cs.CV（2026-04-17）

📊 共 37 篇论文 | 🔗 13 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (11 🔗2) 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) (7 🔗3) 支柱一：机器人控制 (Robot Control) (5 🔗2) 支柱六：视频提取与匹配 (Video Extraction) (2 🔗1) 支柱八：物理动画 (Physics-based Animation) (2 🔗2) 支柱七：动作重定向 (Motion Retargeting) (2 🔗1) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (11 篇)

#题目一句话要点标签🔗⭐
1 Chain-of-Thought Degrades Visual Spatial Reasoning Capabilities of Multimodal LLMs 发现思维链推理降低多模态LLM的视觉空间推理能力 multimodal chain-of-thought
2 neuralCAD-Edit: An Expert Benchmark for Multimodal-Instructed 3D CAD Model Editing 提出neuralCAD-Edit：一个面向多模态指令的3D CAD模型编辑专家基准。 foundation model multimodal ✅
3 PixDLM: A Dual-Path Multimodal Language Model for UAV Reasoning Segmentation 提出PixDLM，用于无人机遥感图像推理分割，并构建DRSeg基准数据集。 multimodal chain-of-thought
4 GAViD: A Large-Scale Multimodal Dataset for Context-Aware Group Affect Recognition from Videos GAViD：大规模多模态数据集，用于视频中上下文感知群体情感识别 multimodal
5 Beyond a Single Frame: Multi-Frame Spatially Grounded Reasoning Across Volumetric MRI 提出SGMRI-VQA基准，用于评估医学VLM在体积MRI上的多帧空间推理能力。 visual grounding chain-of-thought
6 RefereeBench: Are Video MLLMs Ready to be Multi-Sport Referees 提出RefereeBench，评估视频多模态大模型在多体育项目裁判任务中的能力。 large language model multimodal
7 SIMMER: Cross-Modal Food Image--Recipe Retrieval via MLLM-Based Embedding SIMMER：利用MLLM嵌入实现跨模态食物图像-食谱检索 large language model multimodal
8 VEFX-Bench: A Holistic Benchmark for Generic Video Editing and Visual Effects 提出VEFX-Bench，用于评估通用视频编辑和视觉效果的综合基准。 instruction following
9 Information Router for Mitigating Modality Dominance in Vision-Language Models 提出多模态信息路由（MoIR）以缓解视觉-语言模型中的模态主导问题。 large language model
10 Do Vision-Language Models Truly Perform Vision Reasoning? A Rigorous Study of the Modality Gap 提出CrossMath基准，揭示视觉语言模型在视觉推理上的局限性 multimodal ✅
11 Where Do Vision-Language Models Fail? World Scale Analysis for Image Geolocalization 评估视觉-语言模型在图像地理定位中的表现与局限性 multimodal

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

#题目一句话要点标签🔗⭐
12 Splats in Splats++: Robust and Generalizable 3D Gaussian Splatting Steganography 提出Splats in Splats++以解决3D高斯点云隐写问题 3D gaussian splatting 3DGS 3D reconstruction
13 AdaVFM: Adaptive Vision Foundation Models for Edge Intelligence via LLM-Guided Execution AdaVFM：通过LLM引导的自适应视觉基础模型实现边缘智能 open-vocabulary open vocabulary large language model
14 Neural Gabor Splatting: Enhanced Gaussian Splatting with Neural Gabor for High-frequency Surface Reconstruction 提出神经Gabor Splatting，增强高频表面重建的3D高斯溅射 3D gaussian splatting 3DGS 3D reconstruction
15 SENSE: Stereo OpEN Vocabulary SEmantic Segmentation SENSE：利用立体视觉增强开放词汇语义分割，提升空间精度 scene understanding open-vocabulary open vocabulary
16 CLOTH-HUGS: Cloth Aware Human Gaussian Splatting Cloth-HUGS：基于高斯溅射的服装感知人体重建，解耦身体与服装 gaussian splatting splatting SMPL
17 Towards Realistic Open-Vocabulary Remote Sensing Segmentation: Benchmark and Baseline 提出OVRSISBenchV2以解决开放词汇遥感图像分割问题 open-vocabulary open vocabulary ✅
18 PLAF: Pixel-wise Language-Aligned Feature Extraction for Efficient 3D Scene Understanding 提出PLAF，实现像素级语言对齐特征提取，提升高效3D场景理解能力 scene understanding open-vocabulary open vocabulary ✅

🔬 支柱二：RL算法与架构 (RL & Architecture) (7 篇)

#题目一句话要点标签🔗⭐
19 SSMamba: A Self-Supervised Hybrid State Space Model for Pathological Image Classification SSMamba：一种用于病理图像分类的自监督混合状态空间模型 Mamba state space model foundation model
20 Find, Fix, Reason: Context Repair for Video Reasoning 提出Find, Fix, Reason框架，通过上下文修复提升视频推理性能 reinforcement learning spatiotemporal instruction following ✅
21 CollideNet: Hierarchical Multi-scale Video Representation Learning with Disentanglement for Time-To-Collision Forecasting CollideNet：用于碰撞时间预测的分层多尺度解耦视频表示学习 representation learning spatiotemporal ✅
22 MambaBack: Bridging Local Features and Global Contexts in Whole Slide Image Analysis MambaBack：结合局部特征与全局上下文的病理切片图像分析方法 Mamba SSM
23 CLIMB: Controllable Longitudinal Brain Image Generation using Mamba-based Latent Diffusion Model and Gaussian-aligned Autoencoder CLIMB：基于Mamba和高斯对齐自编码器的可控纵向脑部图像生成 Mamba state space model
24 The Amazing Stability of Flow Matching Flow Matching模型展现惊人的稳定性，对数据和架构扰动不敏感 flow matching
25 Repurposing 3D Generative Model for Autoregressive Layout Generation LaviGen：利用3D生成模型进行自回归布局生成，提升场景物理合理性与生成效率 distillation physically plausible ✅

🔬 支柱一：机器人控制 (Robot Control) (5 篇)

#题目一句话要点标签🔗⭐
26 Mind's Eye: A Benchmark of Visual Abstraction, Transformation and Composition for Multimodal LLMs 提出Mind's Eye基准测试，评估多模态LLM的视觉抽象、转换和组合能力 manipulation large language model multimodal
27 DINOv3 Beats Specialized Detectors: A Simple Foundation Model Baseline for Image Forensics 提出基于DINOv3的图像取证基线模型，性能超越专用检测器 manipulation foundation model ✅
28 Continual Hand-Eye Calibration for Open-world Robotic Manipulation 提出一种持续手眼标定框架，解决开放世界机器人操作中的灾难性遗忘问题。 manipulation distillation
29 From Competition to Coopetition: Coopetitive Training-Free Image Editing Based on Text Guidance 提出CoEdit，通过竞争合作训练实现文本引导的免训练图像编辑 manipulation ✅
30 AHS: Adaptive Head Synthesis via Synthetic Data Augmentations 提出AHS，通过合成数据增强实现自适应头部合成，解决现有头部替换方法在真实场景中的局限性。 manipulation

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

#题目一句话要点标签🔗⭐
31 FineCog-Nav: Integrating Fine-grained Cognitive Modules for Zero-shot Multimodal UAV Navigation 提出FineCog-Nav以解决无人机视觉语言导航中的零-shot挑战 egocentric VLN foundation model ✅
32 Watching Movies Like a Human: Egocentric Emotion Understanding for Embodied Companions 提出EgoScreen-Emotion数据集，用于具身智能体在主视角屏幕观看电影时的情感理解。 egocentric multimodal

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

#题目一句话要点标签🔗⭐
33 NeuroLip: An Event-driven Spatiotemporal Learning Framework for Cross-Scene Lip-Motion-based Visual Speaker Recognition NeuroLip：一种事件驱动的时空学习框架，用于跨场景的唇动视觉说话人识别 spatiotemporal ✅
34 LP$^{2}$DH: A Locality-Preserving Pixel-Difference Hashing Framework for Dynamic Texture Recognition 提出LP²DH框架以解决动态纹理识别中的高维特征问题 spatiotemporal ✅

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

#题目一句话要点标签🔗⭐
35 AIFIND: Artifact-Aware Interpreting Fine-Grained Alignment for Incremental Face Forgery Detection 提出AIFIND，通过伪造痕迹对齐实现增量人脸伪造检测 geometric consistency
36 APC: Transferable and Efficient Adversarial Point Counterattack for Robust 3D Point Cloud Recognition 提出对抗点云反击(APC)，提升3D点云识别模型对抗攻击的鲁棒性和迁移性。 geometric consistency ✅

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
37 Motion-Adapter: A Diffusion Model Adapter for Text-to-Motion Generation of Compound Actions 提出Motion-Adapter，解决文本到复合动作生成中的动作覆盖和注意力崩溃问题 motion diffusion model motion diffusion text-to-motion

⬅️ 返回 cs.CV 首页 · 🏠 返回主页