cs.CV（2026-06-22）

📊 共 44 篇论文 | 🔗 10 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (15 🔗2) 支柱九：具身大模型 (Embodied Foundation Models) (12 🔗4) 支柱三：空间感知与语义 (Perception & Semantics) (9) 支柱一：机器人控制 (Robot Control) (4 🔗2) 支柱七：动作重定向 (Motion Retargeting) (3 🔗2) 支柱八：物理动画 (Physics-based Animation) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (15 篇)

#题目一句话要点标签🔗⭐
1 MotionMAR: Multi-scale Auto-Regressive Human Motion Reconstruction from Sparse Observations 提出MotionMAR以解决稀疏观测下的人体运动重建问题 SAC VQ-VAE human motion
2 P-JEPA: Procedural Video Representation Learning via Joint Embedding Predictive Architecture 提出P-JEPA以解决长视频表示学习中的长程依赖问题 JEPA Joint-Embedding Predictive Architecture joint-embedding predictive architecture
3 MeGAS: Thermomechanical Dynamic Gaussian Splatting for Thermophysical Scene Editing 提出MeGAS框架以解决热机械动态场景编辑问题 world model world models 3D gaussian splatting
4 CFPO: Counterfactual Policy Optimization for Multimodal Reasoning 提出CFPO以解决多模态推理中的因果学习问题 reinforcement learning multimodal chain-of-thought ✅
5 Physics-Guided Spatiotemporal State Space Modeling for Lookahead Molten Pool Segmentation in Laser Wire-Feed Welding 提出物理引导的时空状态空间建模以解决激光焊接中的焊池分割问题 Mamba state space model distillation
6 Policy-as-Data: Learning Generalizable HOI Diffusion Models from Simulated Physics 提出基于物理模拟的HOI扩散模型以解决数据稀缺问题 reinforcement learning human-object interaction HOI
7 SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning 提出SingGuard以解决多模态安全评估问题 reinforcement learning multimodal ✅
8 MambaADv2: Evolving Duality-enhanced State Space Model for Unsupervised Anomaly Detection 提出MambaADv2以解决多类无监督异常检测问题 Mamba state space model
9 Compression and Retrieval: Implicit Memory Retrieval for Video World Models 提出压缩与检索机制以解决视频世界模型中的长期记忆问题 world model world models
10 BoxCtrl: 3D-Aware Visual Prompting for Geometric Image Editing 提出BoxCtrl以解决3D几何图像编辑的精确性问题 reinforcement learning large language model multimodal
11 READ More than What You See: Reinforcement Learning for Accurate and Coherent Audio Description Generations 提出READ框架以解决音频描述生成的准确性与连贯性问题 reinforcement learning multimodal
12 Scaling State-Space Models from Lines to Paragraphs: An Ablation of Mamba-based OCR 提出基于Mamba的OCR模型以提升长文本识别效率 Mamba SSM
13 Learning Adaptive Dynamical Features via Multi-$τ$ Liquid-Mamba for All-in-one Image Restoration 提出Multi-$τ$ Liquid-Mamba以解决图像恢复中的适应性问题 Mamba state space model
14 StreamPPG: Low-Latency rPPG Estimation via Consistent Privileged Learning 提出StreamPPG以解决低延迟rPPG估计问题 privileged information PULSE
15 Each Judge Its Own Yardstick: Discovering Per-VLM Taxonomies for Physical Video Evaluation 提出JudgeFit以解决VLM评估一致性问题 world model world models

🔬 支柱九：具身大模型 (Embodied Foundation Models) (12 篇)

#题目一句话要点标签🔗⭐
16 Intend, Reflect, Refine: An Adaptive Multimodal Reflection Framework for Autonomous Driving 提出IRR-Drive框架以解决复杂环境下自主驾驶的可靠性问题 vision-language-action VLA multimodal
17 Attention-Spectrum Regularization for Replay-Free Continual Multimodal LLMs 提出注意力谱正则化以解决持续多模态LLMs遗忘问题 large language model multimodal ✅
18 SPAR: Semantic-Pixel Self-Alignment and Adaptive Routing for Unified Multimodal Models 提出SPAR框架以解决多模态模型的视觉生成问题 large language model multimodal
19 Faithful Grounded Visual Reasoning via Learned Proxy-Tokens 提出Composer以解决视觉推理中的可解释性问题 large language model multimodal visual grounding
20 VideoLatent: Video-Language Learning via Latent Self-Forcing 提出VideoLatent以解决视频理解中的高成本问题 large language model multimodal chain-of-thought
21 Chains That See, Answers That Don't: A Multi-Aspect Evaluation Recipe for Forced Chain-of-Thought on Video-MME 提出多方面评估方法以验证视频问答中的强制思维链假设 chain-of-thought
22 LightSTAR: Efficient Visual Document Retrieval via Lightweight Selection with Vision-Adaptive Refinement 提出LightSTAR以解决视觉文档检索的效率问题 large language model ✅
23 UniverSat: Resolution- and Modality-Agnostic Transformers for Earth Observation 提出UniverSat以解决地球观测中的多模态输入问题 multimodal ✅
24 VideoAgent: All-in-One Framework for Video Understanding and Editing 提出VideoAgent以解决视频理解与编辑的多样性问题 multimodal ✅
25 SteerVTE: Seamless Video Text Editing with Style and Glyph Control 提出SteerVTE以解决视频文本编辑中的风格与字形控制问题 foundation model
26 Three-Step Hierarchical Transformer for Multi-Pedestrian Trajectory Prediction 提出三步层次Transformer以解决多行人轨迹预测问题 multimodal
27 Fursee: Hybrid YOLO-DINOv3 Framework for Fursuit Identity Retrieval and Clustering 提出Fursee框架以解决复杂毛皮服装身份检索与聚类问题 multimodal

🔬 支柱三：空间感知与语义 (Perception & Semantics) (9 篇)

#题目一句话要点标签🔗⭐
28 Temporally Aware Densification for Dynamic 3D Gaussian Splatting 提出可视性感知稠密化框架以解决动态场景重建问题 3D gaussian splatting 3DGS gaussian splatting
29 DrivingVoxels: Compositional Sparse Voxel Rasterization for Dynamic Driving Scene Reconstruction 提出DrivingVoxels以解决动态城市场景重建问题 3D gaussian splatting 3DGS gaussian splatting
30 Dense Reward for Multi-View 3D Reasoning with Global Maps and Local Views 提出DR-MV3D框架以解决多视角3D视觉问答中的稀疏监督问题 VGGT egocentric foundation model
31 Lift4D: Harmonizing Single-View 3D Estimation for 4D Reconstruction In-the-Wild 提出Lift4D以解决动态非刚性物体的4D重建问题 3D gaussian splatting 3D reconstruction gaussian splatting
32 Technical Report for the ICRA 2026 GOOSE 2D Fine-Grained Semantic Segmentation Challenge: Pretraining-Diverse Ensemble of Foundation Vision Encoders for Robust Outdoor Scene Understanding 提出多样化预训练集成方法以解决户外场景理解问题 scene understanding
33 Visual Geometry Transformer in the Wild: Distractor-Free 3D Reconstruction 提出视觉几何变换器以解决真实场景中的3D重建问题 3D reconstruction
34 Learning Stable Canonical Worlds for Novel View Synthesis and Beyond 提出CanonicalGS以解决多视角合成中的不稳定性问题 gaussian splatting splatting
35 G-MASt3R-SfM: Graph-based View Pruning and Multi-stage Optimization for Robust SfM 提出G-MASt3R-SfM以解决SfM中的匹配不可靠问题 3D reconstruction
36 OrthoMotion:Disentangling Camera and Subject Motion via Geometry Semantics Orthogonal Attention 提出OrthoMotion以解决相机与主体运动解耦问题 optical flow

🔬 支柱一：机器人控制 (Robot Control) (4 篇)

#题目一句话要点标签🔗⭐
37 Flow6D: Discrete-to-Continuous Flow Matching for Efficient and Accurate Category-Level 6D Pose Estimation 提出Flow6D以解决高维空间中6D姿态估计的效率与准确性问题 manipulation flow matching 6D pose estimation ✅
38 IMAGIN-4D: Image-Guided Controllable Interaction Generation 提出IMAGIN-4D以解决人机交互生成中的模糊性问题 manipulation human-object interaction HOI
39 Can Single-View Mesh Reconstruction Generalize to Robot Camera Rotation? 提出SAM3D+FoundationPose以解决机器人相机旋转下的单视图网格重建问题 manipulation depth estimation monocular depth
40 AIR: Adaptive Interleaved Reasoning with Code in MLLMs 提出自适应交错推理方法以解决多模态大语言模型的数值计算问题 manipulation reinforcement learning large language model ✅

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

#题目一句话要点标签🔗⭐
41 Polycepta: Object-Centric Appearance Estimation for Multi-Object Tracking 提出Polycepta以解决多目标跟踪中的外观估计问题 motion estimation motion prediction
42 RaysUp: Ultra-light Universal Feature Upsampling via Geometry-Aware Ray Representation 提出RaysUp以解决低分辨率特征上采样问题 geometric consistency foundation model ✅
43 Boosting Neural Video Codec via Scale-Driven Online Flow Refinement 提出无训练的尺度驱动在线流优化方法以提升神经视频编码器性能 motion estimation ✅

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

#题目一句话要点标签🔗⭐
44 PhysFlow: Frequency Decoupled with Dual-Field Rectified Flow for Remote Photoplethysmography 提出PhysFlow以解决复杂干扰下的rPPG估计问题 PULSE

⬅️ 返回 cs.CV 首页 · 🏠 返回主页