cs.CV（2026-10-01）

📊 共 70 篇论文 | 🔗 17 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (25 🔗9) 支柱三：空间感知与语义 (Perception & Semantics) (19 🔗3) 支柱一：机器人控制 (Robot Control) (11 🔗3) 支柱九：具身大模型 (Embodied Foundation Models) (10 🔗2) 支柱七：动作重定向 (Motion Retargeting) (2) 支柱六：视频提取与匹配 (Video Extraction) (1) 支柱四：生成式动作 (Generative Motion) (1) 支柱八：物理动画 (Physics-based Animation) (1)

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

#题目一句话要点标签🔗⭐
1 MEGA: Object-Level Mesh Extraction from 3D Gaussian Splatting via Spatial Visual Distillation 提出MEGA以解决3D高斯点云物体级网格提取问题 distillation 3D gaussian splatting 3DGS
2 FutureWorlds: Learning Robotic World Models from Alternative Futures 提出FutureWorlds框架以解决机器人世界模型学习中的未来预测问题 reinforcement learning world model world models ✅
3 Latent-Foresight: End-to-End Learning Predictable Representations for Latent World Models 提出Latent-Foresight以解决场景未来演化预测问题 world model world models representation learning ✅
4 Where-OPD: Spatially Guided On-Policy Self-Distillation of MLLMs with Synthetic Scenes 提出Where-OPD以解决多模态大语言模型的自我蒸馏问题 distillation privileged information large language model ✅
5 MMVistaReason: Toward Open-Data and Post-Training Recipes for Multimodal Reasoning 提出MMVistaReason以解决多模态推理中的后训练挑战 distillation multimodal
6 HiPhy: Hierarchical Alignment for Physically-Plausible Multi-Principle Video Generation 提出HiPhy以解决视频生成中的物理一致性问题 reinforcement learning physically plausible
7 World Observer: Joint Actor-Observer Generation for Persistent World Modeling 提出World Observer以解决动态环境观察问题 world model world models
8 4Director: Controlling Video World Models with Rigid 3D Geometry 提出4Director以解决视频制作中对象运动控制不精确的问题 world model world models
9 Oneira: From Open-Ended Generation to Open-World Interaction in Video World Models 提出Oneira以解决视频世界模型中的开放世界交互问题 world model world models ✅
10 AutoGUIWorld: Image Generators as Visual World Models for GUI Agent 提出AutoGUIWorld以生成高质量GUI交互轨迹 world model world models
11 OmniSeek: Native Tool Integration for Multi-turn Audio-Visual Reasoning 提出OmniSeek框架以提升多轮音视频推理能力 reinforcement learning large language model chain-of-thought
12 DecomVoxel: Harnessing 3D-Native Priors with Guided In-situ Denoising Optimization for Decompositional Scene Reconstruction 提出DecomVoxel以解决复杂场景重建中的遮挡问题 distillation scene reconstruction ✅
13 Smoother Flow Matching via Contrastive Trajectory Repulsion 提出CoFlow以解决流匹配中的轨迹交叉问题 flow matching contrastive learning distillation ✅
14 One Basis to Animate Them All: Gaussian Blendshape Distillation for Real-Time Avatars 提出GALA方法以解决实时动画中的神经推理瓶颈问题 distillation ✅
15 DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation 提出DMAD以解决快速视觉生成中的分布匹配问题 distillation ✅
16 Omni-Embed-Mini: Binding Modalities Without Forgetting via Dense Distillation 提出Omni-Embed-Mini以解决多模态嵌入模型参数冗余问题 distillation
17 Token-Level Video Reinforcement Learning 提出Token-Level视频强化学习以解决视频生成中的奖励局限性问题 reinforcement learning
18 Unsupervised Domain Adaptation for Enhanced Radiometer Image Precipitation Estimation using Conditional Flow Matching 提出无监督领域适应方法以提升卫星辐射计图像降水估计 flow matching
19 Architectural Sampling: Test-Time Scaling via Computational Diversity in Frozen Vision-Language Models 提出建筑采样方法以提升冻结视觉-语言模型的测试时间扩展能力 reinforcement learning multimodal
20 PAGER: Partial-to-global Alignment via Geometric and Relational Distillation 提出PAGER以解决部分视图与全局语义对齐问题 distillation
21 ROWBench: Do Video Models Render What the Program Specifies? 提出PROWBench以评估视频模型的程序规范遵循性 world model world models
22 iSEE: Object Permanence Through Self-Supervision 提出iSEE框架以解决物体持久性问题 world model world models ✅
23 Spatial Memory Intelligence: Endowing World Models with Understanding-Driven Long-Term Memory 提出空间记忆智能框架以解决长视频生成中的记忆管理问题 world model world models large language model
24 GRAFT: Growing Agglomerative Foundation Models via Continual Teacher Distillation 提出GRAFT框架以实现持续教师蒸馏的基础模型整合 distillation foundation model
25 Confidence-Gated Cloud-Edge Cascade Triage via Variational Risk Minimization for Medical Imaging 提出变分风险最小化框架以解决急诊胸部X光影像的多模态问题 distillation foundation model multimodal

🔬 支柱三：空间感知与语义 (Perception & Semantics) (19 篇)

#题目一句话要点标签🔗⭐
26 Dyna3: VLM-Guided Training-Free 4D Reconstruction via Depth Foundation Models 提出Dyna3以解决动态场景重建问题 depth estimation Depth Anything scene reconstruction
27 EvenSplat: Coupled 2D-3D Decomposition for Gaussian Splatting under Exposure and Illumination Variation 提出EvenSplat框架以解决光照变化下的多视图重建问题 3D gaussian splatting gaussian splatting splatting
28 Lang3DSeg: Annotation-Free Open-Vocabulary 3D Segmentation with Point Transformers 提出Lang3DSeg以解决无注释开放词汇3D分割问题 open-vocabulary open vocabulary
29 Open Vocabulary Word Recognition From Transcribed Bangla Texts 提出深度学习模型以解决孟加拉手写文本识别问题 open-vocabulary open vocabulary
30 Towards Automatic Video Annotation with ASH: Zero-Shot Open-Vocabulary Multi-Object Tracking and Segmentation 提出ASH以解决视频实例分割中的多类别开放词汇跟踪问题 open-vocabulary open vocabulary
31 LiteReality-Agent: An Agentic System for Interactable 3D Indoor Scene Reconstruction 提出LiteReality-Agent以解决室内场景重建问题 3D reconstruction scene reconstruction embodied AI ✅
32 VASC: Value-Aware Sparse Attention with Cross-Layer Memory for Efficient 3D Reconstruction 提出VASC以解决3D重建中的稀疏注意力问题 3D reconstruction scene reconstruction VGGT ✅
33 Semantic RGB--Depth Based Surgical Skill Assessment in Microscopic Stereo Videos 提出语义RGB-深度框架以解决显微外科技能评估问题 monocular depth stereo depth spatial relationship
34 MVDG: Efficient Multi-view 3D Disambiguation on Unconstrained Real-World Images 提出MVDG框架以解决多视图3D重建中的歧义问题 3D reconstruction VGGT foundation model
35 RelationVGGT: Visual Geometry Transformers for 3D Spatial Relation Segmentation 提出RelationVGGT以解决3D空间关系分割问题 3D reconstruction scene understanding foundation model
36 Resolving Mixed Single-Photon LiDAR Returns for Foreground-View and Hidden Scene Reconstruction 提出状态感知框架以解决混合单光子LiDAR返回问题 scene reconstruction
37 HierGF: Hierarchical Gaussian Fields via Geometry-perception Message Passing for Sparse-view 3D Reconstruction 提出HierGF以解决稀疏视图3D重建中的信息不足问题 3D reconstruction
38 CLoSeR: Closing the Loop for Long-Context Streaming Reconstruction 提出CLoSeR以解决长时序重建中的漂移问题 3D reconstruction foundation model ✅
39 Affine-Aligned Atlas for Canonical Gaussian Construction in Video Representation 提出仿射对齐的典型高斯表示以解决视频重建中的全局运动问题 gaussian splatting splatting
40 ARROW: Arbitrary Reconstruction and Tracking of 4D Observations in the Wild 提出ARROW以解决动态场景的3D重建与跟踪问题 3D reconstruction
41 OptimusMesh: Compact Autoregressive Mesh Generation from Point Clouds via Sparse Latent Pivots 提出OptimusMesh以解决点云生成紧凑三角网格的问题 implicit representation
42 Open Vocabulary Word Recognition From Transcribed Bangla Texts 提出深度学习模型以解决孟加拉手写文本识别问题 open-vocabulary open vocabulary
43 SCOPE-4D: Endoscopic 4D Geometry Foundation Models 提出SCOPE-4D以解决内窥镜几何理解问题 depth estimation foundation model
44 OptimusMesh: Compact Autoregressive Mesh Generation from Point Clouds via Sparse Latent Pivots 提出OptimusMesh以解决点云生成紧凑三角网格问题 implicit representation

🔬 支柱一：机器人控制 (Robot Control) (11 篇)

#题目一句话要点标签🔗⭐
45 ATI-VLA: Action-Centric Predictive Vision-Language-Action Models via Actionable Alignment Then Adaptive Injection 提出ATI-VLA以解决机器人操作中的动作预测不足问题 manipulation vision-language-action VLA
46 Ego2Act: Evaluating Goal-Directed Manipulation in Egocentric Video Generation 提出Ego2Act以解决自我中心视频生成中的多步骤物理推理问题 manipulation world model world models
47 CtrlWAM: Controllable World Action Models with Aligned Intent and Foresight 提出CtrlWAM以解决世界动作模型的控制与一致性问题 manipulation world action model world action models
48 Flow Matching Reinforcement for 3D Mesh Generation via Dynamic Homing Optimization 提出动态归向优化方法以提升3D网格生成质量 trajectory optimization reinforcement learning flow matching
49 Do MLLM Judges Judge the Edit? Auditing Bias in Image Editing Evaluation with Verified Quality Preservation 提出EditJudgeBias以解决MLLM评估图像编辑偏差问题 manipulation large language model multimodal
50 Task-Adaptive Grounded 3D-Programmers Using 2D VLMs 提出任务自适应的3D程序生成方法以解决3D理解问题 manipulation open-vocabulary open vocabulary
51 3DROID: A Renderable 3D Gaussian Dataset with Measured Per-Scene Reliability 提出3DROID以解决机器人操作中的3D场景可靠性问题 manipulation 3DGS ✅
52 GenCOPE: Syn2Real Generalized Category-Level Object Pose Estimation for Robotic Picking 提出GenCOPE以解决机器人抓取中的类别级物体姿态估计问题 manipulation scene understanding ✅
53 Exploring Weaknesses of Generative Image Watermarks against Latent Frequency Masking 提出潜在频率掩蔽攻击以解决生成图像水印的脆弱性问题 manipulation
54 SCION: Scene Composition with Instanced Neural Primitives 提出SCION以解决场景表示冗余与操作性不足问题 manipulation 3D gaussian splatting gaussian splatting
55 Physical AI Smart Spaces: A Large-Scale Benchmark for Multi-Camera 3D Perception in Smart Spaces 提出Physical AI Smart Spaces基准以解决室内多摄像头3D感知问题 sim2real VGGT ✅

🔬 支柱九：具身大模型 (Embodied Foundation Models) (10 篇)

#题目一句话要点标签🔗⭐
56 Continuous Conditioning of VLAs with Augmenting EMG and Visual Task Descriptors 提出电生理与视觉注释条件化VLA以解决任务描述不足问题 vision-language-action VLA multimodal
57 Harnessing Domain Specialists in Multimodal Mixture-of-Experts for Efficient Adaptation 提出ExpertLens以实现多模态专家的高效适应 multimodal
58 Form and Void: Entangled Composition through an Autonomous AI Agent 提出FaV-A以解决正负空间生成挑战 large language model multimodal
59 Fewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer Tokens 提出PyRUA-Lean框架以提升机器人任务成功率并减少令牌使用 vision-language-action VLA
60 MWOP: Modality-aware Width-wise Operation Pruning for Efficient MLLMs 提出MWOP以解决多模态大语言模型的高推理成本问题 large language model multimodal ✅
61 Before It Fades: Reinforcing Temporal Representations at Inference Time in VideoLLMs 提出时间激活注入方法以增强视频大语言模型的时间推理能力 large language model ✅
62 Concept Driven Domain Adaptation: Finding an Abstract Needle in a Haystack 提出概念驱动领域适应方法以解决视频检索中的抽象概念问题 multimodal
63 From Fragments to Global Maps: Learning Vectorized Map Aggregation with Large Language Models 提出MapMergeLLM以解决向量化地图聚合问题 large language model
64 Octrees as an Explicit 3D Language 提出OctLLM以解决现有3D语言模型的局限性 large language model multimodal
65 Form and Void: Entangled Composition through an Autonomous AI Agent 提出FaV-A以解决正负空间生成的挑战 large language model multimodal

🔬 支柱七：动作重定向 (Motion Retargeting) (2 篇)

#题目一句话要点标签🔗⭐
66 SuperMotion: Source-Preserving Denoising for Text-Driven Human Motion Editing 提出SuperMotion以解决文本驱动的人体动作编辑问题 human motion
67 Generative Cinematographer: Composing Camera and Object Motion in 3D 提出Generative Cinematographer以解决3D运动控制模糊问题 geometric consistency

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

#题目一句话要点标签🔗⭐
68 EgoFound3R: End-to-End Egocentric Hand Reconstruction in World Space with Point-Wise Interaction Attributes 提出EgoFound3R以解决手部重建与交互属性估计问题 egocentric hand reconstruction

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
69 Joint Branch-Space Transform Coding for Diffusion Activation Quantization with Classifier-Free Guidance 提出分支空间变换编码以提升扩散模型的激活量化精度 classifier-free guidance

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

#题目一句话要点标签🔗⭐
70 PhysDEM: Physics-Defined Energy-Matching Diffusion for Spatiotemporal Field Generation under Scarce Measurements 提出PhysDEM以解决稀疏测量下的时空场生成问题 spatiotemporal

⬅️ 返回 cs.CV 首页 · 🏠 返回主页