cs.CV（2025-09-25）

📊 共 16 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) 支柱二：RL算法与架构 (RL & Architecture) (5 🔗1) 支柱三：空间感知与语义 (Perception & Semantics) (2 🔗1) 支柱八：物理动画 (Physics-based Animation) (1 🔗1) 支柱四：生成式动作 (Generative Motion) (1)

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

#题目一句话要点标签🔗⭐
1 Reasoning-Enhanced Domain-Adaptive Pretraining of Multimodal Large Language Models for Short Video Content Governance 提出推理增强的领域自适应多模态大语言模型预训练方法，用于短视频内容治理 large language model multimodal chain-of-thought
2 Instruction-tuned Self-Questioning Framework for Multimodal Reasoning 提出基于指令调优的自问框架SQ-InstructBLIP，用于增强多模态推理能力 large language model multimodal
3 X-CoT: Explainable Text-to-Video Retrieval via LLM-based Chain-of-Thought Reasoning 提出X-CoT，利用LLM链式思考推理实现可解释的文本到视频检索 chain-of-thought ✅
4 VideoJudge: Bootstrapping Enables Scalable Supervision of MLLM-as-a-Judge for Video Understanding VideoJudge：通过自举法实现MLLM作为视频理解评判器的可扩展监督 large language model multimodal chain-of-thought
5 A Sentinel-3 foundation model for ocean colour 提出基于Sentinel-3的海洋颜色基础模型，提升海洋观测任务性能 foundation model
6 Decipher-MR: A Vision-Language Foundation Model for 3D MRI Representations Decipher-MR：用于3D MRI表征的视觉-语言基础模型 foundation model
7 CompareBench: A Benchmark for Visual Comparison Reasoning in Vision-Language Models 提出CompareBench，用于评估视觉语言模型中的视觉比较推理能力 multimodal

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

#题目一句话要点标签🔗⭐
8 MMR1: Enhancing Multimodal Reasoning with Variance-Aware Sampling and Open Resources MMR1：通过方差感知采样和开放资源增强多模态推理能力 reinforcement learning multimodal chain-of-thought ✅
9 SlideMamba: Entropy-Based Adaptive Fusion of GNN and Mamba for Enhanced Representation Learning in Digital Pathology SlideMamba：结合GNN与Mamba的熵自适应融合框架，提升数字病理学表征学习 predictive model Mamba representation learning
10 FantasyWorld: Geometry-Consistent World Modeling via Unified Video and 3D Prediction FantasyWorld：通过统一视频和3D预测实现几何一致的世界建模 world model foundation model
11 X-Streamer: Unified Human World Modeling with Audiovisual Interaction X-Streamer：提出基于视听交互的统一人类世界建模框架，实现数字人实时交互。 world model multimodal
12 Enhancing Contrastive Learning for Geolocalization by Discovering Hard Negatives on Semivariograms 提出基于Semivariogram的对比学习方法，提升图像地理定位精度 contrastive learning

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

#题目一句话要点标签🔗⭐
13 Dense Semantic Matching with VGGT Prior 提出基于VGGT先验的稠密语义匹配方法，提升几何感知和匹配可靠性 VGGT foundation model
14 Quantized Visual Geometry Grounded Transformer 提出QuantVGGT，解决VGGT量化难题，实现资源受限场景下的高效3D重建。 VGGT ✅

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

#题目一句话要点标签🔗⭐
15 MORPH: PDE Foundation Models with Arbitrary Data Modality 提出MORPH模型以处理多模态偏微分方程数据 spatiotemporal foundation model multimodal ✅

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
16 Does FLUX Already Know How to Perform Physically Plausible Image Composition? 提出SHINE框架，无需训练即可实现物理上合理的图像合成 physically plausible

⬅️ 返回 cs.CV 首页 · 🏠 返回主页