arXivDaily arXiv每日学术速递 周一至周五更新

视觉与机器人

VLA / 视觉-语言-动作模型

视觉-语言-动作模型、机器人基础模型和语言条件机器人控制。

2026-06-24 至 2026-06-24 共收录 14 信号源:cs.RO, cs.CV, cs.AI, cs.LG

1. VLA模型 12 篇

2606.24472 2026-06-24 cs.RO cs.AI 新提交 94%

G$^3$VLA: Geometric inductive bias for Vision-Language-Action Models

G$^3$VLA:视觉-语言-动作模型的几何归纳偏置

Yue Peng, Yongzhe Zhao, Artur Habuda, Khuyen Pham, Yanheng Zhu, Tran Nguyen Le, Fares Abu-Dakka, Li Guo

机构 * New York University Shanghai(上海纽约大学) Technical University of Denmark(丹麦技术大学) MBZUAI - Mohamed bin Zayed University of Artificial Intelligence(穆罕默德·本·扎耶德人工智能大学) New York University Abu Dhabi(纽约大学阿布扎比分校)

专题命中 VLA模型 :VLA(title,title_cn);vision-language-action(title,abstract);action model(title);分类 cs.RO、cs.AI

AI总结 提出G$^3$VLA模块,通过射线嵌入、投影位置编码和跨视图融合为VLA模型注入相机几何先验,无需深度传感器,在多个基准和真实机器人上提升空间敏感任务性能。

Comments Submitted to CoRL 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.24051 2026-06-24 cs.CV 新提交 94%

DriveStack-VLA: Render-Teacher Alignment for BEV-Based DeepStack Vision-Language-Action Model

DriveStack-VLA: 基于BEV的DeepStack视觉-语言-动作模型的渲染-教师对齐

Jingke Wang, Zhenru Zhao, Shuangming Lei, Hao Su, Yuehao Huang, Yijia Xie, Kai Tang, Guanglin Xu, AiXue Ye, Yukai Ma, Yong Liu

机构 * Zhejiang University(浙江大学) The 2012 Labs, Huawei(华为2012实验室)

专题命中 VLA模型 :VLA(title,title_cn);vision-language-action(title,abstract);action model(title);分类 cs.CV

AI总结 提出DriveStack-VLA框架,通过DeepStack连接注入BEV表示并采用渲染-教师对齐增强空间感知,引入自批评模块优化轨迹选择,在多个驾驶基准上取得领先性能。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.24884 2026-06-24 cs.RO cs.AI cs.LG 新提交 89%

InSight: Self-Guided Skill Acquisition via Steerable VLAs

InSight: 通过可引导的VLA实现自主技能获取

Maggie Wang, Lars Osterberg, Stephen Tian, Ola Shorinwa, Jiajun Wu, Mac Schwager

机构 * Stanford University(斯坦福大学) Princeton University(普林斯顿大学)

专题命中 VLA模型 :VLA(title_cn,summary_cn);vision-language-action(abstract);分类 cs.RO、cs.AI、cs.LG

AI总结 提出InSight框架,通过将VLA模型在基本动作层面变得可引导,实现自主技能获取,包括自动分割演示为基本动作和VLM引导的数据飞轮,无需人类演示即可学习新技能。

Comments Project website: https://insight-vla.github.io

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.24448 2026-06-24 cs.RO 新提交 89%

Supervise What Survives: Geometry-Guided VLA Adaptation from Synthetic Robot Videos

监督幸存信息:基于几何引导的合成机器人视频VLA适配

Danze Chen, Yanzhe Chen, Qiming Huang, Zhijun Cao, Chen Gao, Mike Zheng Shou

机构 * Show Lab, National University of Singapore(新加坡国立大学 Show Lab)

专题命中 VLA模型 :VLA(title,title_cn);vision-language-action(abstract);分类 cs.RO

AI总结 提出GRA方法,从合成视频中提取几何信息(2D末端执行器路径点)监督视觉表征,仅用真实演示训练动作头,在真实机器人任务中优于伪动作基线。

Comments 14 pages, 5 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.23938 2026-06-24 cs.AI cs.CL 新提交 86%

Neuro-Symbolic Drive: Rule-Grounded Faithful Reasoning for Driving VLAs

神经符号驱动:基于规则忠实推理的驾驶VLA

Xiangbo Gao, Xiukun Huang, Boyu Lu, Junge Zhang, Mengjie Mao, Jiachen Li, Wei Xiong, Zhengzhong Tu

机构 * Texas A&M University(德克萨斯农工大学) Carnegie Mellon University(卡内基梅隆大学) University of Maryland(马里兰大学) University of California, Riverside(加利福尼亚大学河滨分校) University of Pittsburgh(匹兹堡大学)

专题命中 VLA模型 :VLA(title_cn,summary_cn);分类 cs.AI

AI总结 提出神经符号驱动框架,利用规则规划器的决策轨迹监督驾驶VLA,实现推理与运动生成的因果耦合,显著降低轨迹误差和碰撞率。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.17046 2026-06-24 cs.RO cs.CV cs.LG 新提交 85%

Geometric Action Model for Robot Policy Learning

几何动作模型用于机器人策略学习

Jisang Han, Seonghu Jeon, Jaewoo Jung, René Zurbrügg, Honggyu An, Tifanny Portela, Marco Hutter, Marc Pollefeys, Seungryong Kim, Sunghwan Hong

机构 * KAIST AI(韩国科学技术院人工智能学院) ETH Zurich(苏黎世联邦理工学院) ETH AI Center(苏黎世联邦理工学院人工智能中心)

专题命中 VLA模型 :action model(title,abstract);vision-language-action(abstract);分类 cs.RO、cs.CV、cs.LG

AI总结 提出几何动作模型(GAM),通过重用预训练几何基础模型(GFM)作为共享骨干,实现语言条件下的操作策略,在仿真和真实机器人任务中优于现有方法。

Comments Project page: https://cvlab-kaist.github.io/Geometric-Action-Model/

详情

展开后加载摘要…

URL PDF HTML 收藏
2503.05226 2026-06-24 cs.RO cs.AI 版本更新 82%

Reward-Centered ReST-MCTS: A Robust Decision-Making Framework for Robotic Manipulation in High Uncertainty Environments

以奖励为中心的ReST-MCTS:高不确定性环境下机器人操作的鲁棒决策框架

Xibai Wang

机构 * Xibai Wang(王西拜)

专题命中 VLA模型 :VLA(summary_cn,abstract);分类 cs.RO、cs.AI

AI总结 提出Reward-Centered ReST-MCTS框架,通过分解中间反馈为中心信号并引导搜索,解决高不确定性环境下MCTS因稀疏奖励和噪声导致的决策脆弱性问题,实验验证其在同骨干VLA上的鲁棒性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.22136 2026-06-24 cs.RO 新提交 81%

Wh0: Generative World Models as Scalable Sources of Egocentric Human Hand Manipulation Data

Wh0: 生成式世界模型作为自我中心人类手部操作数据的可扩展来源

Yangtao Chen, Zixuan Chen, Peiyang Wang, Yong-Lu Li, Jing Huo, Jieqi Shi, Yang Gao

机构 * Shanghai Innovation Institute(上海创新研究院) Nanjing University(南京大学) Shanghai Jiaotong University(上海交通大学)

专题命中 VLA模型 :VLA(summary_cn,abstract);分类 cs.RO

AI总结 提出Wh0框架,利用生成式视频世界模型产生自我中心人-物交互数据集WM-H,通过手部运动重建和视觉编辑转化为机器人可训练监督,提升预训练灵巧VLA模型在真实任务上的零样本成功率。

Comments Under review. The first three authors contributed equally to this work

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.23754 2026-06-24 cs.RO cs.LG 新提交 73%

Verifiable Foundation Models for Robot Safety

机器人安全的可验证基础模型

Davide Corsi, Kyungmin Kim, Roy Fox

机构 * University of California, Irvine(加州大学尔湾分校)

专题命中 VLA模型 :vision-language-action(abstract);action model(abstract);分类 cs.RO、cs.LG

AI总结 提出FEARL框架,将策略分解为大控制器和小安全模块,使形式化验证仅应用于安全模块,从而在保持控制器表达能力的同时实现可验证的机器人安全控制。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.23773 2026-06-24 astro-ph.HE gr-qc 新提交 71%

Modern tidal interaction models for rapid binary population synthesis: II. Binary black hole formation, mergers, and spins

快速双星族合成中的现代潮汐相互作用模型:II. 双黑洞形成、并合与自旋

Veome Kapil, Ilya Mandel, Jeff Riley, Evgeni Grishin, Jim Fuller, Emanuele Berti

专题命中 VLA模型 :action model(title)

AI总结 本文在快速双星族合成代码COMPAS中实现新的自洽潮汐耗散模型,预测双黑洞并合率及有效自旋分布,发现第二代黑洞自旋显著依赖于潮汐耗散效率和质量传输历史,高有效自旋系统优先在高红移并合。

Comments 23 pages, 16 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.24184 2026-06-24 cs.LG 新提交 57%

Project Ariadne: Prompt-Conditioned Route Generation for Synthesis Planning

Project Ariadne: 用于合成规划的提示条件路线生成

Anton Morgunov, Victor S. Batista

机构 * Yale University(耶鲁大学)

专题命中 VLA模型 :action model(abstract);分类 cs.LG

AI总结 提出解码器专用路线生成器Ariadne,通过提示-补全序列统一表示目标、约束和路线,在RetroCast/PaRoutes基准上,深度约束和所需叶子约束的Solv-0分别提升13.7和31.2点,推理时间远低于DESP。

Comments Code is available at https://github.com/ischemist/project-ariadne

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.18375 2026-06-24 cs.RO 新提交 57%

PAIWorld: A 3D-Consistent World Foundation Model for Robotic Manipulation

PAIWorld: 用于机器人操作的三维一致世界基础模型

Yuhang Huang, Xuan Lv, Junyan Xu, Zhiyuan Yu, Jiazhao Zhang, Ruizhen Hu, Wancheng Feng, Shilong Zou, Hewen Xiao, Ziqiao Zhou, Kaiyun Huang, Zhiyu Peng, Juzhan Xu, Hang Zhao, Chenyang Zhu, Renjiao Yi, Yifei Huang, Douhui Wu, Yan Zhang, Kexu Cheng, Chunhe Song, Yunzhi Xue, Xiuhong Zhang, Leitao Guo, Yunji Chen, Bin Wu, Haibin Yu, Kai Xu

机构 * Institute of AI for Industries, Chinese Academy of Sciences(中国科学院人工智能产业研究院)

专题命中 VLA模型 :action model(abstract);分类 cs.RO

AI总结 提出PAIWorld框架,通过几何感知交叉注意力、几何旋转位置编码和潜在3D-REPA蒸馏,解决多视图世界模型的3D不一致问题,在机器人操作基准上取得领先性能。

详情

展开后加载摘要…

URL PDF HTML 收藏

2. 语言条件控制 1 篇

2312.10807 2026-06-24 cs.RO 85%

Bridging Language and Action: A Survey of Language-Conditioned Robot Manipulation

连接语言与行动:语言引导的机器人操作综述

Xiangtong Yao, Hongkuan Zhou, Oier Mees, Yuan Meng, Ted Xiao, Yonatan Bisk, Jean Oh, Edward Johns, Mohit Shridhar, Dhruv Shah, Jesse Thomason, Kai Huang, Joyce Chai, Zhenshan Bing, Alois Knoll

机构 * Technical University of Munich(慕尼黑技术大学) Corporate Research, Robert Bosch GmbH(罗伯特·博世集团企业研究部) University of California Berkeley(加州大学伯克利分校) Microsoft(微软) Google DeepMind(谷歌DeepMind) Carnegie Mellon University(卡内基梅隆大学) Imperial College London(伦敦帝国理工学院) Princeton University(普林斯顿大学) University of Southern California(南加州大学) Sun Yat-sen University(中山大学) University of Michigan(密歇根大学) Institute for Artificial Intelligence, University of Stuttgart(斯图加特大学人工智能研究所) The State Key Laboratory for Novel Software Technology, Nanjing University(南京大学新型软件技术国家重点实验室)

专题命中 语言条件控制 :language-conditioned robot(title,abstract);vision-language-action(abstract);action model(abstract);分类 cs.RO

AI总结 本文综述了语言引导的机器人操作领域,探讨了语言如何与机器人系统整合,分析了现有方法的分类及最新进展,指出关键争议和未来研究方向。

详情

展开后加载摘要…

URL PDF HTML 收藏

3. 数据集与评测 1 篇

2606.02800 2026-06-24 cs.CV cs.AI cs.LG cs.MM cs.RO 版本更新 70%

Cosmos 3: Omnimodal World Models for Physical AI

Cosmos 3:面向物理AI的全模态世界模型

NVIDIA, :, Aditi, Niket Agarwal, Arslan Ali, Jon Allen, Martin Antolini, Adeline Aubame, Alisson Azzolini, Junjie Bai, Maciej Bala, Yogesh Balaji, Josh Bapst, Aarti Basant, Mukesh Beladiya, Mohammad Qazim Bhat, Zaid Pervaiz Bhat, Dan Blick, Vanni Brighella, Han Cai, Tiffany Cai, Eric Cameracci, Jiaxin Cao, Yulong Cao, Mark Carlson, Carlos Casanova, Ting-Yun Chang, Yan Chang, Yu-Wei Chao, Prithvijit Chattopadhyay, Roshan Chaudhari, Chieh-Yun Chen, Junyu Chen, Ke Chen, Qizhi Chen, Wenkai Chen, Xiaotong Chen, Yu Chen, An-Chieh Cheng, Click Cheng, Xiu Chia, Jeana Choi, Chaeyeon Chung, Wenyan Cong, Yin Cui, Magdalena Dadela, Nalin Dadhich, Wenliang Dai, Joyjit Daw, Alperen Degirmenci, Rodrigo Vieira Del Monte, Robert Denomme, Sameer Dharur, Marco Di Lucca, Ke Ding, Wenhao Ding, Yifan Ding, Yuzhu Dong, Nicole Drumheller, Yilun Du, Aigul Dzhumamuratova, Aleksandr Efitorov, Hamid Eghbalzadeh, Naomi Eigbe, Imad El Hanafi, Hassan Eslami, Benedikt Falk, Jiaojiao Fan, Jim Fan, Amol Fasale, Sergiy Fefilatyev, Liang Feng, Francesco Ferroni, Sanja Fidler, Xiao Fu, Vikram Fugro, Prashant Gaikwad, TJ Galda, Katelyn Gao, Yihuai Gao, Wenhang Ge, Sreyan Ghosh, Arushi Goel, Vivek Goel, Akash Gokul, Rama Govindaraju, Jinwei Gu, Miguel Guerrero, Elfie Guo, Aryaman Gupta, Siddharth Gururani, Hugo Hadfield, Song Han, Ankur Handa, Zekun Hao, Mohammad Harrim, Ali Hassani, Nathan Hayes-Roth, Yufan He, Chris Helvig, Cyrus Hogg, Madison Huang, Michael Huang, Sophia Huang, Yufan Huang, Jacob Huffman, DeLesley Hutchins, Suneel Indupuru, Boris Ivanovic, Arihant Jain, Joel Jang, Ryan Ji, Yanan Jian, Dongfu Jiang, Jingyi Jin, Atharva Joshi, Nikhilesh Joshi, Pranjali Joshi, Andy Ju, Jaehun Jung, Weiwei Kang, Scott Kassekert, Jan Kautz, Ashna Khetan, Julia Kiczka, Slawek Kierat, Gwanghyun Kim, Kuno Kim, Sunny Kim, Kezhi Kong, Xin Kong, Zhifeng Kong, Tomasz Kornuta, Egor Krivov, Hui Kuang, Saurav Kumar, Chia-Wen Kuo, George Kurian, Wojciech Kutak, JF Lafleche, Himangshu Lahkar, Omar Laymoun, Jayjun Lee, Sanggil Lee, Gabriele Leone, Boyi Li, Freya Li, Jiajun Li, Jinfeng Li, Ling Li, Pengcheng Li, Shangru Li, Tingle Li, Xiaolong Li, Xuan Li, Zhaoshuo Li, Zhiqi Li, Hao Liang, Maosheng Liao, Chen-Hsuan Lin, Tsung-Yi Lin, Ming-Yu Liu, Sifei Liu, Zihan Liu, Hai Loc Lu, Xiangyu Lu, Alice Luo, Ruipu Luo, Wenjie Luo, Jiangran Lyu, Martin Ding Ma, Nic Ma, Qianli Ma, Dawid Majchrowski, Louis Marcoux, Miguel Martin, Qing Miao, Ashkan Mirzaei, Shreyas Misra, Kaichun Mo, Durra Mohsin, Hyejin Moon, Pawel Morkisz, Saeid Motiian, Kirill Motkov, Seungjun Nah, Yashraj Narang, Deepak Narayanan, Thabang Ngazimbi, Julian Ouyang, Shubham Pachori, David Page, Yatian Pang, Sehwi Park, Mahesh Patekar, Mostofa Patwary, Marco Pavone, Trung Pham, Wei Ping, Soha Pouya, Shrimai Prabhumoye, Varun Praveen, Delin Qu, Hesam Rabeti, Morteza Ramezanali, Marilyn Reeb, Xuanchi Ren, Kristen Rumley, Wojciech Rymer, Jun Saito, Yeongho Seol, John Shao, Piyush Shekdar, Tianwei Shen, Humphrey Shi, Min Shi, Stella Shi, Kevin Shih, Mohammad Shoeybi, Mateusz Sieniawski, Shuran Song, Alexander Sotelo, Amir Sotoodeh, Sunil Srinivasa, Vignesh Srinivasakumar, Bartosz Stefaniak, Rahul Heinrich Steiger, Shangkun Sun, Jiaxiang Tang, Shitao Tang, Yangyang Tang, Yue Tang, Tolou Tavakkoli, Kayley Ting, Krzysztof Tomala, Wei-Cheng Tseng, Jibin Varghese, Sergei Vasilev, Thomas Volk, Raju Wagwani, Roger Waleffe, Andrew Z. Wang, Boxiang Wang, Haoxiang Wang, Qiao Wang, Shihao Wang, Shijie Wang, Ting-Chun Wang, Yan Wang, Yu Wang, Rohit Watve, David Wehr, Fangyin Wei, Xinshuo Weng, Jay Zhangjie Wu, Kedi Wu, Hongchi Xia, Summer Xiao, Tianjun Xiao, Kevin Xie, Daguang Xu, Jiashu Xu, Mengyao Xu, Ruqing Xu, Xingqian Xu, Yao Xu, Dinghao Yang, Dong Yang, Hans Yang, Xiaodong Yang, Xuning Yang, Yichu Yang, Yurong You, Zhiding Yu, Hao Yuan, Simon Yuen, Xiaohui Zeng, Pengcuo Zeren, Cindy Zha, Haotian Zhang, Jenny Zhang, Jing Zhang, Liangkai Zhang, Paris Zhang, Shun Zhang, Xuanmeng Zhang, Zhizheng Zhang, Ann Zhao, Yilin Zhao, Yuliya Zhautouskaya, Charles Zhou, Fengzhe Zhou, Shilin Zhu, Yuke Zhu, Dima Zhylko, Artur Zolkowski

机构 * NVIDIA

专题命中 数据集与评测 :action model(abstract);分类 cs.RO、cs.CV、cs.AI

AI总结 提出基于统一混合Transformer架构的全模态世界模型Cosmos 3,联合处理语言、图像、视频、音频和动作序列,在理解和生成任务上达到新最优,为具身智能体提供可扩展的通用骨干。

详情

展开后加载摘要…

URL PDF HTML 收藏