cs.CV（2026-10-06）

📊 共 35 篇论文 | 🔗 8 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (12 🔗4) 支柱三：空间感知与语义 (Perception & Semantics) (8 🔗2) 支柱九：具身大模型 (Embodied Foundation Models) (8) 支柱一：机器人控制 (Robot Control) (3 🔗1) 支柱四：生成式动作 (Generative Motion) (2 🔗1) 支柱七：动作重定向 (Motion Retargeting) (1) 支柱六：视频提取与匹配 (Video Extraction) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (12 篇)

#题目一句话要点标签🔗⭐
1 DensiTok: Making Feed-Forward 3D Gaussian Splatting See More Views Than It Is Given 提出DensiTok以解决稀疏视图重建质量下降问题 flow matching 3D gaussian splatting 3DGS
2 RSJEV: Discriminative Remote Sensing Scene Classification with Multimodal Large Language Models 提出RSJEV以解决遥感场景分类中的多模态决策问题 Mamba large language model multimodal ✅
3 Multimodal Knowledge Distillation for Gastric Adenocarcinoma Classification from Whole-Slide Images 提出多模态知识蒸馏框架以提高胃腺癌分类精度 distillation large language model multimodal ✅
4 World Models' Last Exam in Physics 提出物理一致性基准以评估视频世界模型的可靠性 world model world models embodied AI
5 MacJEPA: Missingness-Robust Audio-Visual Recognition from Untrimmed Egocentric Videos 提出MacJEPA以解决长视频中的缺失模态问题 JEPA egocentric multimodal
6 VisionWeave: Weaving Elastic Visual Representations as a Native Capability of MLLMs 提出VisionWeave以解决多模态大语言模型的视觉表示效率问题 distillation large language model foundation model
7 Dynamic Alignment and Calibration for Multimodal Learning 提出ACML框架以解决多模态学习中的对齐与校准问题 representation learning multimodal
8 4D-HOF: Hand-Object Flow Matching for Feed-Forward 4D Interaction Reconstruction 提出4D-HOF以解决4D手物体重建中的优化问题 flow matching hand-object reconstruction foundation model
9 CtrlCache: Accelerating Interactive Video World Models with Control-Aware Caching 提出CtrlCache以解决交互视频世界模型的计算效率问题 world model world models
10 UP-MOPD: Update Projection in Multi-Teacher On-Policy Distillation 提出UP-MOPD以解决多教师在线蒸馏中的梯度冲突问题 distillation instruction following
11 Unlocking Fine-Grained Perception in CLIP via Structurally-Aware Latent Masked Modeling 提出SALM框架以解决CLIP细粒度感知不足问题 distillation large language model multimodal ✅
12 Revisiting Numerical Forecasting Models for Language-Based Trajectory Prediction 提出MoRE框架以解决语言基础轨迹预测中的数值预测问题 reinforcement learning PPO ✅

🔬 支柱三：空间感知与语义 (Perception & Semantics) (8 篇)

#题目一句话要点标签🔗⭐
13 M3SunAgent: Monocular 3D Spatial Understanding Agent for Metric Depth Estimation and 3D Visual Grounding 提出M3SunAgent以解决单目3D空间理解问题 depth estimation metric depth large language model
14 Post-Training Semantic Lifting for 3D Gaussian Splatting: Separating Detector, Lifting and Representation Error 提出后训练语义提升方法以解决3D高斯点云分类问题 3D gaussian splatting gaussian splatting splatting
15 CueRator: Agentic Search for Symbolic Rules to Adapt Frozen Multimodal Encoders 提出CueRator以解决多模态编码器适应性与泛化性之间的矛盾问题 open-vocabulary open vocabulary large language model ✅
16 Efficient Gaussian Splatting Sequence Compression with Standard Video Codecs 提出高效的高斯点云序列压缩方法以解决视频编码问题 gaussian splatting splatting ✅
17 Revar3r: gauge-aware perturbation uncertainty for feed-forward 3d reconstruction 提出Revar3r以解决3D重建中的不确定性问题 3D reconstruction VGGT
18 Image-Space Refraction Correction for Underwater 3D Reconstruction: Warping Flat-Port Views into Pinhole Perspective 提出基于物理的折射校正方法以解决水下三维重建问题 3D reconstruction
19 Towards benchmarking Western Bluebird detection in the wild 提出新基准数据集以解决西部蓝鸟检测挑战 open-vocabulary open vocabulary
20 View Matters: Keyframe-Guided Text-Driven 3D Gaussian Editing 提出视图重要性感知框架以优化文本驱动的3D高斯编辑 3DGS

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

#题目一句话要点标签🔗⭐
21 DIPrune: Task-Aware Token Pruning with Dual Importance for Efficient Multimodal Language Models 提出DIPrune以解决多模态语言模型的任务感知剪枝问题 large language model multimodal
22 LiDAR Resolution Recovery via Foundation-Model-Guided Diffusion 通过基础模型引导扩散实现LiDAR分辨率恢复 foundation model
23 Beyond Perturbation Magnitude: Direction-Dependent Responses in Multimodal Geometric Representations 提出方向依赖的几何响应模型以解决多模态一致性问题 multimodal
24 Beyond Training from Scratch: Foundation Models for Data-Efficient and Generalizable Cardiac MRI Reconstruction 提出基于预训练模型的心脏MRI重建方法以提高数据效率 foundation model
25 GeoPID: Decomposing and Steering Visual Information in Vision-Language Models 提出GeoPID以增强视觉语言模型中的视觉信息利用 multimodal visual grounding
26 Forensic Reserve: Eliciting Latent Knowledge for Image Forgery Detection 提出Reserve-Guided Elicitation以解决图像伪造检测问题 foundation model
27 Image Bitstream Fine-grained Understanding for Privacy-Friendly AIoT 提出图像比特流细粒度理解以解决隐私友好AIoT问题 foundation model
28 Decide Before You Look: Learning Which Retrieved Memories Deserve Pixels 提出PixelTriage以优化多模态助手的记忆检索效率 multimodal

🔬 支柱一：机器人控制 (Robot Control) (3 篇)

#题目一句话要点标签🔗⭐
29 Digital Twin-Driven Real2Sim2Real: Simulator-Conditioned Generation via Paired Driving-Scene Reconstruction 提出数字双胞胎驱动的Real2Sim2Real以解决数据收集成本问题 sim2real real2sim scene reconstruction
30 EmbodiedSmith: Scaling Embodied Data through Recursive Self-Improvement Flywheel in Simulation 提出EmbodiedSmith以解决机器人数据生成的多样性与评估环境不足问题 humanoid dexterous hand foundation model
31 Can We Model the Artifacts Explicitly? Disentangle Artifacts via Pairwise Edit Relations for Image Manipulation Localization 提出显式建模伪影的方法以解决图像操控定位问题 manipulation ✅

🔬 支柱四：生成式动作 (Generative Motion) (2 篇)

#题目一句话要点标签🔗⭐
32 PhysLDM: Latent Diffusion for High-Fidelity Deformable Simulation 提出PhysLDM以解决高保真可变形体模拟问题 physically plausible spatiotemporal
33 Geometry-Constrained Bidirectional Point Cloud Registration for Thin, Sheet-Like Heritage Artifacts 提出几何约束双向点云配准方法以解决薄型文物重建问题 physically plausible ✅

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

#题目一句话要点标签🔗⭐
34 HuC-VideoMAE: Human-Centric Video Masked Autoencoding from synthetic data 提出人本视频掩码自编码以解决伦理数据问题 human motion

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

#题目一句话要点标签🔗⭐
35 RenderBench: Benchmarking Render-to-Real Video Transfer with Reconstructed Digital Twins 提出RenderBench以解决渲染到真实视频转移评估问题 egocentric

⬅️ 返回 cs.CV 首页 · 🏠 返回主页