cs.CV（2026-01-20）

📊 共 33 篇论文 | 🔗 11 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (17 🔗6) 支柱三：空间感知与语义 (Perception & Semantics) (7 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (6) 支柱八：物理动画 (Physics-based Animation) (1 🔗1) 支柱七：动作重定向 (Motion Retargeting) (1 🔗1) 支柱四：生成式动作 (Generative Motion) (1 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (17 篇)

#题目一句话要点标签🔗⭐
1 FantasyVLN: Unified Multimodal Chain-of-Thought Reasoning for Vision-Language Navigation 提出FantasyVLN，用于视觉语言导航中统一的多模态链式思考推理，提升效率与性能。 VLA VLN multimodal
2 Vision Also You Need: Navigating Out-of-Distribution Detection with Multimodal Large Language Model 提出MM-OOD以解决图像空间的OOD检测问题 large language model multimodal
3 LLM Augmented Intervenable Multimodal Adaptor for Post-operative Complication Prediction in Lung Cancer Surgery MIRACLE：融合临床与影像数据，可干预的LLM增强多模态适配器，用于肺癌术后并发症预测。 large language model multimodal
4 Weather-R1: Logically Consistent Reinforcement Fine-Tuning for Multimodal Reasoning in Meteorology 提出LoCo-RFT，解决气象领域多模态推理中逻辑不一致问题，并构建Weather-R1模型。 multimodal ✅
5 Scaling Test-time Inference for Visual Grounding 提出EGM：通过扩展测试时计算量提升视觉定位小模型的性能与效率。 visual grounding
6 The Side Effects of Being Smart: Safety Risks in MLLMs' Multi-Image Reasoning 提出MIR-SafetyBench，揭示多图推理能力增强的大语言模型安全风险。 large language model multimodal ✅
7 Fine-Grained Zero-Shot Composed Image Retrieval with Complementary Visual-Semantic Integration 提出CVSI模型，通过互补的视觉-语义融合实现细粒度零样本组合图像检索 large language model multimodal ✅
8 XD-MAP: Cross-Modal Domain Adaptation using Semantic Parametric Mapping 提出XD-MAP，利用语义参数化映射实现图像到LiDAR的跨模态领域自适应 foundation model
9 OmniTransfer: All-in-one Framework for Spatio-temporal Video Transfer OmniTransfer：用于时空视频迁移的统一框架，提升视频生成灵活性和保真度 multimodal
10 OCCAM: Class-Agnostic, Training-Free, Prior-Free and Multi-Class Object Counting 提出OCCAM，一种免训练、无先验、类别无关的多类别物体计数方法。 foundation model ✅
11 Insight: Interpretable Semantic Hierarchies in Vision-Language Encoders Insight：在视觉-语言编码器中构建可解释的语义层级结构 foundation model ✅
12 HiT: History-Injection Transformers for Onboard Continuous Flood Change Detection 提出历史注入Transformer（HiT），用于星载连续洪水变化检测，实现实时灾害评估。 foundation model ✅
13 Hierarchical Long Video Understanding with Audiovisual Entity Cohesion and Agentic Search 提出HAVEN框架，通过视听实体关联和Agent搜索实现层级长视频理解 multimodal
14 Reasoning or Pattern Matching? Probing Large Vision-Language Models with Visual Puzzles 利用视觉谜题探究大型视觉语言模型的推理能力，揭示其模式匹配局限性 multimodal
15 VIAFormer: Voxel-Image Alignment Transformer for High-Fidelity Voxel Refinement 提出VIAFormer，用于多视角图像引导下的高保真体素精细化 foundation model
16 Face-Voice Association with Inductive Bias for Maximum Class Separation 提出基于归纳偏置的最大类间分离人脸-语音关联方法 multimodal
17 ChartVerse: Scaling Chart Reasoning via Reliable Programmatic Synthesis from Scratch ChartVerse：通过可靠的程序化从零合成，扩展图表推理能力 chain-of-thought

🔬 支柱三：空间感知与语义 (Perception & Semantics) (7 篇)

#题目一句话要点标签🔗⭐
18 Learning Fine-Grained Correspondence with Cross-Perspective Perception for Open-Vocabulary 6D Object Pose Estimation 提出FiCoP框架，通过跨视角感知和细粒度匹配，提升开放词汇6D物体姿态估计的鲁棒性。 open-vocabulary open vocabulary ✅
19 Rig-Aware 3D Reconstruction of Vehicle Undercarriages using Gaussian Splatting 提出一种基于高斯溅射的车辆底盘三维重建方法，用于提升检测效率和买家信心。 gaussian splatting splatting
20 OmniOVCD: Streamlining Open-Vocabulary Change Detection with SAM 3 提出OmniOVCD框架，利用SAM进行开放词汇变化检测，实现SOTA性能。 open-vocabulary open vocabulary
21 One-Shot Refiner: Boosting Feed-forward Novel View Synthesis via One-Step Diffusion 提出One-Shot Refiner，通过单步扩散提升前馈式新视角合成质量 3D gaussian splatting 3DGS gaussian splatting
22 ParkingTwin: Training-Free Streaming 3D Reconstruction for Parking-Lot Digital Twins ParkingTwin：无需训练的停车场数字孪生在线3D重建 3D gaussian splatting 3DGS gaussian splatting ✅
23 Vision-Based Natural Language Scene Understanding for Autonomous Driving: An Extended Dataset and a New Model for Traffic Scene Description Generation 提出一种基于视觉的自然语言场景理解框架，用于自动驾驶交通场景描述生成。 scene understanding
24 Two-Stream temporal transformer for video action classification 提出双流时序Transformer，用于视频动作分类，提升时空信息利用率。 optical flow

🔬 支柱二：RL算法与架构 (RL & Architecture) (6 篇)

#题目一句话要点标签🔗⭐
25 Revisiting Multi-Task Visual Representation Learning 提出MTV多任务视觉预训练框架，融合视觉-语言和自监督学习优势，提升空间推理能力。 representation learning MAE visual pre-training
26 PAS-Mamba: Phase-Amplitude-Spatial State Space Model for MRI Reconstruction 提出PAS-Mamba模型，解耦相位-幅度信息，提升MRI重建质量 Mamba state space model
27 Glance-or-Gaze: Incentivizing LMMs to Adaptively Focus Search via Reinforcement Learning 提出Glance-or-Gaze框架，通过强化学习自适应地聚焦搜索，提升LMMs在知识密集型视觉问答中的性能。 reinforcement learning multimodal
28 DIS2: Disentanglement Meets Distillation with Classwise Attention for Robust Remote Sensing Segmentation under Missing Modalities 提出DIS2框架，通过解耦和蒸馏学习，提升遥感图像分割在模态缺失下的鲁棒性。 distillation multimodal
29 Gaussian Based Adaptive Multi-Modal 3D Semantic Occupancy Prediction 提出基于高斯模型的自适应多模态3D语义占据预测方法，提升自动驾驶安全性。 Mamba state space model multimodal
30 ASBA: A-line State Space Model and B-line Attention for Sparse Optical Doppler Tomography Reconstruction 提出ASBA网络，利用A线状态空间模型和B线注意力机制，实现稀疏光多普勒断层扫描重建。 state space model

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

#题目一句话要点标签🔗⭐
31 Facial Spatiotemporal Graphs: Leveraging the 3D Facial Surface for Remote Physiological Measurement 提出面部时空图STGraph，用于利用3D面部表面进行远程生理信号测量 spatiotemporal ✅

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

#题目一句话要点标签🔗⭐
32 Interp3D: Correspondence-aware Interpolation for Generative Textured 3D Morphing Interp3D：提出一种基于对应关系的插值方法，用于生成具有纹理的3D形变 geometric consistency ✅

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
33 Motion 3-to-4: 3D Motion Reconstruction for 4D Synthesis Motion 3-to-4：单目视频驱动的4D动态物体高质量合成框架 motion latent ✅

⬅️ 返回 cs.CV 首页 · 🏠 返回主页