cs.CV（2026-02-20）

📊 共 23 篇论文 | 🔗 4 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (7 🔗1) 支柱一：机器人控制 (Robot Control) (6 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (5 🔗1) 支柱四：生成式动作 (Generative Motion) (2) 支柱三：空间感知与语义 (Perception & Semantics) (2) 支柱八：物理动画 (Physics-based Animation) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (7 篇)

#题目一句话要点标签🔗⭐
1 Comparative Assessment of Multimodal Earth Observation Data for Soil Moisture Estimation 结合多模态遥感数据，提出一种高分辨率土壤湿度估计框架，适用于欧洲农田监测。 foundation model multimodal
2 JAEGER: Joint 3D Audio-Visual Grounding and Reasoning in Simulated Physical Environments JAEGER：提出基于神经强度向量的3D音视频联合理解框架，解决复杂物理环境中空间推理难题。 large language model visual grounding
3 Rodent-Bench Rodent-Bench：用于评估多模态大模型在啮齿动物行为标注能力的新基准 large language model multimodal
4 On the Adversarial Robustness of Discrete Image Tokenizers 研究离散图像Tokenizers的对抗鲁棒性，提出高效攻击与无监督防御方法。 foundation model multimodal
5 3DMedAgent: Unified Perception-to-Understanding for 3D Medical Analysis 提出3DMedAgent，利用2D MLLM实现3D医学影像的统一感知与理解。 large language model multimodal ✅
6 Towards LLM-centric Affective Visual Customization via Efficient and Precise Emotion Manipulating 提出基于LLM的情感视觉定制框架L-AVC，实现高效精确的情感操控。 foundation model multimodal
7 Spatio-temporal Decoupled Knowledge Compensator for Few-Shot Action Recognition 提出DiST框架，利用解耦时空知识补偿器提升少样本动作识别性能 large language model

🔬 支柱一：机器人控制 (Robot Control) (6 篇)

#题目一句话要点标签🔗⭐
8 BLM-Guard: Explainable Multimodal Ad Moderation with Chain-of-Thought and Policy-Aligned Rewards 提出BLM-Guard，利用CoT和策略对齐奖励进行可解释的多模态广告审核。 manipulation reinforcement learning multimodal
9 ROCKET: Residual-Oriented Multi-Layer Alignment for Spatially-Aware Vision-Language-Action Models ROCKET：面向残差的多层对齐框架，提升具身智能体空间感知能力 manipulation vision-language-action VLA ✅
10 UAOR: Uncertainty-aware Observation Reinjection for Vision-Language-Action Models 提出不确定性感知观测重注入(UAOR)模块，提升VLA模型在机器人操作任务中的性能。 manipulation vision-language-action VLA
11 Generated Reality: Human-centric World Simulation using Interactive Video Generation with Hand and Camera Control 提出基于手部和相机控制的交互式视频生成方法，用于人机交互世界模拟。 dexterous hand world model egocentric
12 CapNav: Benchmarking Vision Language Models on Capability-conditioned Indoor Navigation 提出Capability-Conditioned Navigation以解决室内导航中的能力约束问题 quadruped VLN ✅
13 Dual-Channel Attention Guidance for Training-Free Image Editing Control in Diffusion Transformers 提出双通道注意力引导（DCAG），用于Diffusion Transformer的免训练图像编辑控制。 manipulation

🔬 支柱二：RL算法与架构 (RL & Architecture) (5 篇)

#题目一句话要点标签🔗⭐
14 VLANeXt: Recipes for Building Strong VLA Models VLANeXt：通过系统性实验，为构建强大的视觉-语言-动作模型提供有效方案。 policy learning vision-language-action VLA
15 MUOT_3M: A 3 Million Frame Multimodal Underwater Benchmark and the MUTrack Tracking Method 提出MUOT_3M水下多模态数据集与MUTrack跟踪方法，提升水下目标跟踪性能。 distillation multimodal
16 Spatio-Spectroscopic Representation Learning using Unsupervised Convolutional Long-Short Term Memory Networks 提出基于卷积LSTM自编码器的无监督时空光谱表征学习框架，用于星系演化研究。 representation learning
17 Self-Aware Object Detection via Degradation Manifolds 提出基于退化流形的自感知目标检测框架，解决恶劣条件下检测器失效问题 contrastive learning zero-shot transfer
18 SARAH: Spatially Aware Real-time Agentic Humans SARAH：提出一种空间感知实时Agentic人类建模方法，用于VR和数字人应用。 flow matching classifier-free guidance ✅

🔬 支柱四：生成式动作 (Generative Motion) (2 篇)

#题目一句话要点标签🔗⭐
19 Temporal Consistency-Aware Text-to-Motion Generation 提出TCA-T2M框架，解决文本到动作生成中时序一致性问题。 text-to-motion motion generation VQ-VAE
20 A Self-Supervised Approach on Motion Calibration for Enhancing Physical Plausibility in Text-to-Motion 提出失真感知运动校准器DMC，提升文本到动作生成中的物理真实性。 text-to-motion motion generation physically plausible

🔬 支柱三：空间感知与语义 (Perception & Semantics) (2 篇)

#题目一句话要点标签🔗⭐
21 Diff2DGS: Reliable Reconstruction of Occluded Surgical Scenes via 2D Gaussian Splatting Diff2DGS：基于2D高斯溅射可靠重建遮挡手术场景 gaussian splatting splatting
22 Unifying Color and Lightness Correction with View-Adaptive Curve Adjustment for Robust 3D Novel View Synthesis Luminance-GS++：提出视角自适应曲线调整，统一校正颜色和亮度，实现鲁棒的3D新视角合成 3D gaussian splatting 3DGS gaussian splatting

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

#题目一句话要点标签🔗⭐
23 Narrating For You: Prompt-guided Audio-visual Narrating Face Generation Employing Multi-entangled Latent Space 提出一种基于多重纠缠潜在空间的提示引导式音视频叙事人脸生成方法 spatiotemporal

⬅️ 返回 cs.CV 首页 · 🏠 返回主页