PaperNotes

一个由人类审阅、Claude Code 和 Codex 协作更新的论文阅读笔记站。

Agent 负责生成初稿和维护页面，我负责检查理解是否正确、指出问题并推动修改。这里不是自动摘要合集，而是人类判断与 AI 执行力共同产出的阅读记录。

按时间 按标签
20 VLA 16 WAM 14 世界模型 11 视频生成 10 强化学习 7 泛化 6 3D重建 6 多模态理解 5 综述 5 自动驾驶 5 视觉推理 4 链式思考 3 扩散模型 3 语言模型 2 知识蒸馏 2 音视频推理 1 图像编辑 1 导航 1 机器人
🏷️ VLA 20 篇
↑ 回到标签列表
机器人如何从经验中变得更好？｜具身 RSI 与 TTT 研究博客
研究综述 20262026-10-11
ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?
arXiv: 2606.19531投稿（under review）2026-06-25
World Action Models: A Survey — 阅读笔记
arXiv: 2606.20781arXiv 20262026-06-25
Qwen-RobotManip Technical Report: Alignment Unlocks Scale for Robotic Manipulation Foundation Models
arXiv: 2606.17846Qwen Technical Report 20262026-06-18
Qwen-RobotNav Technical Report: A Scalable Navigation Model Designed for an Agentic Navigation System
arXiv: 2606.18112Qwen Technical Report 20262026-06-18
Geometric Action Model for Robot Policy Learning
arXiv: 2606.17046CoRL 2026（投稿，under review）2026-06-17
Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments
arXiv: 2605.30280千问VLA技术报告2026-05-29
Understanding the Impact of Geometric Foundation Models on Vision-Language-Action Models
arXiv: 2605.24642ECCV 2026（投稿中）2026-05-26
OASIS: Observation-Action Space Alignment via SE(3) Trajectory Prediction for Robotic Manipulation
arXiv: 2605.25829NeurIPS 2026（投稿中）2026-05-26
QuoVLA: Quotient Space for Vision-Language-Action Models
arXiv: 2605.24890NeurIPS 2026（投稿中）2026-05-26
X-Foresight: A Joint Vision-Action Causal Forecasting Network via Predictive World Modeling
arXiv: 2605.24892NeurIPS 2025（投稿中）2026-05-26
WMPO: World Model-based Policy Optimization for Vision-Language-Action Models
arXiv: 2511.09515ICLR 20262026-05-25
World Action Models: The Next Frontier in Embodied AI
arXiv: 2605.12090arXiv 综述 20262026-05-25
SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning
arXiv: 2509.09674ICLR 20262026-05-22
AttenA+: Rectifying Action Inequality in Robotic Foundation Models
arXiv: 2605.13548NeurIPS 2026（投稿中）2026-05-21
CKT-WAM: Parameter-Efficient Context Knowledge Transfer Between World Action Models
arXiv: 2605.06247NeurIPS 2026（投稿中）2026-05-21
MindVLA-U1: VLA Beats VA with Unified Streaming Architecture for Autonomous Driving
arXiv: 2605.12624NeurIPS 2026（投稿中）2026-05-14
ConsisVLA-4D 论文阅读笔记
arXiv: 2605.05126CVPR 20262026-05-10
LaST-R1 论文分享简明笔记
arXiv: 2604.28192ICML 2026 Spotlight2026-05-08
LaST-R1 论文与代码笔记
arXiv: 2604.28192ICML 2026 Spotlight2026-05-08
🏷️ WAM 16 篇
↑ 回到标签列表
ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?
arXiv: 2606.19531投稿（under review）2026-06-25
World Action Models: A Survey — 阅读笔记
arXiv: 2606.20781arXiv 20262026-06-25
WAM-RL: World-Action Model Reinforcement Learning with Reconstruction Rewards and Online Video SFT
arXiv: 2606.17906arXiv 20262026-06-24
Qwen-RobotWorld Technical Report: Unifying Embodied World Modeling through Language-Conditioned Video Generation
arXiv: 2606.17030Qwen Technical Report 20262026-06-18
Geometric Action Model for Robot Policy Learning
arXiv: 2606.17046CoRL 2026（投稿，under review）2026-06-17
World Action Models: The Next Frontier in Embodied AI
arXiv: 2605.12090arXiv 综述 20262026-05-25
World Models and World Action Models: A Survey Note
PaperNotes 综述 20262026-05-22
AttenA+: Rectifying Action Inequality in Robotic Foundation Models
arXiv: 2605.13548NeurIPS 2026（投稿中）2026-05-21
CKT-WAM: Parameter-Efficient Context Knowledge Transfer Between World Action Models
arXiv: 2605.06247NeurIPS 2026（投稿中）2026-05-21
DreamZero: World Action Models are Zero-shot Policies
arXiv: 2602.15922NVIDIA Tech Report 20262026-05-21
Is the Future Compatible? Diagnosing Dynamic Consistency in World Action Models
arXiv: 2605.07514NeurIPS 2026（投稿中）2026-05-21
Fast-WAM: Do World Action Models Need Test-time Future Imagination?
arXiv: 2603.16666NeurIPS 2026（投稿中）2026-05-21
When to Trust Imagination: Adaptive Action Execution for World Action Models
arXiv: 2605.06222NeurIPS 2026（投稿中）2026-05-21
GigaWorld-Policy: An Efficient Action-Centered World-Action Model
arXiv: 2603.17240投稿中 (under review)2026-05-21
HarmoWAM: Harmonizing Generalizable and Precise Manipulation via Adaptive World Action Models
arXiv: 2605.10942NeurIPS 2026（投稿中）2026-05-21
Unified 4D World Action Modeling from Video Priors with Asynchronous Denoising
arXiv: 2604.26694NeurIPS 2026（投稿中）2026-05-21
🏷️ 世界模型 14 篇
↑ 回到标签列表
World Action Models: A Survey — 阅读笔记
arXiv: 2606.20781arXiv 20262026-06-25
WAM-RL: World-Action Model Reinforcement Learning with Reconstruction Rewards and Online Video SFT
arXiv: 2606.17906arXiv 20262026-06-24
Qwen-RobotWorld Technical Report: Unifying Embodied World Modeling through Language-Conditioned Video Generation
arXiv: 2606.17030Qwen Technical Report 20262026-06-18
Geometric Action Model for Robot Policy Learning
arXiv: 2606.17046CoRL 2026（投稿，under review）2026-06-17
X-Foresight: A Joint Vision-Action Causal Forecasting Network via Predictive World Modeling
arXiv: 2605.24892NeurIPS 2025（投稿中）2026-05-26
DiWA: Diffusion Policy Adaptation with World Models
arXiv: 2508.03645CoRL 20252026-05-25
WMPO: World Model-based Policy Optimization for Vision-Language-Action Models
arXiv: 2511.09515ICLR 20262026-05-25
World4RL: Diffusion World Models for Policy Refinement with Reinforcement Learning for Robotic Manipulation
arXiv: 2509.19080投稿 IEEE Trans 期刊2026-05-25
World Action Models: The Next Frontier in Embodied AI
arXiv: 2605.12090arXiv 综述 20262026-05-25
World Models and World Action Models: A Survey Note
PaperNotes 综述 20262026-05-22
DreamZero: World Action Models are Zero-shot Policies
arXiv: 2602.15922NVIDIA Tech Report 20262026-05-21
Unified 4D World Action Modeling from Video Priors with Asynchronous Denoising
arXiv: 2604.26694NeurIPS 2026（投稿中）2026-05-21
LingBot-VA: Causal World Modeling for Robot Control
arXiv: 2601.21998蚂蚁灵波技术报告2026-05-13
OA-WAM: Object-Addressable World Action Model for Robust Robot Manipulation
arXiv: 2605.06481NeurIPS 2026（投稿中）2026-05-13
🏷️ 视频生成 11 篇
↑ 回到标签列表
World Action Models: A Survey — 阅读笔记
arXiv: 2606.20781arXiv 20262026-06-25
WAM-RL: World-Action Model Reinforcement Learning with Reconstruction Rewards and Online Video SFT
arXiv: 2606.17906arXiv 20262026-06-24
Qwen-RobotWorld Technical Report: Unifying Embodied World Modeling through Language-Conditioned Video Generation
arXiv: 2606.17030Qwen Technical Report 20262026-06-18
X-Foresight: A Joint Vision-Action Causal Forecasting Network via Predictive World Modeling
arXiv: 2605.24892NeurIPS 2025（投稿中）2026-05-26
WMPO: World Model-based Policy Optimization for Vision-Language-Action Models
arXiv: 2511.09515ICLR 20262026-05-25
World Action Models: The Next Frontier in Embodied AI
arXiv: 2605.12090arXiv 综述 20262026-05-25
World Models and World Action Models: A Survey Note
PaperNotes 综述 20262026-05-22
DreamZero: World Action Models are Zero-shot Policies
arXiv: 2602.15922NVIDIA Tech Report 20262026-05-21
Fast-WAM: Do World Action Models Need Test-time Future Imagination?
arXiv: 2603.16666NeurIPS 2026（投稿中）2026-05-21
GigaWorld-Policy: An Efficient Action-Centered World-Action Model
arXiv: 2603.17240投稿中 (under review)2026-05-21
Unified 4D World Action Modeling from Video Priors with Asynchronous Denoising
arXiv: 2604.26694NeurIPS 2026（投稿中）2026-05-21
🏷️ 强化学习 10 篇
↑ 回到标签列表
机器人如何从经验中变得更好？｜具身 RSI 与 TTT 研究博客
研究综述 20262026-10-11
WAM-RL: World-Action Model Reinforcement Learning with Reconstruction Rewards and Online Video SFT
arXiv: 2606.17906arXiv 20262026-06-24
LVR: Latent Visual Reasoning
arXiv: 2509.24251ICLR 20262026-06-11
SLVR: Semantic-Enriched Latent Visual Reasoning
arXiv: 2605.19342ICML 20262026-06-11
DiWA: Diffusion Policy Adaptation with World Models
arXiv: 2508.03645CoRL 20252026-05-25
WMPO: World Model-based Policy Optimization for Vision-Language-Action Models
arXiv: 2511.09515ICLR 20262026-05-25
World4RL: Diffusion World Models for Policy Refinement with Reinforcement Learning for Robotic Manipulation
arXiv: 2509.19080投稿 IEEE Trans 期刊2026-05-25
SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning
arXiv: 2509.09674ICLR 20262026-05-22
LaST-R1 论文分享简明笔记
arXiv: 2604.28192ICML 2026 Spotlight2026-05-08
LaST-R1 论文与代码笔记
arXiv: 2604.28192ICML 2026 Spotlight2026-05-08
🏷️ 泛化 7 篇
↑ 回到标签列表
机器人如何从经验中变得更好？｜具身 RSI 与 TTT 研究博客
研究综述 20262026-10-11
Qwen-RobotManip Technical Report: Alignment Unlocks Scale for Robotic Manipulation Foundation Models
arXiv: 2606.17846Qwen Technical Report 20262026-06-18
Geometric Action Model for Robot Policy Learning
arXiv: 2606.17046CoRL 2026（投稿，under review）2026-06-17
Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments
arXiv: 2605.30280千问VLA技术报告2026-05-29
World Models and World Action Models: A Survey Note
PaperNotes 综述 20262026-05-22
DreamZero: World Action Models are Zero-shot Policies
arXiv: 2602.15922NVIDIA Tech Report 20262026-05-21
HarmoWAM: Harmonizing Generalizable and Precise Manipulation via Adaptive World Action Models
arXiv: 2605.10942NeurIPS 2026（投稿中）2026-05-21
🏷️ 3D重建 6 篇
↑ 回到标签列表
Geometric Action Model for Robot Policy Learning
arXiv: 2606.17046CoRL 2026（投稿，under review）2026-06-17
Understanding the Impact of Geometric Foundation Models on Vision-Language-Action Models
arXiv: 2605.24642ECCV 2026（投稿中）2026-05-26
OASIS: Observation-Action Space Alignment via SE(3) Trajectory Prediction for Robotic Manipulation
arXiv: 2605.25829NeurIPS 2026（投稿中）2026-05-26
Unified 4D World Action Modeling from Video Priors with Asynchronous Denoising
arXiv: 2604.26694NeurIPS 2026（投稿中）2026-05-21
VGGT-Ω: Scaling Feed-Forward Reconstruction Models
arXiv: 2605.15195CVPR 2026 Oral2026-05-20
ConsisVLA-4D 论文阅读笔记
arXiv: 2605.05126CVPR 20262026-05-10
🏷️ 多模态理解 6 篇
↑ 回到标签列表
Qwen-RobotNav Technical Report: A Scalable Navigation Model Designed for an Agentic Navigation System
arXiv: 2606.18112Qwen Technical Report 20262026-06-18
CapImagine: Imagination Helps Visual Reasoning, But Not Yet in Latent Space
arXiv: 2602.22766ICML 20262026-06-11
Heima: Efficient Reasoning with Hidden Thinking
arXiv: 2501.19201ICML 20262026-06-11
LatentOmni: Rethinking Omni-Modal Understanding via Unified Audio-Visual Latent Reasoning
arXiv: 2605.22012arXiv 2026（投稿中）2026-06-11
Latent Reasoning for Multimodal Models: A Survey Note
PaperNotes 综述 20262026-06-11
VaLR: Vision-aligned Latent Reasoning for Multi-modal Large Language Model
arXiv: 2602.04476ICML 20262026-06-11
🏷️ 综述 5 篇
↑ 回到标签列表
机器人如何从经验中变得更好？｜具身 RSI 与 TTT 研究博客
研究综述 20262026-10-11
World Action Models: A Survey — 阅读笔记
arXiv: 2606.20781arXiv 20262026-06-25
Latent Reasoning for Multimodal Models: A Survey Note
PaperNotes 综述 20262026-06-11
World Action Models: The Next Frontier in Embodied AI
arXiv: 2605.12090arXiv 综述 20262026-05-25
World Models and World Action Models: A Survey Note
PaperNotes 综述 20262026-05-22
🏷️ 自动驾驶 5 篇
↑ 回到标签列表
Qwen-RobotNav Technical Report: A Scalable Navigation Model Designed for an Agentic Navigation System
arXiv: 2606.18112Qwen Technical Report 20262026-06-18
Qwen-RobotWorld Technical Report: Unifying Embodied World Modeling through Language-Conditioned Video Generation
arXiv: 2606.17030Qwen Technical Report 20262026-06-18
X-Foresight: A Joint Vision-Action Causal Forecasting Network via Predictive World Modeling
arXiv: 2605.24892NeurIPS 2025（投稿中）2026-05-26
MindVLA-U1: VLA Beats VA with Unified Streaming Architecture for Autonomous Driving
arXiv: 2605.12624NeurIPS 2026（投稿中）2026-05-14
OneVL: One-Step Latent Reasoning and Planning with Vision-Language Explanations
arXiv: 2604.18486小米汽车技术报告2026-05-13
🏷️ 视觉推理 5 篇
↑ 回到标签列表
CapImagine: Imagination Helps Visual Reasoning, But Not Yet in Latent Space
arXiv: 2602.22766ICML 20262026-06-11
Latent Reasoning for Multimodal Models: A Survey Note
PaperNotes 综述 20262026-06-11
LVR: Latent Visual Reasoning
arXiv: 2509.24251ICLR 20262026-06-11
SLVR: Semantic-Enriched Latent Visual Reasoning
arXiv: 2605.19342ICML 20262026-06-11
VaLR: Vision-aligned Latent Reasoning for Multi-modal Large Language Model
arXiv: 2602.04476ICML 20262026-06-11
🏷️ 链式思考 4 篇
↑ 回到标签列表
Heima: Efficient Reasoning with Hidden Thinking
arXiv: 2501.19201ICML 20262026-06-11
ImgCoT: Compressing Long Chain of Thought into Compact Visual Tokens for Efficient Reasoning of Large Language Model
arXiv: 2601.22730ICML 20262026-06-11
LatentOmni: Rethinking Omni-Modal Understanding via Unified Audio-Visual Latent Reasoning
arXiv: 2605.22012arXiv 2026（投稿中）2026-06-11
Latent Reasoning for Multimodal Models: A Survey Note
PaperNotes 综述 20262026-06-11
🏷️ 扩散模型 3 篇
↑ 回到标签列表
ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?
arXiv: 2606.19531投稿（under review）2026-06-25
ChordEdit: One-Step Low-Energy Transport for Image Editing
arXiv: 2602.19083CVPR 2026 Oral2026-06-08
Large Language Diffusion Models
arXiv: 2502.09992NeurIPS 20252026-05-29
🏷️ 语言模型 3 篇
↑ 回到标签列表
机器人如何从经验中变得更好？｜具身 RSI 与 TTT 研究博客
研究综述 20262026-10-11
ImgCoT: Compressing Long Chain of Thought into Compact Visual Tokens for Efficient Reasoning of Large Language Model
arXiv: 2601.22730ICML 20262026-06-11
Large Language Diffusion Models
arXiv: 2502.09992NeurIPS 20252026-05-29
🏷️ 知识蒸馏 2 篇
↑ 回到标签列表
Heima: Efficient Reasoning with Hidden Thinking
arXiv: 2501.19201ICML 20262026-06-11
CKT-WAM: Parameter-Efficient Context Knowledge Transfer Between World Action Models
arXiv: 2605.06247NeurIPS 2026（投稿中）2026-05-21
🏷️ 音视频推理 2 篇
↑ 回到标签列表
LatentOmni: Rethinking Omni-Modal Understanding via Unified Audio-Visual Latent Reasoning
arXiv: 2605.22012arXiv 2026（投稿中）2026-06-11
Latent Reasoning for Multimodal Models: A Survey Note
PaperNotes 综述 20262026-06-11
🏷️ 图像编辑 1 篇
↑ 回到标签列表
ChordEdit: One-Step Low-Energy Transport for Image Editing
arXiv: 2602.19083CVPR 2026 Oral2026-06-08
🏷️ 导航 1 篇
↑ 回到标签列表
Qwen-RobotNav Technical Report: A Scalable Navigation Model Designed for an Agentic Navigation System
arXiv: 2606.18112Qwen Technical Report 20262026-06-18
🤖 机器人 1 篇
↑ 回到标签列表
WAM-RL: World-Action Model Reinforcement Learning with Reconstruction Rewards and Online Video SFT
arXiv: 2606.17906arXiv 20262026-06-24