 提出位置感知诊断与强化学习以解决长文本图像描述问题 reinforcement learning visual grounding
8 WovenAnchor Matcher: Specialized Intra- and Inter-Image Context Modeling for Feature Matching 提出WovenAnchor Matcher以解决特征匹配中的上下文建模问题 Mamba feature matching ✅
9 Reading, Writing, Sharing, and Interacting with the World through Video 通过视频基础模型实现世界的阅读、写作与交互 world model world models foundation model
10 RECAP: Looking Once Is Not Enough for Vision-Language Reasoning 提出RECAP以解决视觉语言推理中的时间基础失效问题 reinforcement learning chain-of-thought
11 One-shot federated unsupervised domain adaptation with Smoothed Knowledge Distillation and teacher refinement 提出SKD-ETR框架以解决FUDA中的标签噪声与通信效率问题 distillation ✅
12 CODI: Improving Self-Supervised Vision Transformers with Cross Distillation 提出CODI以改善自监督视觉变换器的密集预测能力 distillation ✅
13 Sensorimotor Alignment: A Platonic Proxy for Embodied Policy Success 提出传感运动对齐方法以评估机器人模型能力 world action model world action models vision-language-action
14 Learning from Next-Frame Prediction: Towards Unified and Scalable Visual World Modeling 提出VANE框架以解决视频世界建模中的未来预测问题 world model world models JEPA
15 PixGR: Pixel-Space Generative Decoding for Fast Feed-Forward 3D Reconstruction 提出PixGR以解决快速前馈3D重建中的场景不确定性问题 flow matching 3D reconstruction ✅
16 StackGen3D: Stable 3D Block Stack Generation from Multiview Silhouettes 提出StackGen3D以解决稳定3D块堆叠生成问题 flow matching penetration multimodal ✅
17 VisFlow: Interpretable Flow Matching for Visibility Reconstruction, Characterization, and Recognition 提出VisFlow以解决可见性重建与表征问题 flow matching representation learning
18 Interpreting Physics in Video Encoders using Multi-Object Interaction Videos 提出多物体交互视频解析物理属性的方法以提升视频编码器的可解释性 world model world models

🔬 支柱三：空间感知与语义 (Perception & Semantics) (15 篇)

#题目一句话要点标签🔗⭐
19 RecallGS: Geometry-Grounded Recall for Online Semantic Gaussian Splatting 提出RecallGS以解决在线语义高斯点云重建中的几何上下文缺失问题 3D gaussian splatting 3DGS gaussian splatting
20 MVLoMa: Multi-View Local Feature Matching with Any Geometry Foundation Model 提出MVLoMa以解决多视图局部特征匹配问题 NeRF VGGT feature matching
21 H-2DGS: Height-Field Gaussian Splatting for Sharp and Deformable Neurosurgical Surface Reconstruction 提出H-2DGS以解决神经外科表面重建中的模糊问题 3D gaussian splatting 3DGS gaussian splatting
22 Evaluation of Front-End Feature Matching for Lunar Visual SLAM 评估前端特征匹配以解决月球视觉SLAM问题 visual odometry visual SLAM feature matching
23 Query-Time Calibration of CLIP Features for Open-Vocabulary 3D Segmentation 提出基于场景校准的CLIP特征以提升开放词汇3D分割性能 open-vocabulary open vocabulary
24 Refining Open-Vocabulary Semantic Segmentation via Regional Semantics and Visual Prototypes 通过区域语义和视觉原型提升开放词汇语义分割性能 open-vocabulary open vocabulary
25 WildDepth-v2: A Synchronized RGB--LiDAR--Radar Dataset and Benchmark for Metric Depth, Motion, and Tracking of Wildlife 提出WildDepth-v2以解决野生动物运动测量的精度问题 monocular depth metric depth
26 A Unified Framework for Image-to-3D Part Generation via Variable Granularity 提出FlexPart以解决可控3D部件生成问题 sam 3D SAM 3D geometric consistency ✅
27 OmniRay: Camera-Agnostic Monocular 3D Reconstruction with Composite Ray Fields 提出OmniRay以解决单目3D重建中的相机几何问题 3D reconstruction
28 HiP3R: Hierarchical Pose Graphs for Long-Horizon 3D Reconstruction 提出HiP3R以解决长视频3D重建中的一致性问题 3D reconstruction ✅
29 Moving Regions, Not Moving Objects: Segmenting What Actually Moves under Camera Motion 提出MoRSe以解决动态场景下运动区域分割问题 optical flow foundation model
30 WorldForge V2: Eliciting Camera-Trajectory Alignment in Video Diffusion Models 提出WorldForge V2以解决视频扩散模型中的相机轨迹对齐问题 optical flow
31 AnyGameRenderer: Entity-Consistent Generative Rendering Across Games and an Entity-Level Benchmark 提出AnyGameRenderer以解决游戏视觉质量不足问题 semantic map
32 Stream4D: Streaming 4D Reconstruction with Training-free Motion Decoupling via Causal KV-Cache Attention 提出Stream4D以解决动态场景重建中的运动解耦问题 scene reconstruction
33 GSHeadRelight: Fast Relightability for 3D Gaussian Head Synthesis 提出GSHeadRelight以解决3D人头合成的快速重光照问题 NeRF

🔬 支柱九：具身大模型 (Embodied Foundation Models) (6 篇)

#题目一句话要点标签🔗⭐
34 A Visual Grounding Benchmark for Retinal OCT with Multimodal Large Language Models 提出RetinaGroundBench以解决视网膜OCT中的视觉定位问题 large language model multimodal visual grounding
35 Beyond Watching and Searching: Benchmarking Agents for Video-centered Deep Research Reports 提出视频中心深度研究报告生成方法以解决现有基准不足问题 multimodal visual grounding ✅
36 Making MLLM Latents Matter: Factor-Aware Predictive Learning for Cinematic Video Generation 提出因子感知预测学习以提升电影视频生成效果 large language model multimodal
37 Asking Less, Adapting Better: Feedback-Efficient Test-Time Adaptation for Vision-and-Language Navigation 提出FETA以解决视觉与语言导航中的反馈效率问题 VLN
38 Making MLLM Latents Matter: Factor-Aware Predictive Learning for Cinematic Video Generation 提出因子感知预测学习以提升电影视频生成效果 large language model multimodal
39 LOCI: Localized Continuous Control for Multi-Subject Image Editing 提出LOCI以解决多主体图像编辑中的局部控制问题 instruction following

🔬 支柱一：机器人控制 (Robot Control) (6 篇)

#题目一句话要点标签🔗⭐
40 DeXplicit: Making Physical State Explicit in Dexterous Manipulation World Models 提出DeXplicit以解决物理状态显式化问题 manipulation dexterous manipulation world model
41 PiHand: Latent Hand-Object Interaction Priors for Physics-Based Policy Learning 提出PiHand以解决物理基础下的手-物体交互合成问题 manipulation dexterous hand dexterous manipulation
42 DRILL: Training World Models to Improve Policies, Not Predict Pixels 提出DRILL框架以提升政策优化而非像素预测 manipulation reinforcement learning world model
43 SmolVLA-MoE: Sparse Mixture-of-Experts for Efficient Multi-Task Robotic Manipulation 提出SmolVLA-MoE以解决多任务机器人操作中的计算效率问题 manipulation vision-language-action VLA ✅
44 AgentDreamer: Agentic 3D Motion Reasoning for Text-Controllable Physics Simulation 提出AgentDreamer以解决文本控制的物理仿真问题 motion planning dreamer
45 F2G-Pose: Geometry-Aware Foundation Feature Lifting for Direct RGB-D Category-Level Object Pose Estimation 提出F2G-Pose以解决类别级RGB-D物体姿态估计问题 manipulation ✅

🔬 支柱六：视频提取与匹配 (Video Extraction) (2 篇)

#题目一句话要点标签🔗⭐
46 Retrieval-Augmented Online Textual Memory for Episodic Memory Video Question Answering 提出检索增强的在线文本记忆以解决视频问答挑战 egocentric Ego4D multimodal
47 AgentVQA: A Unified Benchmark for Agentic Visual Understanding 提出AgentVQA以解决代理视觉理解评估不足问题 egocentric spatial relationship

🔬 支柱八：物理动画 (Physics-based Animation) (2 篇)

#题目一句话要点标签🔗⭐
48 Direct Forcing: A Single-Stage Training Recipe for Few-Step Autoregressive Text-to-Video Generation 提出Direct Forcing以解决视频生成中的训练效率问题 spatiotemporal
49 SyncVid: World-Equivariant Diffusion for Synchronized Video Generation 提出SyncVid以解决视频生成中的视角一致性问题 spatiotemporal ✅

🔬 支柱七：动作重定向 (Motion Retargeting) (1 篇)

#题目一句话要点标签🔗⭐
50 AudioLift: Audio-Driven Single Image to Full-Head Gaussian Avatar via Multi-view Video Generation 提出AudioLift以解决传统3D虚拟头像生成的局限性 motion representation

⬅️ 返回 cs.CV 首页 · 🏠 返回主页