cs.CV（2026-10-08）

📊 共 87 篇论文 | 🔗 16 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (30 🔗6) 支柱二：RL算法与架构 (RL & Architecture) (28 🔗6) 支柱三：空间感知与语义 (Perception & Semantics) (13 🔗3) 支柱一：机器人控制 (Robot Control) (10) 支柱四：生成式动作 (Generative Motion) (3) 支柱六：视频提取与匹配 (Video Extraction) (2) 支柱八：物理动画 (Physics-based Animation) (1 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (30 篇)

#题目一句话要点标签🔗⭐
1 OneSearch-VL: Unified Multimodal Deep Research Agent for Image and Video 提出OneSearch-VL以解决多模态深度研究中的依赖性问题 multimodal visual grounding ✅
2 Look Back, Think Ahead: Visual Memory on Demand for Efficient Multimodal Reasoning 提出ViMoD框架以解决多模态推理中的视觉记忆问题 large language model multimodal
3 SAGE: Sink-Aware Guided Emphasis for Visual Grounding in Vision-Language Decoders 提出SAGE以解决视觉语言解码器中的注意力路径问题 large language model visual grounding
4 Improving Image-Based Nutrition Estimation Through Multimodal Food-Item Verification and Recovery 提出多模态框架以提升图像营养估计的准确性 large language model multimodal
5 DataVista: Diagnosing Multimodal LLMs on Data Video Understanding 提出DataVista以解决数据视频理解的评估问题 multimodal ✅
6 Multimodal Remote Sensing Image Registration: A Comprehensive Review, Challenges and Prospects 综述多模态遥感图像配准方法以应对高精度挑战 multimodal
7 Diffusion Meta-Prompting and Steering for Generalizable Foundation Model Adaptation 提出扩散元提示模型以解决基础模型适应性不足问题 foundation model ✅
8 Learning What to Trust in Multimodal Learning under Noisy Supervision 提出REFINE框架以解决多模态学习中的标签噪声问题 multimodal
9 OmniCapBench: A Deep-Structured Evaluation Framework for Fine-Grained Audio-Visual Captioning 提出OmniCapBench以解决音视频字幕评估的局限性问题 large language model multimodal
10 Is In-Domain Training Enough for Fine-Grained Industrial Anomaly Understanding? 提出SiGMA框架以提升工业异常理解的精细化能力 large language model multimodal
11 From What to Which: Decoding Modifier Grounding in Frozen MLLMs 提出OTTER以解决多模态大语言模型中的修饰符定位问题 large language model multimodal
12 DVD: Dynamic Vector Decoding for Efficient MLLM-based Perception 提出DVD方法以解决MLLM感知中的高开销和精度问题 large language model multimodal
13 ContourVLA: A Closed-Loop Perception-Action Contour Policy for Generalized Referring Expression Segmentation 提出ContourVLA以解决通用指称表达分割问题 vision-language-action multimodal
14 Mid-Training Language Models on Raw Video 提出利用原始视频进行语言模型中期训练以提升多模态性能 large language model multimodal
15 GATOR: Generative and Agentic 3D Object Reconstruction From Casual Images 提出GATOR框架以解决从随意图像重建3D物体的问题 multimodal
16 FastBench: Can Streaming VLMs Perceive High-Dynamic Real-World Streams? 提出FastBench以解决高动态视频流理解问题 large language model ✅
17 Pumpire: Unified Benchmark for Metric Distance Estimation 提出Pumpire基准以评估3D模型的距离估计能力 foundation model
18 Perception Test 2026: Challenge Summary and Extension to City-scale Audio-Visual Reasoning 提出城市规模音视频推理的新基准以解决空间智能问题 multimodal
19 Beyond Visual Enhancement: Adaptive Multi-Context Steering to Mitigate LVLM Hallucinations 提出AIMS框架以缓解LVLM中的幻觉问题 multimodal
20 From Suppression to Repair: Mitigating Object Hallucination in Large Vision-Language Models via Localized Distribution Alignment 提出ResOT以解决大规模视觉语言模型中的物体幻觉问题 multimodal
21 Seek-and-View Reasoning for Multi-View Spatial Understanding 提出Seek-and-View推理以解决多视角空间理解问题 foundation model ✅
22 Towards Unified Evaluation of Prompt Enhancers for Video Generation 提出PEBench以解决视频生成中提示增强器评估问题 multimodal
23 MSGAT: Multi-Head Spiking Graph Attention with Similarity-Space Fusion for Image-Text Retrieval 提出MSGAT以解决图像-文本检索中的跨模态对齐问题 multimodal
24 GroundSight at GroundLM 2026 Shared Tasks: GoldenViewVQA 提出CoVeR-VQA以解决多视角视觉问答中的证据定位问题 multimodal
25 FastJEV: Understanding Redundancy for Compact JEV Inference 提出FastJEV以解决JEV推理中的冗余问题 multimodal
26 SepGen: Multi-Stem Audio-Video Separation and Generation in a Single Model 提出SepGen以解决音视频分离与生成问题 language conditioned
27 MiniVer-V: Identifying Minimal Sufficient Evidence for Short Video Verification 提出MiniVer-V以解决短视频验证中的证据充分性问题 multimodal
28 Transforming Image Editors into Video Editors 提出基于锚点生成的视频编辑方法以解决视频编辑复杂性问题 instruction following ✅
29 Shift-Aware Identity-Guided Latent Refinement for Referring Audio--Visual Segmentation 提出Shift-Aware Identity-Guided Latent Refinement以解决Ref-AVS中的目标转移问题 multimodal
30 MIRAGE: Metric Inference and Reliability-Aware Gaussian Estimation from Feed-Forward Vision 提出MIRAGE以解决视觉重建中的度量尺度和轨迹漂移问题 multimodal

🔬 支柱二：RL算法与架构 (RL & Architecture) (28 篇)

#题目一句话要点标签🔗⭐
31 WOVEN: Weaving Visual World Modeling into Multimodal LLMs 提出WOVEN以解决多模态大语言模型的视觉推理问题 world model world models large language model
32 EchoDiST: Self-distillation-based joint learning for diffusion-conditioned echocardiographic myocardial motion estimation 提出EchoDiST以解决超声心动图心肌运动估计问题 distillation motion estimation
33 VGGTWorld-VLA: Intent-Conditioned 3D World Evolution for Autonomous Driving 提出VGGTWorld-VLA以解决自主驾驶中的3D世界演变问题 world model world models VGGT
34 AffordDrive3D: Affordance-Aware World-Action Modeling with Spatial Understanding 提出AffordDrive3D以解决自主驾驶中的空间理解问题 world model world models affordance
35 Controllable Exaggeration for Generative Motion Models via Training-Time Adaptation and Inference-Time Guidance 提出可控夸张方法以增强生成运动模型的表现力 flow matching text-to-motion motion generation
36 TAM: Task-Aware Memory Distillation for Efficient Spatiotemporal Prediction 提出TAM框架以提升时空预测的效率与准确性 teacher-student distillation spatiotemporal
37 Autoregressive Retriever: Improving Query Understanding from Item Feedback for Universal Multimodal Retrieval 提出自回归检索器以提升多模态检索中的查询理解 reinforcement learning multimodal
38 Rendering-Free Lookahead for Question-Guided Active Vision 提出无渲染前瞻策略以解决问答引导的主动视觉问题 distillation 3D gaussian splatting 3DGS
39 OuroWorld: Bringing Any 3D World Alive as Diverse, Endlessly Looping 3D Cinemagraphs 提出OuroWorld以解决静态3D场景动态化问题 world model world models 3D gaussian splatting
40 WorldGuide: Goal-Directed Video World Model for Procedural Task Execution 提出WorldGuide以解决长时程程序任务执行中的生成适应性问题 world model world models
41 WorldCast: Distributed Multiplayer World Models 提出WorldCast以解决多玩家世界模型生成的效率问题 world model world models
42 Right Screen, Wrong Transition: World Models as Verifiers for GUI Agents 提出LGWM以解决GUI代理安全性验证问题 world model world models
43 Learning to Retrieve: Internalizing Memory Retrieval for Video World Models 提出内化记忆检索机制以解决视频世界模型的场景一致性问题 world model world models ✅
44 IntactWorld: Joint World Modeling with Intact Features 提出IntactWorld以解决视频生成模型的结构信息损失问题 world model world models
45 CARE: Constrained Attention Refinement for Fine-Grained Visual Classification via Teacher-Student Distillation 提出CARE框架以解决细粒度视觉分类中的解释性与性能问题 teacher-student distillation
46 TAP3D: Thermal-Assisted 3D Human Point Clouds 提出TAP3D以解决传统人类点云重建的隐私与成本问题 MAE depth estimation human mesh recovery ✅
47 Rubric-CEPR: Self-Evolving Image Editing via Reward-Verified Self-Distillation 提出Rubric-CEPR以解决图像编辑自我进化问题 distillation ✅
48 DVLA-RL++: Dual-Level Vision-Language Alignment with Reinforcement Learning Gating for Few-Shot Learning 提出DVLA-RL++以解决少样本学习中的上下文污染问题 reinforcement learning ✅
49 Parametric Trajectory Distillation for Few-Step Video Generation 提出参数化轨迹蒸馏以解决视频生成计算成本问题 distillation ✅
50 Collaboratively Guided Adversarial Robust Distillation with Teacher-Favorable Examples 提出协同引导对抗鲁棒蒸馏以提升学生模型的鲁棒性 distillation
51 V-CoLA: Vision Token Compression with Linear Attention 提出V-CoLA以解决视觉语言模型的计算开销问题 linear attention
52 DiscoVL: Unveiling Disentangled C ross-Modal Representation Learning via Orthogonal Adversarial Regularization for V ision-Language Models 提出DiscoVL以解决视觉语言模型的跨模态表示学习问题 representation learning
53 No Distillation Needed: Single-Pass Real-Time Talking Heads via Acausal Noise Shaping 提出FaceGAN以解决实时音频驱动面部动画问题 distillation
54 ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills 提出ViSkill框架以增强视觉语言模型代理的技能学习 PPO reward shaping ✅
55 Zebra3D: Efficient Hybrid Foundation Models For Scalable 3D Reconstruction 提出Zebra3D以解决大规模3D重建效率问题 distillation 3D reconstruction VGGT
56 Scaling Multi-Teacher Distillation for Digital Pathology 提出多教师蒸馏方法以提升数字病理学模型的效率 distillation foundation model
57 Mind the Temporal Gap: Evaluating Video Models for Temporal Understanding in Driving 提出轻量化微调方法以提升驾驶视频理解中的时间动态捕捉能力 world model world models chain-of-thought
58 ForcingBooster: Adaptive Patch-wise Denoising for Fast Autoregressive Video Diffusion 提出ForcingBooster以解决视频生成中的计算冗余问题 world model world models

🔬 支柱三：空间感知与语义 (Perception & Semantics) (13 篇)

#题目一句话要点标签🔗⭐
59 MAMHOI: Factorizing Scene-Aware Human-Object Interaction through Affordances 提出MAMHOI以解决复杂场景下人机交互生成问题 scene understanding affordance motion synthesis ✅
60 PAM-ToD: Plug-and-Play Appearance Modeling for Cross-Time-of-Day 3D Gaussian Splatting 提出PAM-ToD以解决跨时间段3D高斯点云渲染问题 3D gaussian splatting 3DGS gaussian splatting
61 FlyMark: Training-Free Invisible Watermarking of 3D Gaussian Splatting via a Fruit Fly Connectome 提出FlyMark以解决3D高斯点云水印的训练依赖问题 3D gaussian splatting 3DGS gaussian splatting
62 Open-Vocabulary Audio-Visual Event Localization via Complex-Valued Fusion 提出复杂值融合方法以解决开放词汇音视频事件定位问题 open-vocabulary open vocabulary foundation model
63 TKCAM: Text and Keyframe to Camera Trajectory Generation 提出TKCAM以解决高质量相机运动生成问题 scene understanding motion synthesis motion matching ✅
64 OX-NeRF: 3D X-ray Tomography Reconstruction from Sparse Views Using Implicit Neural Representation 提出OX-NeRF以解决超稀疏X射线重建问题 3D reconstruction gaussian splatting splatting
65 PointVGGT: Zero-Shot Multiview RGB-D Point Cloud Registration with Visual Geometry Foundation Priors 提出PointVGGT以解决多视角RGB-D点云配准问题 metric depth 3D reconstruction VGGT
66 Slot3R: Set-Associative Spatial Memory for Streaming 3D Reconstruction 提出Slot3R以解决流式3D重建中的空间记忆问题 depth estimation 3D reconstruction
67 GenIA: Generative Reconstruction with Test-Time Input Alignment 提出GenIA框架以解决3D物体重建中的输入对齐问题 sam 3D SAM 3D foundation model ✅
68 LVS: Local View Synthesis from Relative Camera Pose by Reusing Previous Views 提出基于相对相机姿态的局部视图合成方法以提高场景探索效率 3D gaussian splatting gaussian splatting splatting
69 SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models 提出SpaceCast-Bench以评估视觉语言模型的预测空间推理能力 scene understanding
70 Domain Adaptation of SAM 2.1 for Building-Component Segmentation in Image-Based 3D Reconstruction 提出SAM 2.1的领域适应方法以解决建筑组件分割问题 3D reconstruction neural radiance field foundation model
71 Reference-Free Food Volume Estimation from Egocentric Stereo Video 提出一种无参考物体的食品体积估计方法以解决用户负担问题 VIO egocentric

🔬 支柱一：机器人控制 (Robot Control) (10 篇)

#题目一句话要点标签🔗⭐
72 EgoPhys: Estimating Peak Contact Force and Mechanical Work from Egocentric Manipulation Video 提出EgoPhys以解决从视频中估计接触力和机械功的问题 manipulation MAE contact-aware
73 VersaCamVLA: Camera-Configurable VLA Policies for Robotic Manipulation 提出VersaCamVLA以解决固定相机配置带来的机器人操作脆弱性问题 manipulation vision-language-action VLA
74 Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction 提出多智能体自我中心世界模型以解决细粒度交互问题 locomotion world model world models
75 LIVIN: Benchmarking Spatial and Embodied Intelligence in Digital Twins of Lived-In Homes 提出LIVIN基准以解决真实家庭环境中的空间与具身智能问题 manipulation loco-manipulation 3D reconstruction
76 RiCo: Neural Simulation of Rigid-Body Interactions via Local Contact Reasoning 提出RiCo以解决刚体相互作用模拟中的局部接触推理问题 sim-to-real world model world models
77 What 30,000 Hours of Ego-centric Video Does Not Teach 提出新监督机制以提升物体动态建模能力 humanoid world model world models
78 CRISP: Fixing Flying Pixels in Latent LiDAR Generation via Diffusion Decoding 提出CRISP以解决潜在LiDAR生成中的飞行像素问题 sim-to-real world model world models
79 Fast Pose Tracking of Rigid Objects with Compact Pose Graph Optimization 提出轻量级跟踪模块以解决长时间物体姿态跟踪问题 manipulation
80 MATE4D: Matrix-Guided Editable 4D Generation from a Single Image 提出MATE4D以解决单图生成动态4D内容的挑战 manipulation
81 LIVIN: Benchmarking Spatial and Embodied Intelligence in Digital Twins of Lived-In Homes 提出LIVIN基准以解决真实家庭环境中的空间与具身智能问题 manipulation loco-manipulation 3D reconstruction

🔬 支柱四：生成式动作 (Generative Motion) (3 篇)

#题目一句话要点标签🔗⭐
82 Stop My Dancing! Understanding, Detecting and Attributing Motion-Aware Deepfake Videos 提出MoDA框架以检测和归因运动感知深度伪造视频 motion synthesis zero-shot transfer
83 Phase-aware video generation for physics-grounded dynamics and interactions 提出PAVG以解决固体-气体动态生成问题 physically plausible spatiotemporal
84 Memory Forcing: Attendable Mid-Horizon History for Streaming Video Generation 提出Memory Forcing以解决视频生成中的中期遗忘问题 physically plausible

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

#题目一句话要点标签🔗⭐
85 LEGO: A Lifting-Free Approach for Exocentric-to-Egocentric Video Generation 提出无提升方法以解决外部视角到自我视角视频生成问题 egocentric
86 FloorSAV: Elucidating Spatial Audio-Visual Context with 2D Floormap for AV-LLMs 提出FloorSAV以解决AV-LLMs空间推理不足问题 egocentric large language model

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

#题目一句话要点标签🔗⭐
87 Revisiting Identity and Spectra Dispersion in Media-Bridged Time Series Forecasting: Linking Multivariate Signals and Narrative Flows 提出MIDAPN以解决多模态时间序列预测中的身份与谱散问题 spatiotemporal foundation model multimodal ✅

⬅️ 返回 cs.CV 首页 · 🏠 返回主页