cs.CV（2026-03-24）

📊 共 55 篇论文 | 🔗 17 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (20 🔗6) 支柱二：RL算法与架构 (RL & Architecture) (11 🔗3) 支柱三：空间感知与语义 (Perception & Semantics) (9 🔗3) 支柱一：机器人控制 (Robot Control) (5 🔗1) 支柱六：视频提取与匹配 (Video Extraction) (3) 支柱五：交互与反应 (Interaction & Reaction) (2 🔗2) 支柱八：物理动画 (Physics-based Animation) (2 🔗2) 支柱七：动作重定向 (Motion Retargeting) (2) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (20 篇)

#题目一句话要点标签🔗⭐
1 ENC-Bench: A Benchmark for Evaluating Multimodal Large Language Models in Electronic Navigational Chart Understanding 提出ENC-Bench，用于评估多模态大语言模型在电子海图理解中的能力。 large language model multimodal symbolic grounding
2 YOLOv10 with Kolmogorov-Arnold networks and vision-language foundation models for interpretable object detection and trustworthy multimodal AI in computer vision perception 提出基于Kolmogorov-Arnold网络和视觉-语言模型的YOLOv10，用于可解释的目标检测和可信赖的多模态AI foundation model multimodal
3 MLLM-HWSI: A Multimodal Large Language Model for Hierarchical Whole Slide Image Understanding MLLM-HWSI：用于分层全切片图像理解的多模态大语言模型 large language model multimodal ✅
4 ForestPrune: High-ratio Visual Token Compression for Video Multimodal Large Language Models via Spatial-Temporal Forest Modeling ForestPrune：通过时空森林建模实现视频多模态大语言模型的高比例视觉Token压缩 large language model multimodal
5 SpecEyes: Accelerating Agentic Multimodal LLMs via Speculative Perception and Planning SpecEyes：通过推测性感知与规划加速Agentic多模态LLM large language model multimodal
6 GeoTikzBridge: Advancing Multimodal Code Generation for Geometric Perception and Reasoning GeoTikzBridge：通过Tikz代码生成增强多模态大模型几何感知与推理能力 large language model multimodal ✅
7 3DCity-LLM: Empowering Multi-modality Large Language Models for 3D City-scale Perception and Understanding 提出3DCity-LLM，赋能多模态大语言模型进行3D城市级感知与理解 large language model ✅
8 Multimodal Industrial Anomaly Detection via Geometric Prior 提出基于几何先验的多模态工业异常检测网络，提升复杂几何缺陷检测精度。 multimodal
9 UniFunc3D: Unified Active Spatial-Temporal Grounding for 3D Functionality Segmentation UniFunc3D：统一的主动时空定位框架，用于3D功能分割 large language model multimodal
10 ViKey: Enhancing Temporal Understanding in Videos via Visual Prompting ViKey：通过视觉提示增强视频大语言模型的时间理解能力 large language model multimodal
11 SMSP: A Plug-and-Play Strategy of Multi-Scale Perception for MLLMs to Perceive Visual Illusions 提出SMSP多尺度感知策略，提升MLLM对视觉错觉的识别能力。 large language model multimodal ✅
12 Cog3DMap: Multi-View Vision-Language Reasoning with 3D Cognitive Maps Cog3DMap：利用3D认知地图实现多视角视觉-语言推理 large language model multimodal
13 ForeSea: AI Forensic Search with Multi-modal Queries for Video Surveillance ForeSea：面向视频监控的多模态查询AI取证搜索系统 large language model multimodal
14 Focus, Don't Prune: Identifying Instruction-Relevant Regions for Information-Rich Image Understanding PinPoint：聚焦而非剪枝，识别信息密集图像中指令相关区域，提升视觉语言模型效率。 large language model multimodal
15 Know3D: Prompting 3D Generation with Knowledge from Vision-Language Models Know3D：利用视觉-语言模型知识提示3D生成，实现可控后视图生成。 large language model multimodal
16 OccAny: Generalized Unconstrained Urban 3D Occupancy OccAny：首个广义无约束城市3D Occupancy预测模型，提升泛化性和几何补全能力。 foundation model ✅
17 DetPO: In-Context Learning with Multi-Modal LLMs for Few-Shot Object Detection DetPO：利用多模态LLM的上下文学习进行少样本目标检测，提升泛化能力。 visual grounding ✅
18 AgentFoX: LLM Agent-Guided Fusion with eXplainability for AI-Generated Image Detection AgentFoX：基于LLM Agent引导的AI生成图像检测与可解释性融合框架 large language model
19 SOUPLE: Enhancing Audio-Visual Localization and Segmentation with Learnable Prompt Contexts SOUPLE：利用可学习提示上下文增强音视频定位与分割 multimodal
20 Think 360°: Evaluating the Width-centric Reasoning Capability of MLLMs Beyond Depth 提出Think 360°基准，评估多模态大模型在推理宽度上的能力。 multimodal

🔬 支柱二：RL算法与架构 (RL & Architecture) (11 篇)

#题目一句话要点标签🔗⭐
21 Rethinking Token-Level Policy Optimization for Multimodal Chain-of-Thought 提出感知-探索策略优化(PEPO)，提升多模态CoT推理中视觉 grounding 和探索推理的平衡。 reinforcement learning multimodal visual grounding ✅
22 PhotoAgent: A Robotic Photographer with Spatial and Aesthetic Understanding PhotoAgent：融合空间与美学理解的机器人摄影师 world model 3D gaussian splatting 3DGS
23 EVA: Efficient Reinforcement Learning for End-to-End Video Agent 提出EVA：基于强化学习的高效端到端视频Agent，用于解决长视频理解难题。 reinforcement learning large language model multimodal ✅
24 Cross-Slice Knowledge Transfer via Masked Multi-Modal Heterogeneous Graph Contrastive Learning for Spatial Gene Expression Inference 提出SpaHGC模型，通过跨切片知识迁移提升空间基因表达推断精度。 contrastive learning spatial relationship foundation model
25 UniGRPO: Unified Policy Optimization for Reasoning-Driven Visual Generation 提出UniGRPO，用于推理驱动的视觉生成统一策略优化 reinforcement learning flow matching classifier-free guidance
26 Mamba-driven MRI-to-CT Synthesis for MRI-only Radiotherapy Planning 提出基于Mamba的MRI-to-CT合成方法，用于MRI引导的放疗计划。 Mamba geometric consistency
27 Conformal Cross-Modal Active Learning 提出CCMA，利用多模态知识提升视觉主动学习的数据效率。 teacher-student foundation model multimodal
28 WildWorld: A Large-Scale Dataset for Dynamic World Modeling with Actions and Explicit State toward Generative ARPG 提出WildWorld：一个大规模动作条件动态世界建模数据集，用于生成式ARPG。 reinforcement learning world model ✅
29 FCL-COD: Weakly Supervised Camouflaged Object Detection with Frequency-aware and Contrastive Learning 提出FCL-COD框架，通过频率感知和对比学习解决弱监督伪装目标检测难题。 representation learning contrastive learning
30 Curriculum-Driven 3D CT Report Generation via Language-Free Visual Grafting and Zone-Constrained Compression 提出Ker-VLJEPA-3B，通过课程学习和无语言视觉嫁接生成3D CT报告 curriculum learning large language model
31 Dual-Teacher Distillation with Subnetwork Rectification for Black-Box Domain Adaptation 提出DDSR模型，通过双教师蒸馏和子网络校正解决黑盒领域自适应问题 distillation

🔬 支柱三：空间感知与语义 (Perception & Semantics) (9 篇)

#题目一句话要点标签🔗⭐
32 Pose-Free Omnidirectional Gaussian Splatting for 360-Degree Videos with Consistent Depth Priors 提出PFGS360，实现无位姿全景视频的3D高斯重建与高质量新视角合成 monocular depth 3D gaussian splatting 3DGS ✅
33 Predictive Photometric Uncertainty in Gaussian Splatting for Novel View Synthesis 针对新视角合成，提出高斯溅射光度不确定性预测框架，提升空间地图可靠性 3D gaussian splatting gaussian splatting splatting
34 Looking Beyond the Window: Global-Local Aligned CLIP for Training-free Open-Vocabulary Semantic Segmentation 提出GLA-CLIP，通过全局-局部对齐CLIP模型，实现免训练开放词汇语义分割。 open-vocabulary open vocabulary ✅
35 Generative Event Pretraining with Foundation Model Alignment 提出GEP：通过对齐视觉基础模型进行生成式事件预训练 depth estimation foundation model
36 DA-Flow: Degradation-Aware Optical Flow Estimation with Diffusion Models DA-Flow：基于扩散模型的退化感知光流估计，提升真实场景下的鲁棒性 optical flow
37 UniQueR: Unified Query-based Feedforward 3D Reconstruction UniQueR：一种用于高效精确三维重建的统一查询式前馈框架 NeRF VGGT
38 One View Is Enough! Monocular Training for In-the-Wild Novel View Generation 提出OVIE，仅用单视角图像训练，实现野外场景的新视角生成 monocular depth ✅
39 SLARM: Streaming and Language-Aligned Reconstruction Model for Dynamic Scenes SLARM：用于动态场景的流式语言对齐重建模型 scene reconstruction
40 Group Editing : Edit Multiple Images in One Go 提出GroupEditing框架，用于对一组相关图像进行一致性编辑。 VGGT

🔬 支柱一：机器人控制 (Robot Control) (5 篇)

#题目一句话要点标签🔗⭐
41 ABot-PhysWorld: Interactive World Foundation Model for Robotic Manipulation with Physics Alignment ABot-PhysWorld：基于物理对齐的交互式世界基础模型，用于机器人操作 manipulation DPO world model
42 TETO: Tracking Events with Teacher Observation for Motion Estimation and Frame Interpolation TETO：利用教师观测追踪事件，用于运动估计和帧插值 sim-to-real teacher-student distillation
43 VLA-IAP: Training-Free Visual Token Pruning via Interaction Alignment for Vision-Language-Action Models VLA-IAP：通过交互对齐实现免训练视觉Token剪枝，加速VLA模型推理。 manipulation vision-language-action VLA ✅
44 Gaze-Regularized Vision-Language-Action Models for Robotic Manipulation 提出基于注视正则化的VLA模型，提升机器人操作任务性能。 manipulation vision-language-action VLA
45 RealMaster: Lifting Rendered Scenes into Photorealistic Video RealMaster：利用视频扩散模型将渲染场景提升为照片级真实视频 sim-to-real

🔬 支柱六：视频提取与匹配 (Video Extraction) (3 篇)

#题目一句话要点标签🔗⭐
46 Unleashing Spatial Reasoning in Multimodal Large Language Models via Textual Representation Guided Reasoning TRACE：通过文本引导多模态大模型进行3D空间推理 egocentric large language model multimodal
47 GSwap: Realistic Head Swapping with Dynamic Neural Gaussian Field GSwap：利用动态神经高斯场实现逼真头部替换 SMPL SMPL-X
48 Gaze-Regularized VLMs for Ego-Centric Behavior Understanding 提出基于注视正则化的VLM，用于提升以自我为中心的行为理解能力 egocentric

🔬 支柱五：交互与反应 (Interaction & Reaction) (2 篇)

#题目一句话要点标签🔗⭐
49 InterDyad: Interactive Dyadic Speech-to-Video Generation by Querying Intermediate Visual Guidance 提出InterDyad框架，通过查询中间视觉引导实现交互式双人语音到视频生成。 two-person interaction dyadic interaction large language model ✅
50 A Feature Shuffling and Restoration Strategy for Universal Unsupervised Anomaly Detection 提出特征洗牌与恢复策略以解决通用无监督异常检测问题 ReMoS ✅

🔬 支柱八：物理动画 (Physics-based Animation) (2 篇)

#题目一句话要点标签🔗⭐
51 WaveSFNet: A Wavelet-Based Codec and Spatial--Frequency Dual-Domain Gating Network for Spatiotemporal Prediction WaveSFNet：基于小波编解码和空频双域门控网络的时空预测 spatiotemporal ✅
52 MultiCam: On-the-fly Multi-Camera Pose Estimation Using Spatiotemporal Overlaps of Known Objects 提出MultiCam，利用时空重叠的已知物体进行动态多相机位姿估计 spatiotemporal ✅

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

#题目一句话要点标签🔗⭐
53 A Synchronized Audio-Visual Multi-View Capture System 提出一种同步音视频多视角采集系统，用于细粒度会话行为分析。 human motion
54 Object Pose Transformer: Unifying Unseen Object Pose Estimation Object Pose Transformer：统一的无监督物体姿态估计框架 geometric consistency

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
55 SIMART: Decomposing Monolithic Meshes into Sim-ready Articulated Assets via MLLM SIMART：通过MLLM将整体网格分解为可用于仿真的铰接资产 VQ-VAE embodied AI

⬅️ 返回 cs.CV 首页 · 🏠 返回主页