cs.CV（2026-10-02）

📊 共 34 篇论文 | 🔗 5 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (15) 支柱九：具身大模型 (Embodied Foundation Models) (8 🔗1) 支柱一：机器人控制 (Robot Control) (5 🔗2) 支柱三：空间感知与语义 (Perception & Semantics) (4 🔗1) 支柱四：生成式动作 (Generative Motion) (1) 支柱五：交互与反应 (Interaction & Reaction) (1 🔗1)

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

#题目一句话要点标签🔗⭐
1 Parasitic Co-Denoising: Unlocking 3D Human Motion Generation in a Frozen Video Diffusion Model 提出寄生共去噪以实现视频中的3D人类动作生成 flow matching motion generation human motion
2 FlowHMR: Physically Plausible Motion Capture from Video 提出FlowHMR以解决单目视频中人类运动捕捉问题 flow matching motion generation physically plausible
3 SymRegFlow: Symmetry-Regularized Flow Matching for Video World Models 提出SymRegFlow以解决多视角视频生成中的相机姿态变化问题 flow matching world model world models
4 Native Action-Prior Learning from Videos for World Action Models 提出NAVA-WAM以解决机器人动作预测的标注依赖问题 flow matching world action model world action models
5 EVEWorld: Physical Evolution Supervision for Embodied World Models 提出EVEWorld以解决机器人学习中的物理推理不足问题 world model world models JEPA
6 SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching 提出SpectralCache以解决扩散模型推理效率低下问题 world model world models
7 A Benchmark for Spatially Grounded Gesture Generation 提出空间基础手势生成基准以解决手势指向不明确问题 flow matching human motion
8 Beyond Entropy: Self-Diagnostic Multi-Role Token Optimization for Video Reasoning 提出DyCPO框架以解决视频推理中的模糊信号问题 reinforcement learning contrastive learning multimodal
9 OuroReward: Sequential Reward Scheduling for Reinforcement Learning in Text-to-3D Generation 提出OuroReward以解决文本到3D生成中的奖励调度问题 reinforcement learning
10 VDOT++: Unified Few-Step Video Generation via Unbalanced Optimal Transport Distillation 提出VDOT++以解决视频生成中的高成本与不稳定性问题 distillation
11 Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation 提出COFLOW以解决流模型推理效率低下问题 flow matching
12 Beyond Single Videos: Benchmarking and Active Evidence Seeking for E-Commerce Cross-Video Reasoning 提出AdsCVR基准与AdSeek框架以解决电商跨视频推理问题 reinforcement learning multimodal
13 ViTok: Improving Dense Semantics in AM-RADIO-Style Multi-Teacher Distillation with PHI-S and Masked Image Modelling 提出ViTok以优化AM-RADIO风格多教师蒸馏中的稠密语义 distillation
14 World Embedding Benchmark 提出世界嵌入基准以评估视频生成中的物理信息编码 world model world models
15 Does Physics Live in the Activations? Localizing Physical Quantities in Video Diffusion Models 通过激活定位物理量以提升视频生成模型的物理推理能力 world model world models

🔬 支柱九：具身大模型 (Embodied Foundation Models) (8 篇)

#题目一句话要点标签🔗⭐
16 Seeing, Saying, but Not Using: From Reportable Spatial Facts to Usable States in Multimodal Large Language Models 提出Operational State Supervision以解决多模态大语言模型的空间状态利用问题 large language model multimodal
17 CHASE-VLA: Post-Training Quantization Framework for Vision-Language-Action Models with Chunk-Aware Scale Estimation 提出CHASE-VLA以解决低比特后训练量化问题 vision-language-action VLA
18 Kinematics-Induced Multimodal 3D Human Pose Estimation with Subject-Level Privacy 提出一种多模态3D人体姿态估计框架以解决隐私问题 multimodal
19 Behavior Pack Optimization for Video MLLM Post-Training 提出行为包优化以提升视频多模态大语言模型的推理能力 large language model multimodal
20 Revealing Epistemic Uncertainty in MLLMs via Causal-Invariant Masking 提出因果不变掩蔽方法以揭示多模态大语言模型的不确定性问题 large language model multimodal
21 FiberGeoText: A Vision-Language Model for Population- Level Organization of Superficial White Matter 提出FiberGeoText模型以解决白质组织的个体差异问题 large language model multimodal
22 OmniAct3D: Leveraging Foundation Geometry and Evidence-Grounded Reasoning for Panoramic 3D Detection 提出OmniAct3D以解决全景3D检测中的几何适配问题 foundation model ✅
23 A Vision-Language Model (VLM)-based Pipeline for End-to-End Procedural Modeling of Field-Grown Maize from Point Clouds 提出基于视觉-语言模型的管道以自动化生成玉米3D模型 multimodal

🔬 支柱一：机器人控制 (Robot Control) (5 篇)

#题目一句话要点标签🔗⭐
24 Imagine the Future, Internalize the Gist: Efficient VLA Reasoning via Internalized Spatiotemporal Imagination 提出IG-VLA框架以解决机器人操作中的未来场景推理问题 manipulation spatiotemporal vision-language-action
25 EmbPASS: Towards Cross-Embodiment Open Panoramic Segmentation 提出EmbPASS以解决跨平台全景图像分割问题 quadruped scene understanding open-vocabulary ✅
26 ManifoldSplat: Language-Guided Semantic Shape Editing of 3D Gaussian Head Avatars 提出ManifoldSplat以解决3D头像编辑中的几何一致性问题 manipulation 3D gaussian splatting gaussian splatting ✅
27 Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis 提出SNAP以解决现有NVS方法的几何表示学习问题 manipulation representation learning depth estimation
28 CrowdOcc: Monocular Semantic Scene Completion for Quadruped Robots in Crowded Indoor Environments 提出CrowdOcc以解决拥挤室内环境中的单目语义场景补全问题 quadruped

🔬 支柱三：空间感知与语义 (Perception & Semantics) (4 篇)

#题目一句话要点标签🔗⭐
29 Depth Hypothesis Guided Iterative Refinement for Event-Image Monocular Depth Estimation 提出HypoDepth框架以解决事件相机单目深度估计问题 depth estimation monocular depth
30 UniDynamics: Event-RGB Fusion for Unified Future 4D Dynamic Scene Generation 提出UniDynamics以解决未来4D动态场景生成问题 optical flow physically plausible spatiotemporal
31 ForestQuery: Boundary-Aware and Spatially Anchored Query Learning for Unified Forest Point Cloud Segmentation 提出ForestQuery以解决森林点云分割中的边界不确定性问题 scene understanding ✅
32 One Photon, Many Worlds: Posteriors and Predictions with Single-Photon Cameras 提出条件生成框架以解决单光子相机的逆问题 scene understanding

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
33 Rethinking Fixed Temporal Grids: Frequency-Disentangled Motion Generation 提出FreqMo以解决人类运动生成中的时间网格问题 motion generation human motion human motion generation

🔬 支柱五：交互与反应 (Interaction & Reaction) (1 篇)

#题目一句话要点标签🔗⭐
34 COSMI: COmpositional Synthesis of Multi-object Interactions 提出COSMI以解决多对象交互生成模型的数据限制问题 human-object interaction ✅

⬅️ 返回 cs.CV 首页 · 🏠 返回主页