A VLM backbone with a flow-matching action expert, trained across many robots, doing long dexterous tasks like laundry folding.VLM 主干加流匹配动作专家，跨多种机器人训练，能完成叠衣服等长时程灵巧任务。
@misc{black2024vision,
title = {{$π_0$: A Vision-Language-Action Flow Model for General Robot Control}},
author = {Kevin Black and Noah Brown and Danny Driess and Adnan Esmail and Michael Equi and Chelsea Finn and Niccolo Fusai and Lachy Groom and Karol Hausman and Brian Ichter and Szymon Jakubczak and Tim Jones and Liyiming Ke and Sergey Levine and Adrian Li-Bell and Mohith Mothukuri and Suraj Nair and Karl Pertsch and Lucy Xiaoyang Shi and James Tanner and Quan Vuong and Anna Walling and Haohuan Wang and Ury Zhilinsky},
year = {2024},
eprint = {2410.24164},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2410.24164},
}
1M+ trajectories from 22 robot types pooled by 30+ labs; RT-X models trained on it beat single-robot baselines.30 多个实验室汇集 22 种机器人的 100 多万条轨迹；在其上训练的 RT-X 优于单机器人基线。
@misc{collaboration2023open,
title = {{Open X-Embodiment: Robotic Learning Datasets and RT-X Models}},
author = {Open X-Embodiment Collaboration and Abby O'Neill and Abdul Rehman and Abhinav Gupta and Abhiram Maddukuri and Abhishek Gupta and Abhishek Padalkar and Abraham Lee and Acorn Pooley and Agrim Gupta and Ajay Mandlekar and Ajinkya Jain and Albert Tung and Alex Bewley and Alex Herzog and Alex Irpan and Alexander Khazatsky and Anant Rai and Anchit Gupta and Andrew Wang and Andrey Kolobov and Anikait Singh and Animesh Garg and Aniruddha Kembhavi and Annie Xie and Anthony Brohan and Antonin Raffin and Archit Sharma and Arefeh Yavary and Arhan Jain and Ashwin Balakrishna and Ayzaan Wahid and Ben Burgess-Limerick and Beomjoon Kim and Bernhard Schölkopf and Blake Wulfe and Brian Ichter and Cewu Lu and Charles Xu and Charlotte Le and Chelsea Finn and Chen Wang and Chenfeng Xu and Cheng Chi and Chenguang Huang and Christine Chan and Christopher Agia and Chuer Pan and Chuyuan Fu and Coline Devin and others},
year = {2023},
eprint = {2310.08864},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2310.08864},
}
Blinded pairwise trials at 124 sites compare VLAs and frontier agents. Rankings barely agree across robots (W = 0.24), and open rankings poorly predict results after fine-tuning; no policy finishes a G1 task.在 124 个站点做盲测两两对比，横跨四种机器人评测 VLA 与前沿智能体。不同机器人的排名几乎不一致（Kendall W = 0.24），开箱排名也难以预测微调后的表现；G1 人形上没有策略完成任何任务。
@misc{nguyen2026robocolosseum,
title = {{RoboColosseum: Evaluating Physical Intelligence at Scale In-The-Wild}},
author = {Tan Sang Nguyen and Yi Ru Wang and Shailesh Shailesh and Gang Yang and Jiamo Hu and Jianuo Jiang and Haoquan Fang and Zhongzheng Ren and Zhongyi Luo and Ian Chong and Aaron Tung and Santiago Pravisani and Minghui Lin and Yao Mo and Ishneet Singh and Benjamin Tee and Harold Soh and Siddhartha Srinivasa and Dieter Fox and David Hsu and Haoming Song and Maxime Ellerbach and Thomas Wolf and Pepijn Kooijmans and Ruijie He and Michael Cho and Jiafei Duan},
year = {2026},
eprint = {2611.02497},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.02497},
}
LingBot-VLA: an Apache-2.0 foundation model trained on about 20,000 hours of real data from nine dual-arm setups, tested on four platforms with 100 tasks each; code, base model and benchmark data are released.LingBot-VLA：Apache-2.0 开源的基础模型，用九种双臂构型约 2 万小时真实数据训练，并在四个平台各 100 项任务上评测；代码、基座模型和评测数据一并开放。
@misc{wu2026pragmatic,
title = {{A Pragmatic VLA Foundation Model}},
author = {Wei Wu and Fan Lu and Yunnan Wang and Shuai Yang and Shi Liu and Fangjing Wang and Qian Zhu and He Sun and Yong Wang and Shuailei Ma and Yiyu Ren and Kejia Zhang and Hui Yu and Jingmei Zhao and Shuai Zhou and Zhenqi Qiu and Houlong Xiong and Ziyu Wang and Zechen Wang and Ran Cheng and Yong-Lu Li and Yongtao Huang and Xing Zhu and Yujun Shen and Kecheng Zheng},
year = {2026},
eprint = {2601.18692},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2601.18692},
}
NEO-VLA drops the visual encoder: images and language evolve in one backbone and a flow-matching action expert reads several depths. Layer probing shows action signals often peak before the final visual layer.NEO-VLA 去掉独立视觉编码器：图像与语言在同一主干中共同演化，流匹配动作专家从多个深度读取特征。逐层探测显示，对动作关键的信号往往在最终视觉层之前达到峰值。
@misc{fan2026from,
title = {{From Pixels to Actions - Towards Native Vision-Language-Action Policies}},
author = {Weichen Fan and Haiwen Diao and Penghao Wu and Siyuan Hu and Ziqi Huang and Tianqi Liu and Mingze Sun and Ziyi Wang and Zhongang Cai and Hanming Deng and Quan Wang and Lei Yang and Lewei Lu and Fangzhou Hong and Zhaoxi Chen and Dahua Lin and Ziwei Liu},
year = {2026},
eprint = {2611.06918},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.06918},
}
Opens a 50,000-hour annotated egocentric video corpus with pipelines and checkpoints. Reports a log-linear law between pretraining hours and robot success, and 84.6% on RoboTwin 2.0 Clean2Random.开源 5 万小时带标注的第一视角视频语料及流程和权重。报告预训练时长与机器人成功率之间的对数线性规律，后训练后在 RoboTwin 2.0 Clean2Random 上达到 84.6%。
@misc{ma2026open,
title = {{Open-HumanWAM: An Open Recipe for Scaling Human and Robot Experience for Embodied Foundation Models}},
author = {Juncheng Ma and Jiaxin Li and Jiankai Tu and Bo Liang and Guodong Lv and Yufan Deng and Junzhe Tang and Liangzhu Yang and Chen Bian and Zicong Cao and Baohao Zhu and Jianxin Bi and Tengfei Yan and Yinyuan Zhang and Kewei Zhang and Xinglong Chen and Kaiqi Chen and Duomin Wang and Xiaotian Tang and Liang Tan and Li Li and Yuan Xu and Haosen Yang and Yuqi Wang and Jingfei Xia and Songyan Huang and Zhiyang Dou and Zhen Dong and Enze Xie and Tat-Seng Chua and Daquan Zhou},
year = {2026},
eprint = {2611.06817},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.06817},
}
Learned from 16,151 hours of handheld gripper demos with no robot data; reports 78% average success over ten tasks versus 27% for the best other VLA, and releases the dataset, weights and post-training recipe.从 16,151 小时手持夹爪演示中学习，全程不采机器人数据；报告十项任务平均成功率 78%，其他 VLA 最好为 27%，并开放数据集、权重和后训练方案。
@misc{wu2026airoa,
title = {{AIRoA-0: Long-Horizon, Precise, Cross-Embodiment Manipulation from Handheld Data}},
author = {Yueh-Hua Wu and Kei Ota and Hanlong Li and Hayata Yuasa and Jason Wang and Kengo Ikeuchi and Kensuke Chiyoma and Makoto Sugiura and Masafumi Takahashi and Masatoshi Tateno and Naoaki Kanazawa and Nicola Scianca and Petr Khrapchenkov and Satoshi Otsubo and Takehiko Ohkawa and Takuya Okubo and Tatsuya Matsushima and Yuki Noguchi and Yuma Shin and Zixuan Zhong},
year = {2026},
eprint = {2611.03815},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.03815},
}
An open research stack that splits world-action models into composable modules, then runs controlled studies on what to inherit from video priors and how world and action learning interact.开放研究栈，把世界动作模型拆成可组合模块，并通过对照实验研究该从视频先验继承什么、世界学习与动作学习如何相互作用。
@misc{wang2026openwam,
title = {{OpenWAM: An Open, Modular Exploration Towards Systematic World-Action Model Pretraining}},
author = {Yuran Wang and Siqiao Huang and Mingleyang Li and Chenhao Zhang and Jiaqi Liang and Weiyang Jin and Yue Chen and Xuemin Chi and Donghao Zhou and Qize Yu and Yu-Kai Wang and Yuhan Rui and Shenzhe Yao and Zhen Yuan and Zhenhao Shen and Kefei Zhu and Zijie Zhu and Ning Gao and Xiaowei Chi and Guanqi He and Shanghang Zhang and Hao Dong and Lin Shao and Hang Zhao},
year = {2026},
eprint = {2609.07398},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.07398},
}
RECAP trains a VLA with RL via advantage conditioning on demos, on-robot rollouts and teleop corrections; reports it more than doubles throughput on hard tasks like laundry folding and espresso making.RECAP 通过优势条件化，用演示、机器人自主运行数据和遥操作纠正对 VLA 做强化学习；论文报告在叠衣服、做意式咖啡等难任务上吞吐量提升一倍以上、失败率约减半。
@misc{intelligence2025vla,
title = {{π*0.6: a VLA That Learns From Experience}},
author = {Physical Intelligence and Ali Amin and Raichelle Aniceto and Ashwin Balakrishna and Kevin Black and Ken Conley and Grace Connors and James Darpinian and Karan Dhabalia and Jared DiCarlo and Danny Driess and Michael Equi and Adnan Esmail and Yunhao Fang and Chelsea Finn and Catherine Glossop and Thomas Godden and Ivan Goryachev and Lachy Groom and Hunter Hancock and Karol Hausman and Gashon Hussein and Brian Ichter and Szymon Jakubczak and Rowan Jen and Tim Jones and Ben Katz and Liyiming Ke and Chandra Kuchi and Marinda Lamb and Devin LeBlanc and Sergey Levine and Adrian Li-Bell and Yao Lu and Vishnu Mano and Mohith Mothukuri and Suraj Nair and Karl Pertsch and Allen Z. Ren and Charvi Sharma and Lucy Xiaoyang Shi and Laura Smith and Jost Tobias Springenberg and Kyle Stachowicz and Will Stoeckle and Alex Swerdlow and James Tanner and Marcel Torne and Quan Vuong and Anna Walling and others},
year = {2025},
eprint = {2511.14759},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2511.14759},
}
A 7B open VLA built on Llama 2 + DINOv2/SigLIP; the de-facto open baseline for VLA research and LoRA fine-tuning.基于 Llama 2 与 DINOv2/SigLIP 的 7B 开源 VLA，是 VLA 研究与 LoRA 微调的事实开源基线。
@misc{kim2024openvla,
title = {{OpenVLA: An Open-Source Vision-Language-Action Model}},
author = {Moo Jin Kim and Karl Pertsch and Siddharth Karamcheti and Ted Xiao and Ashwin Balakrishna and Suraj Nair and Rafael Rafailov and Ethan Foster and Grace Lam and Pannag Sanketi and Quan Vuong and Thomas Kollar and Benjamin Burchfiel and Russ Tedrake and Dorsa Sadigh and Sergey Levine and Percy Liang and Chelsea Finn},
year = {2024},
eprint = {2406.09246},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2406.09246},
}
@misc{brohan2023rt,
title = {{RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control}},
author = {Anthony Brohan and Noah Brown and Justice Carbajal and Yevgen Chebotar and Xi Chen and Krzysztof Choromanski and Tianli Ding and Danny Driess and Avinava Dubey and Chelsea Finn and Pete Florence and Chuyuan Fu and Montse Gonzalez Arenas and Keerthana Gopalakrishnan and Kehang Han and Karol Hausman and Alexander Herzog and Jasmine Hsu and Brian Ichter and Alex Irpan and Nikhil Joshi and Ryan Julian and Dmitry Kalashnikov and Yuheng Kuang and Isabel Leal and Lisa Lee and Tsang-Wei Edward Lee and Sergey Levine and Yao Lu and Henryk Michalewski and Igor Mordatch and Karl Pertsch and Kanishka Rao and Krista Reymann and Michael Ryoo and Grecia Salazar and Pannag Sanketi and Pierre Sermanet and Jaspiar Singh and Anikait Singh and Radu Soricut and Huong Tran and Vincent Vanhoucke and Quan Vuong and Ayzaan Wahid and Stefan Welker and Paul Wohlhart and Jialin Wu and Fei Xia and Ted Xiao and others},
year = {2023},
eprint = {2307.15818},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2307.15818},
}
Drives robots with a frozen frontier LLM plus in-context 'knowledge packages' on 12 sim benchmarks + 8 real robots; reports beating prior bests on 9 of 12, but weak real-robot results (competence, not yet intelligence).用冻结的前沿大模型加上下文“知识包”直接控制机器人，覆盖 12 个仿真基准和 8 台真机；报告在 12 个基准中的 9 个超过此前最好结果，但真机表现较弱——结论是具备了具身“能力”，尚未达到具身“智能”。
@misc{sensenova2026has,
title = {{Has GPT-6-Astra Achieved Embodied Intelligence?}},
author = {SenseNova and Joint Research Team and Zhongang Cai and Ziang Cao and Tailai Chen and Han Chong Chua and Linjun Dai and Hanming Deng and Hokin Deng and Haiwen Diao and Chenjing Ding and Weichen Fan and Xiao Fu and Zerui Gong and Chenyang Gu and Zexin He and Zhi Hou and Bo Li and Qinyue Li and Quanzhou Li and Xiaowan Li and Zhichao Liao and Dahua Lin and Zilong Lin and Bo Liu and Chunxiao Liu and Pengfei Liu and Tianqi Liu and Ziwei Liu and Chen Change Loy and Lewei Lu and Jiapeng Luo and Yuexiao Ma and Liang Pan and Xingang Pan and Oscar Qian and Ziyi Shang and Yuyao Shao and Dawei Su and Dacheng Tao and Qingyi Tao and Quan Wang and Ruisi Wang and Xiaogang Wang and Ziyi Wang and Beichen Wen and Donghao Wu and Jiaxu Wu and Silei Wu and Xiaoming Wu and others},
year = {2026},
eprint = {2610.22471},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.22471},
}
Post-trains a VLA (GigaBrain-0.5, 10k+ h of robot data) with RL conditioned on a video world model's predictions (RAMP); reports ~30% over the RECAP baseline on laundry folding, box packing and espresso making.在 GigaBrain-0.5（预训练于 1 万多小时机器人操作数据）基础上，用以视频世界模型预测为条件的强化学习（RAMP）做后训练；论文报告在叠衣服、装箱和制作浓缩咖啡等难任务上比 RECAP 基线提升约 30%。
@misc{team2026gigabrain,
title = {{GigaBrain-0.5M*: a VLA That Learns From World Model-Based Reinforcement Learning}},
author = {GigaBrain Team and Boyuan Wang and Bohan Li and Chaojun Ni and Guan Huang and Guosheng Zhao and Hao Li and Jie Li and Jindi Lv and Jingyu Liu and Lv Feng and Mingming Yu and Peng Li and Qiuping Deng and Tianze Liu and Xinyu Zhou and Xinze Chen and Xiaofeng Wang and Yang Wang and Yifan Li and Yifei Nie and Yilong Li and Yukun Zhou and Yun Ye and Zhichao Liu and Zheng Zhu},
year = {2026},
eprint = {2602.12099},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2602.12099},
}
VLA that reasons in discrete motion tokens (a chain of dynamics) shared across embodiments; predicted futures are scored against what actually happened, so action-free human videos give RL signal. Sim and real tests.VLA 用跨本体共享的离散运动 token 链（Chain-of-Dynamics）表达动作思考；预测每条链导致的未来并与真实结果比对来强化，使没有动作标注的人类视频也能提供监督。在仿真与真机上验证人到机器人的迁移。
@misc{kim2026thinkworld,
title = {{ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning}},
author = {Chaehyun Kim and Chi-Pin Huang and Jisoo Kim and Min-Hung Chen and Yu-Chiang Frank Wang and Seungryong Kim and Fu-En Yang},
year = {2026},
eprint = {2610.37666},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.37666},
}
@misc{wang2026pdrvla,
title = {{PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations}},
author = {Hanqing Wang and Zhaotian Cai and Ruichuan An and Bohan Zeng and Tao Zhang and Yilin Chen and Xu Jiang and Kaiquan Chen and Chongyang Wu and Haoran Li and Fan Yang and Mingan Lin and Wentao Zhang and Zenan Zhou and Youzheng Wu},
year = {2026},
eprint = {2610.37381},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.37381},
}
Wall-OSS-0.5: open 4B VLA on a Qwen2.5-VL-3B backbone, pretrained across 20+ embodiments; reports 60.5% task progress after fine-tuning on 15 real tasks, against 43.0% for π0.5.Wall-OSS-0.5：自变量机器人（X Square Robot）开源的 40 亿参数 VLA 模型，基于 Qwen2.5-VL-3B 视觉语言主干，在 20 余种机器人形态上预训练；报告在 15 个真机任务上微调后平均任务进度 60.5%，π0.5 为 43.0%。
@misc{yu2026wall,
title = {{Wall-OSS-0.5 Technical Report}},
author = {Ryan Yu and Pushi Zhang and Starrick Liu and Brae Liu and Miracle Kang and Shalfun Li and Lights Shi and Ellie Ma and Ping Yang and Chris Pan and Jerry Chen and Dongxiu Liu and Rain Sun and Miles Guo and Byron Zhang and Hugo Zhou and Zach Xu and Vincent Chen and Harrison Huang and James Wang and Dance Kuzi and Andy Zhai and Hang Su and Roy Gan and Lucy Liang and Hao Wang and Qian Wang},
year = {2026},
eprint = {2605.30877},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2605.30877},
}
@misc{yang2026eaglevla,
title = {{EagleVLA: Towards Onboard Real-Time Robot Control via Foresight-Aligned Asynchronous Inference}},
author = {Zebin Yang and Qi Wang and Yunhe Wang and Xiurui Guo and Bo Yu and Shaoshan Liu and Jiafeng Xu and Hao Dong and Meng Li},
year = {2026},
eprint = {2607.12659},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.12659},
}
@misc{wang2026never,
title = {{Never Too Late for Force: Accelerating VLA Post-Training with Reactive Force Injection}},
author = {Yi Wang and Wendi Chen and Zimo Wen and Han Xue and Xueqi Li and Wenye Yu and Zhijie Chen and Hao Yang and Jun Lv and Chuan Wen and Cewu Lu},
year = {2026},
eprint = {2607.14236},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.14236},
}
CoRL 2026 Spotlight. Instead of regenerating action chunks from scratch, keeps a persistent action buffer with staggered flow timesteps, so flow-based VLAs get more reactive while preserving long-horizon context.CoRL 2026 Spotlight。不再每次从头生成动作块，而是维护一个流时间步错开的持久动作缓冲区，让基于流的 VLA 更具反应性，同时保留长时域上下文。
@misc{xiong2026react,
title = {{REACT: Rolling Denoising and Dual Decoupling for Reactive Robot Control with VLA Models}},
author = {Houlong Xiong and Zhenqi Qiu and Zechen Wang and Suohang Zhang and Yiyu Ren and Wanting Xu and Hongfei Niu and Chengyang He and Ge Sun and Ran Cheng and Qian Zhu},
year = {2026},
eprint = {2610.12007},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.12007},
}
CoRL 2026 Spotlight. Studies why VLA models struggle in contact-rich manipulation and separates two failure modes: precision failures, where policies struggle to produce small corrective actions, and force failures.CoRL 2026 Spotlight。分析 VLA 模型为何难以完成接触丰富的操作任务，并区分两类失败模式：难以输出接触时所需微小修正动作的精度失败，以及力失败。
@misc{paresmorlans2026demystifying,
title = {{Demystifying When and Why VLAs Fail in Contact-Rich Tasks and How to Fix Them}},
author = {Carlota Parés-Morlans and Nils Kuhn and Y. Isabel Liu and Alberta Longhini and Jeannette Bohg},
year = {2026},
eprint = {2608.01402},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.01402},
}
CoRL 2026 Oral. Cross-embodiment VLA that treats human interaction traces as a shared language, pre-trained on UniHand-2.0: over 35,000 hours of multimodal data across 30 robot embodiments.CoRL 2026 Oral。跨形态 VLA，把人类交互轨迹当作通用“母语”，在 UniHand-2.0 上预训练：超过 3.5 万小时多模态数据，覆盖 30 种机器人形态。
@misc{luo2026being,
title = {{Being-H0.5: Scaling Human-Centric Robot Learning for Cross-Embodiment Generalization}},
author = {Hao Luo and Ye Wang and Wanpeng Zhang and Sipeng Zheng and Ziheng Xi and Chaoyi Xu and Haiweng Xu and Haoqi Yuan and Chi Zhang and Yiqing Wang and Yicheng Feng and Zongqing Lu},
year = {2026},
eprint = {2601.12993},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2601.12993},
}
Commercially licensed 3B-parameter humanoid VLA: a Cosmos-Reason2 backbone plans and a diffusion transformer outputs motor commands. NVIDIA reports training on about 20,854 hours of human egocentric video.可商用的 30 亿参数人形 VLA：Cosmos-Reason2 主干负责推理规划，扩散 Transformer 输出电机指令。NVIDIA 称用约 20,854 小时人类第一视角视频训练。
H2RB adapts a pretrained VLA from unpaired human and robot demonstrations by aligning observations on geometry and sharing action latents; reports object transfer score rising from 8.0 to 51.8 on real tasks.H2RB 用非配对的人类与机器人示教微调预训练 VLA：按几何对齐观测，并共享动作潜变量；报告真机任务的物体迁移得分从 8.0 升至 51.8。
@misc{park2026human,
title = {{Human-to-Robot Transfer via Bridging Representation and Action Space in VLAs}},
author = {Junhyeong Park and Wook Jung and Taeyoung Kim and Huiwon Jang and Jinwoo Shin},
year = {2026},
eprint = {2611.04391},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2611.04391},
}
VLMs plan with points, boxes and trajectories one coordinate token at a time; VFlash drafts a whole primitive in one pass. Reports lower trajectory error, ~25x faster 3D point forecasting, a WAM gain (62.6% → 70.2%).用点、框、轨迹做视觉规划的视觉语言模型需要逐个坐标 token 解码；VFlash 用轻量草稿器一次前向生成整个图元。报告轨迹误差更低、三维点运动预测约快 25 倍，接入世界动作模型后成功率 62.6% → 70.2%。
@misc{chou2026vflash,
title = {{VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding}},
author = {Yu-Cheng Chou and Shihao Wang and Jin Huang and Chi-Pin Huang and Chiao-An Yang and Xianzhe Fan and Jing Wang and Zhiqi Li and Yuanguo Kuang and Ruijie Zheng and Johan Bjorck and Sifei Liu and Hang Su and Abhishek Badki and Lei Zhang and Alan Yuille and Linxi "Jim'' Fan and Jan Kautz and Zhiding Yu},
year = {2026},
eprint = {2611.08449},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08449},
}
A generalist AI model commands the arm directly through a reference harness (AMP1), with no demos or built-in grasp skills; reports 85.6% zero-shot on RoboLab-120 vs 49.2% for the best VLA, and 104 of 108 real trials.让通用 AI 模型通过参考框架 AMP1 直接指挥机械臂，无需示教，也不预置抓取技能；在 RoboLab-120 上零样本成功率 85.6%（最佳 VLA 为 49.2%），真机 108 次试验成功 104 次。
@misc{chen2026agentic,
title = {{On Agentic Manipulation Policies}},
author = {Chen Chen and Omid Bagherieh and Naveen Nikamanth Appiah Balaji and Jesse T. Gonzalez and Negin Heravi and Fahad Islam and Xueting Li and Jean-Baptiste Regli and Kejia Ren and Petr Švec and David J. Yoon and Matthias Müller and Vladlen Koltun},
year = {2026},
eprint = {2611.09892},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09892},
}
A coding agent drives a bimanual robot through one command line; after each failure it reads the logs and writes or fixes small reusable tool programs.编程智能体通过一条命令行驱动双臂机器人；每次失败后读取日志，编写或修复可复用的小工具程序。
@misc{gao2026hexashell,
title = {{HexaShell: A Universal Harness for Robot Coding Agents in Simulation and the Real World}},
author = {Hongcheng Gao and Alex Zhou and Jay Zhu and Zelin Zheng and Shijia Ge and Xiping Li and Di Wu and Yazhe Wang and Jianshu Zeng and Xuan Shangguan and Yongli Wang and Yexing Wan and Lingyu He and Zhiqi Jia and Sihang Wu and Xiao He},
year = {2026},
eprint = {2610.23479},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.23479},
}
One world model that is also the policy, across 20 robot families mapped to a shared joint-space action format; reports strong WorldArena, RoboTwin 2.0 and LIBERO-plus results and a one-step distilled humanoid policy.同一个世界模型兼作策略，用共享的关节坐标动作空间覆盖 20 类机器人；论文报告其在 WorldArena、RoboTwin 2.0、LIBERO-plus 上表现突出，并蒸馏出可实时运行的单步人形机器人策略。
@misc{sidorov2026greenwam,
title = {{GreenWAM: A World Foundation Model for Robotics}},
author = {Gennady Sidorov and Valerii Kornilov and Denis Grankin and Polina Fedotova and Aleksey Postnikov},
year = {2026},
eprint = {2610.18774},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.18774},
}
LingBot-VA: an autoregressive diffusion framework that learns frame prediction and policy together, with shared video-action latents, closed-loop rollout and asynchronous inference.LingBot-VA：同时学习帧预测和策略的自回归扩散框架，具有视频动作共享潜空间、闭环推演和异步推理。
@misc{li2026causal,
title = {{Causal World Modeling for Robot Control}},
author = {Lin Li and Qihang Zhang and Yiming Luo and Shuai Yang and Ruilin Wang and Fei Han and Mingrui Yu and Zelin Gao and Nan Xue and Xing Zhu and Yujun Shen and Yinghao Xu},
year = {2026},
eprint = {2601.21998},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2601.21998},
}
Asks whether WAMs need future imagination at test time: keeps video co-training during training but skips video generation at inference, separating the two effects in a controlled comparison.追问世界动作模型在测试时是否需要想象未来：训练时保留视频联合训练，推理时跳过视频生成，以对照实验区分两者的作用。
@misc{yuan2026fast,
title = {{Fast-WAM: Do World Action Models Need Test-time Future Imagination?}},
author = {Tianyuan Yuan and Zibin Dong and Yicheng Liu and Hang Zhao},
year = {2026},
eprint = {2603.16666},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2603.16666},
}
An action-centered WAM that predicts actions from the observation and makes video generation optional, so motion accuracy depends less on forecast quality and inference is cheaper.以动作为中心的世界动作模型：由观测预测动作，视频生成变为可选，动作精度不再依赖视频预测质量，推理开销也更低。
@misc{ye2026gigaworld,
title = {{GigaWorld-Policy: An Efficient Action-Centered World--Action Model}},
author = {Angen Ye and Boyuan Wang and Chaojun Ni and Guan Huang and Guosheng Zhao and Hao Li and Hengtao Li and Jie Li and Jindi Lv and Jingyu Liu and Min Cao and Peng Li and Qiuping Deng and Wenjun Mei and Xiaofeng Wang and Xinze Chen and Xinyu Zhou and Yang Wang and Yifan Chang and Yifan Li and Yukun Zhou and Yun Ye and Zhichao Liu and Zheng Zhu},
year = {2026},
eprint = {2603.17240},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2603.17240},
}
Uses semantically coherent action events rather than fixed-length chunks as the unit of learning, to resolve the timescale mismatch between language, vision and control.用语义完整的动作事件代替固定长度动作块作为学习单元，以解决语言、视觉和控制之间的时间尺度不匹配。
@misc{li2026wall,
title = {{WALL-WM: Carving World Action Modeling at the Event Joints}},
author = {Shalfun Li and Victor Yao and Charles Yang and Truth Qu and Regis Cheng and Ryan Yu and Howard Lu and Newton Von and Vincent Chen and Yohann Tang and Maeve Zhang and Ellie Ma and Gody Li and Starrick Liu and Sage Yang and Lorien Shu and J. W. Gao and Ethan Chen and Colin Ye and Yu Sun and Elise Mon and PS Zhang and Neo Li and Lily Li and James Wang and Ping Yang and Chris Pan and Lucy Liang and Hang Su and Roy Gan and Hao Wang and Qian Wang},
year = {2026},
eprint = {2606.01955},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.01955},
}
OpenVLA-OFT — parallel decoding, action chunking and continuous actions make VLA fine-tuning much faster and more accurate.OpenVLA-OFT：并行解码、动作分块与连续动作让 VLA 微调更快、成功率更高。
@misc{kim2025fine,
title = {{Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success}},
author = {Moo Jin Kim and Chelsea Finn and Percy Liang},
year = {2025},
eprint = {2502.19645},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.19645},
}
Co-training on heterogeneous data lets a mobile manipulator clean kitchens and bedrooms in homes it has never seen.通过异构数据联合训练，让移动操作机器人在从未见过的家庭中整理厨房和卧室。
@misc{intelligence2025vision,
title = {{$π_{0.5}$: a Vision-Language-Action Model with Open-World Generalization}},
author = {Physical Intelligence and Kevin Black and Noah Brown and James Darpinian and Karan Dhabalia and Danny Driess and Adnan Esmail and Michael Equi and Chelsea Finn and Niccolo Fusai and Manuel Y. Galliker and Dibya Ghosh and Lachy Groom and Karol Hausman and Brian Ichter and Szymon Jakubczak and Tim Jones and Liyiming Ke and Devin LeBlanc and Sergey Levine and Adrian Li-Bell and Mohith Mothukuri and Suraj Nair and Karl Pertsch and Allen Z. Ren and Lucy Xiaoyang Shi and Laura Smith and Jost Tobias Springenberg and Kyle Stachowicz and James Tanner and Quan Vuong and Homer Walke and Anna Walling and Haohuan Wang and Lili Yu and Ury Zhilinsky},
year = {2025},
eprint = {2504.16054},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2504.16054},
}
@misc{nvidia2025gr,
title = {{GR00T N1: An Open Foundation Model for Generalist Humanoid Robots}},
author = {NVIDIA and : and Johan Bjorck and Fernando Castañeda and Nikita Cherniadev and Xingye Da and Runyu Ding and Linxi "Jim" Fan and Yu Fang and Dieter Fox and Fengyuan Hu and Spencer Huang and Joel Jang and Zhenyu Jiang and Jan Kautz and Kaushil Kundalia and Lawrence Lao and Zhiqi Li and Zongyu Lin and Kevin Lin and Guilin Liu and Edith Llontop and Loic Magne and Ajay Mandlekar and Avnish Narayan and Soroush Nasiriany and Scott Reed and You Liang Tan and Guanzhi Wang and Zu Wang and Jing Wang and Qi Wang and Jiannan Xiang and Yuqi Xie and Yinzhen Xu and Zhenjia Xu and Seonghyeon Ye and Zhiding Yu and Ao Zhang and Hao Zhang and Yizhou Zhao and Ruijie Zheng and Yuke Zhu},
year = {2025},
eprint = {2503.14734},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2503.14734},
}
@misc{shukor2025smolvla,
title = {{SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics}},
author = {Mustafa Shukor and Dana Aubakirova and Francesco Capuano and Pepijn Kooijmans and Steven Palma and Adil Zouitine and Michel Aractingi and Caroline Pascal and Martino Russi and Andres Marafioti and Simon Alibert and Matthieu Cord and Thomas Wolf and Remi Cadene},
year = {2025},
eprint = {2506.01844},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2506.01844},
}
Open embodied VLMs (7B and 32B) trained for spatial grounding, affordances, trajectories and multi-step task planning — a "brain" that sits above low-level policies.开源具身视觉语言模型（7B 与 32B），面向空间定位、可供性、轨迹预测与多步任务规划，作为底层策略之上的“大脑”。
@misc{team2025robobrain,
title = {{RoboBrain 2.0 Technical Report}},
author = {BAAI RoboBrain Team and Mingyu Cao and Huajie Tan and Yuheng Ji and Xiansheng Chen and Minglan Lin and Zhiyu Li and Zhou Cao and Pengwei Wang and Enshen Zhou and Yi Han and Yingbo Tang and Xiangqi Xu and Wei Guo and Yaoxu Lyu and Yijie Xu and Jiayu Shi and Mengfei Du and Cheng Chi and Mengdi Zhao and Xiaoshuai Hao and Junkai Zhao and Xiaojie Zhang and Shanyu Rong and Huaihai Lyu and Zhengliang Cai and Yankai Fu and Ning Chen and Bolun Zhang and Lingfeng Zhang and Shuyi Zhang and Dong Liu and Xi Feng and Songjing Wang and Xiaodan Liu and Yance Jiao and Mengsi Lyu and Zhuo Chen and Chenrui He and Yulong Ao and Xue Sun and Zheqi He and Jingshu Zheng and Xi Yang and Donghai Shi and Kunchang Xie and Bochao Zhang and Shaokai Nie and Chunlei Men and Yonghua Lin and others},
year = {2025},
eprint = {2507.02029},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2507.02029},
}
Learns task-centric latent actions from videos (in DINO feature space, filtered by language) so one policy can be pretrained across robots and human video; reports beating OpenVLA with far less pretraining compute.在 DINO 特征空间中、借助语言指令过滤，从视频里学习“任务相关”的隐动作，使策略可跨机器人与人类视频预训练；报告以少得多的预训练算力超过 OpenVLA。
@misc{bu2025univla,
title = {{UniVLA: Learning to Act Anywhere with Task-centric Latent Actions}},
author = {Qingwen Bu and Yanting Yang and Jisong Cai and Shenyuan Gao and Guanghui Ren and Maoqing Yao and Ping Luo and Hongyang Li},
year = {2025},
eprint = {2505.06111},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.06111},
}
One model for UI navigation and robot manipulation, pretrained on images, videos and robot data labelled with Set-of-Mark (what to act on) and Trace-of-Mark (how things move); model and code are public.同一个模型同时做界面操作与机器人操作：在图像、视频和机器人数据上预训练，用 Set-of-Mark 标注可操作对象、用 Trace-of-Mark 标注运动轨迹；模型与代码公开。
@misc{yang2025magma,
title = {{Magma: A Foundation Model for Multimodal AI Agents}},
author = {Jianwei Yang and Reuben Tan and Qianhui Wu and Ruijie Zheng and Baolin Peng and Yongyuan Liang and Yu Gu and Mu Cai and Seonghyeon Ye and Joel Jang and Yuquan Deng and Lars Liden and Jianfeng Gao},
year = {2025},
eprint = {2502.13130},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2502.13130},
}
WALL-OSS: an open embodied foundation model that links instruction reasoning, subgoal decomposition and action generation in one differentiable model ('unified cross-level CoT'); training code and checkpoints are public.WALL-OSS：开源具身基础模型，在同一可微模型中串联指令推理、子目标分解与细粒度动作生成（"统一跨层级思维链"）；训练代码与模型权重已开源。
@misc{zhai2025igniting,
title = {{Igniting VLMs toward the Embodied Space}},
author = {Andy Zhai and Brae Liu and Bruno Fang and Chalse Cai and Ellie Ma and Ethan Yin and Hao Wang and Hugo Zhou and James Wang and Lights Shi and Lucy Liang and Make Wang and Qian Wang and Roy Gan and Ryan Yu and Shalfun Li and Starrick Liu and Sylas Chen and Vincent Chen and Zach Xu},
year = {2025},
eprint = {2509.11766},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2509.11766},
}
Trains a VLA largely on world-model-generated data (re-rendered real videos, new viewpoints, sim2real, human-to-robot video); reports gains over π0 on dexterous and mobile tasks, plus a 402M variant for Jetson Orin.VLA 训练大量依赖世界模型生成的数据（真实视频重渲染、新视角、仿真转真实、人类视频转机器人）；论文报告在灵巧与移动操作任务上优于 π0，并提供可在 Jetson Orin 上运行的 4.02 亿参数小模型。
@misc{team2025gigabrain,
title = {{GigaBrain-0: A World Model-Powered Vision-Language-Action Model}},
author = {GigaBrain Team and Angen Ye and Boyuan Wang and Chaojun Ni and Guan Huang and Guosheng Zhao and Haoyun Li and Jie Li and Jiagang Zhu and Lv Feng and Peng Li and Qiuping Deng and Runqi Ouyang and Wenkang Qin and Xinze Chen and Xiaofeng Wang and Yang Wang and Yifan Li and Yilong Li and Yiran Ding and Yuan Xu and Yun Ye and Yukun Zhou and Zhehao Dong and Zhenan Wang and Zhichao Liu and Zheng Zhu},
year = {2025},
eprint = {2510.19430},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.19430},
}
One autoregressive model that is both a VLA policy and an image-predicting world model; the two help each other on LIBERO, and an attention mask hiding earlier actions curbs error build-up in action chunks.单个自回归模型同时充当 VLA 策略与预测未来图像的世界模型，二者在 LIBERO 上相互促进；论文还发现自回归生成动作块会累积误差，用屏蔽先前动作的注意力掩码可以缓解。
@misc{cen2025worldvla,
title = {{WorldVLA: Towards Autoregressive Action World Model}},
author = {Jun Cen and Chaohui Yu and Hangjie Yuan and Yuming Jiang and Siteng Huang and Jiayan Guo and Xin Li and Yibing Song and Hao Luo and Fan Wang and Deli Zhao and Hao Chen},
year = {2025},
eprint = {2506.21539},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.21539},
}
Online RL for VLAs with GRPO and only binary success rewards on OpenVLA-OFT, plus exploration tweaks; reports 99.1% on LIBERO and 70.4% on RoboTwin 1.0 (vs 39.8% for the SFT baseline).在 OpenVLA-OFT 上用 GRPO 与仅"成功/失败"二值奖励做在线强化学习，并加入若干促进探索的改动；论文报告 LIBERO 成功率 99.1%、RoboTwin 1.0 为 70.4%（SFT 基线 39.8%）。
@misc{li2025simplevla,
title = {{SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning}},
author = {Haozhan Li and Yuxin Zuo and Jiale Yu and Yuhao Zhang and Zhaohui Yang and Kaiyan Zhang and Xuekai Zhu and Yuchen Zhang and Tianxing Chen and Ganqu Cui and Dehui Wang and Dingxiang Luo and Yuchen Fan and Youbang Sun and Jia Zeng and Jiangmiao Pang and Shanghang Zhang and Yu Wang and Yao Mu and Bowen Zhou and Ning Ding},
year = {2025},
eprint = {2509.09674},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2509.09674},
}
Open RL training framework for VLAs (PPO/GRPO across GPU-parallel simulators); reports 98.11% on 130 LIBERO tasks and 97.66% on 25 ManiSkill tasks after RL — a common base for VLA+RL work.面向 VLA 的开源强化学习训练框架（PPO/GRPO，支持 GPU 并行仿真器）；论文报告经 RL 后在 130 个 LIBERO 任务上成功率 98.11%、25 个 ManiSkill 任务上 97.66%，是 VLA+RL 研究常用的底座。
@misc{zang2025rlinf,
title = {{RLinf-VLA: A Unified and Efficient Framework for Reinforcement Learning of Vision-Language-Action Models}},
author = {Hongzhi Zang and Mingjie Wei and Si Xu and Yongji Wu and Zhen Guo and Yuanqing Wang and Hao Lin and Peihong Wang and Liangzhi Shi and Yuqing Xie and Zhexuan Xu and Zhihao Liu and Kang Chen and Wenhao Tang and Quanlu Zhang and Weinan Zhang and Chao Yu and Yu Wang},
year = {2025},
eprint = {2510.06710},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.06710},
}
A small open diffusion-head Transformer pretrained on Open X-Embodiment that fine-tunes quickly to new robots and sensors.在 Open X-Embodiment 上预训练的开源扩散头 Transformer 策略，可快速微调到新机器人和新传感器。
@misc{team2024octo,
title = {{Octo: An Open-Source Generalist Robot Policy}},
author = {Octo Model Team and Dibya Ghosh and Homer Walke and Karl Pertsch and Kevin Black and Oier Mees and Sudeep Dasari and Joey Hejna and Tobias Kreiman and Charles Xu and Jianlan Luo and You Liang Tan and Lawrence Yunliang Chen and Pannag Sanketi and Quan Vuong and Ted Xiao and Dorsa Sadigh and Chelsea Finn and Sergey Levine},
year = {2024},
eprint = {2405.12213},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2405.12213},
}
@misc{liu2024rdt,
title = {{RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation}},
author = {Songming Liu and Lingxuan Wu and Bangguo Li and Hengkai Tan and Huayu Chen and Zhengyi Wang and Ke Xu and Hang Su and Jun Zhu},
year = {2024},
eprint = {2410.07864},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2410.07864},
}
@misc{li2024evaluating,
title = {{Evaluating Real-World Robot Manipulation Policies in Simulation}},
author = {Xuanlin Li and Kyle Hsu and Jiayuan Gu and Karl Pertsch and Oier Mees and Homer Rich Walke and Chuyuan Fu and Ishikaa Lunawat and Isabel Sieh and Sean Kirmani and Sergey Levine and Jiajun Wu and Chelsea Finn and Hao Su and Quan Vuong and Ted Xiao},
year = {2024},
eprint = {2405.05941},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2405.05941},
}
Self-supervised image features that work frozen for depth, segmentation and retrieval; a common visual backbone in VLAs (e.g. OpenVLA) and latent world models.自监督学习的图像特征，冻结即可用于深度估计、分割和检索；是 VLA（如 OpenVLA）和潜空间世界模型常用的视觉骨干。
@misc{oquab2023dinov,
title = {{DINOv2: Learning Robust Visual Features without Supervision}},
author = {Maxime Oquab and Timothée Darcet and Théo Moutakanni and Huy Vo and Marc Szafraniec and Vasil Khalidov and Pierre Fernandez and Daniel Haziza and Francisco Massa and Alaaeldin El-Nouby and Mahmoud Assran and Nicolas Ballas and Wojciech Galuba and Russell Howes and Po-Yao Huang and Shang-Wen Li and Ishan Misra and Michael Rabbat and Vasu Sharma and Gabriel Synnaeve and Hu Xu and Hervé Jegou and Julien Mairal and Patrick Labatut and Armand Joulin and Piotr Bojanowski},
year = {2023},
eprint = {2304.07193},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2304.07193},
}
Tokenized actions plus a Transformer trained on 130k real demos showed that robot policies scale with data diversity.将动作离散化为 token，用 13 万条真实演示训练 Transformer，证明机器人策略会随数据多样性扩展。
@misc{brohan2022rt,
title = {{RT-1: Robotics Transformer for Real-World Control at Scale}},
author = {Anthony Brohan and Noah Brown and Justice Carbajal and Yevgen Chebotar and Joseph Dabis and Chelsea Finn and Keerthana Gopalakrishnan and Karol Hausman and Alex Herzog and Jasmine Hsu and Julian Ibarz and Brian Ichter and Alex Irpan and Tomas Jackson and Sally Jesmonth and Nikhil J Joshi and Ryan Julian and Dmitry Kalashnikov and Yuheng Kuang and Isabel Leal and Kuang-Huei Lee and Sergey Levine and Yao Lu and Utsav Malla and Deeksha Manjunath and Igor Mordatch and Ofir Nachum and Carolina Parada and Jodilyn Peralta and Emily Perez and Karl Pertsch and Jornell Quiambao and Kanishka Rao and Michael Ryoo and Grecia Salazar and Pannag Sanketi and Kevin Sayed and Jaspiar Singh and Sumedh Sontakke and Austin Stone and Clayton Tan and Huong Tran and Vincent Vanhoucke and Steve Vega and Quan Vuong and Fei Xia and Ted Xiao and Peng Xu and Sichun Xu and Tianhe Yu and others},
year = {2022},
eprint = {2212.06817},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2212.06817},
}
A 2.4B VLM + flow-matching VLA pretrained only on public data; reports beating all but ~2× larger models, plus cheap register-token memory.仅用公开数据预训练的 2.4B VLM + 流匹配 VLA；报告称只输给约两倍大的模型，并用少量寄存器 token 实现低成本记忆。
@misc{marrie2026reki,
title = {{REKI: Registering Knowledge in VLAs}},
author = {Juliette Marrie and Paul Pacaud and Adrien Ramanana Rahary and Shizhe Chen and Moritz Böhle},
year = {2026},
eprint = {2610.22178},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.22178},
}
Gives π0.5 memory by caching compact latent transitions between observations, aimed at tasks where key objects go out of view.通过缓存相邻观测之间的紧凑潜在转移，为 π0.5 加入记忆，面向关键物体移出视野的任务。
@misc{sun2026ltm,
title = {{LTM: Latent Transition Based In-Context Memory for Vision-Language-Action Models}},
author = {Qixin Sun and WenLong Gao and Xiansheng Chen and Zhongyuan Wang and Si Liu and Pengwei Wang},
year = {2026},
eprint = {2610.21362},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.21362},
}
Speeds up VLA inference by keeping only the visual tokens the current task needs, with a token budget that changes from step to step.只保留当前任务所需的视觉 token、并让 token 预算随时间步变化，以加快 VLA 推理。
@misc{xu2026fastvla,
title = {{FastVLA: Task-Conditioned and Temporally Adaptive Visual Token Compression for Efficient Vision-Language-Action Models}},
author = {Jiuyi Xu and Chengyuan Qian and Zhengzhong Tu and Yang Sui and Yangming Shi},
year = {2026},
eprint = {2610.22146},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.22146},
}
Learns skills separately, then consolidates them after the fact with a nested hypernetwork that generates VLA parameters instead of merging weights.先独立学习各项技能，再用嵌套超网络“生成”VLA 参数来事后整合技能，而不是直接合并权重。
Argues VLAs overfit observation→action pairs and adds explicit supervision of the physical state change an action should cause, rather than predicting future states in language or pixels.认为 VLA 容易过拟合“观测→动作”的对应关系，因此显式监督动作应引起的物理状态变化，而不是在语言或像素空间里预测未来状态。
@misc{lu2026vla,
title = {{VLA-$Δ$: Learning Physical State Transition for Vision-Language-Action Models}},
author = {Qi Lu and Jiarun Li and Ce Zhang and Miao Pan and Yi Pan and Xu Wang and Wenqi Zhang and Xuhong Zhang},
year = {2026},
eprint = {2610.21306},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.21306},
}
Keeps π0.5's pretrained vision-language encoder but swaps its Transformer action expert — run many times per denoising loop — for a compact Mamba expert to cut inference cost.保留 π0.5 预训练的视觉语言编码器，把在去噪循环中被反复调用的 Transformer 动作专家换成紧凑的 Mamba 专家，以降低推理开销。
Instead of feeding a VLA a longer frame history, learns to condense past frames into a few tokens that keep short-lived cues such as brief contacts and occlusions.不靠简单加长历史帧窗口，而是学习把过去的帧压缩成少量 token，保留短暂接触、遮挡等转瞬即逝的线索。
@misc{hong2026condensevla,
title = {{CondenseVLA: Learnable History Condensation for Efficient Multi-Frame VLA}},
author = {Feiyang Hong and Yujie Wei and Bo Zhao and Xiu Su and Hongxun Yao and Shuo Yang},
year = {2026},
eprint = {2610.25927},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.25927},
}
Post-hoc kernel-variance heads (Gaussian-process variance over training-demo embeddings) estimate one VLA policy's epistemic uncertainty without an ensemble, so the robot knows when to ask for help.在训练完成后为单个 VLA 策略加上轻量的核方差头（基于训练示教嵌入的高斯过程方差）来估计认知不确定性，无需训练和运行策略集成，让机器人知道何时该求助或补充示教。
@misc{keller2026post,
title = {{Post-hoc Epistemic Uncertainty for Vision-Language-Action Policies}},
author = {Johannes Keller and Albrecht M. Bloße and Adrian Lindenmeyer and Thomas Neumuth and Daniel Schneider},
year = {2026},
eprint = {2610.29730},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.29730},
}
Tests six flow-matching VLAs and reports that one-step Euler action generation largely matches multi-step sampling at scale — a useful check before adding special one-shot objectives to speed up control (CoRL 2026).在六个基于流匹配的 VLA 上实验，报告单步 Euler 生成动作在规模足够时与多步采样效果基本相当——在为提速引入专门的单步生成目标之前值得先看（CoRL 2026）。
@misc{sujit2026when,
title = {{When One Step Is Enough: An Empirical Study of Euler Integration in Flow-Matching Policies}},
author = {Shivakanth Sujit and Luca Nunziante and Rousslan Fernand Julien Dossa and Kai Arulkumaran},
year = {2026},
eprint = {2610.30735},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.30735},
}
Reuses the unexecuted tail of a flow-matching VLA's action chunk to start the next one, with a small confidence model setting how much to re-denoise; reports lower inference cost on π0.5, SmolVLA and others.流匹配 VLA 每次只执行动作块的前几步，剩余部分通常丢弃；本文把这段"动作尾巴"作为下一块的起点，并用小型置信度模型决定还需去噪多少步；论文报告在 π0.5、SmolVLA 等模型上降低推理开销。
@misc{he2026vla,
title = {{VLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action Reuse}},
author = {Sicheng He and Yi Pan and Qiuting Yu and Jiaming Huang and Qi Lu and Shuyan Wu and Hongyan Feng and Xuanyu Liu and Jun Zhang and Jie Zhang and Wenqi Zhang and Xuhong Zhang and Jianwei Yin},
year = {2026},
eprint = {2610.25043},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2610.25043},
}
Adds a 257M latent world model to a pretrained VLA so predicted future features condition the action expert, with no pixel generation or video pretraining; reports 98.7% on LIBERO and real-robot gains over π0.5.在预训练 VLA 上加一个 2.57 亿参数的潜空间世界模型，用预测的未来表征去条件化动作专家，无需像素生成或大规模视频预训练；报告 LIBERO 成功率 98.7%，真机任务上明显优于 π0.5。
@misc{wang2026wave,
title = {{WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction}},
author = {Haomin Wang and Anqiong Ji and Ke Wang and Zhongshan Liu and Ru Ying and Ran Zheng and Zhipeng Zhang},
year = {2026},
eprint = {2610.18568},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.18568},
}
Instead of one fixed set of VLA weights, generates context-dependent low-rank updates for both the vision-language and action modules; reports gains up to 16.3% on LIBERO/Meta-World and up to 14% over π0.5 on real tasks.不再让 VLA 在所有交互情境下共用固定参数，而是按当前情境为视觉语言模块和动作模块生成耦合的低秩参数更新；报告在 LIBERO/Meta-World 上最高提升 16.3%，真机上相对 π0.5 最高提升 14%。
@misc{liu2026morphact,
title = {{MorphAct: Robot Policies Morphing with the Worlds They Act In}},
author = {Ziyan Liu and Jiting Liu and Weihong Zhong and Wenda Yu and Yeqiu Chen and Jason Li and Xiaojiang Peng and Shuyuan Hu and Kai Wang and Bo Zhao},
year = {2026},
eprint = {2610.20019},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.20019},
}
Wraps a frozen VLA in two loops — one keeps user preferences and scene memory for planning, one blocks hazardous actions and calls recovery skills; reports violations down from 52.6% to 6.1%, recovery up from 13% to 80%.给冻结的 VLA 套上双环框架：规划环保存用户偏好与场景记忆以做个性化任务分解，执行环拦截危险接触并调用可复用的恢复技能；报告安全违规率从 52.6% 降到 6.1%，失败恢复成功率从 13% 提升到 80%。
@misc{wang2026carryon,
title = {{CarryOn: A Dual-Loop Harness for Personalized and Reliable Robot Assistance}},
author = {Zixuan Wang and Yuqi Liu and Yuxin Chen and Sengqiao Yang and Jinhui Ye and Pengguang Chen and Changsheng Lu and Shu Liu and Bei Yu and Jiaya Jia},
year = {2026},
eprint = {2610.21453},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.21453},
}
Survey of 'agentic robots' — foundation-model agents that plan, collect data, build environments and recover from errors — organised into 17 categories under model, data, environment and harness; companion paper list.关于"智能体化机器人"的综述：基础模型智能体负责规划、采集与筛选数据、构建环境与测试并自主纠错；把相关研究归为模型、数据、环境、执行框架四大领域下的 17 个类别，附配套论文列表。
@misc{qin2026towards,
title = {{Towards Agentic Robots: Model, Data, Environment, and Harness}},
author = {Chonghan Qin and Dongming Wu and Kaituo Feng and Jiaxu Wang and Xinlei Yu and Peiwen Sun and Yijia Fan and Xilin He and Bingao Chen and Weitong Lian and Ryan Ji and Wenqi Ren and Xiaobin Hu and Dahua Lin and Yu-Gang Jiang and Xiangyu Yue},
year = {2026},
eprint = {2610.20442},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.20442},
}
Survey framing computer-use and robot agents with five functions — perceive, anticipate, plan, act, verify; across 64 benchmarks it finds only 5 test anticipation and 3 test verification. Companion paper list.用"感知、预判、规划、执行、验证"五种能力统一看待电脑操作智能体和机器人智能体的综述；统计 64 个基准后发现只有 5 个评测预判、3 个评测验证。附配套论文列表。
@misc{chen2026multimodal,
title = {{Multimodal Embodied Agents: A Unified Capability-centric Perspective from Computer-Use to Robot-Use}},
author = {Yanzhe Chen and Ziyi Yang and Jifeng Zhu and Qiming Huang and Ruihe An and Peiyao Xu and Hesen Yang and Runda Liu and Chang Gong and Zhijun Cao and Zechen Bai and Wenzheng Zeng and Yiqi Lin and Guoqiang Liang and Kevin Yuchen Ma and Kevin Qinghong Lin and Mike Zheng Shou},
year = {2026},
eprint = {2610.30108},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.30108},
}
Fully open successor to MolmoAct: a flow-matching action expert on an embodied-reasoning VLM, the OpenFAST tokenizer and a 720 h bimanual YAM dataset; reports beating π0.5 on 7 sim and real benchmarks.MolmoAct 的全开源后继：在具身推理 VLM（MolmoER）上挂接流匹配连续动作专家，并发布开源动作分词器 OpenFAST 与 720 小时双臂 YAM 遥操作数据集；论文报告在 7 个仿真与真机基准上超过 π0.5 等基线，权重、训练代码与数据全部开放。
@misc{fang2026molmoact,
title = {{MolmoAct2: Action Reasoning Models for Real-world Deployment}},
author = {Haoquan Fang and Jiafei Duan and Donovan Clay and Sam Wang and Shuo Liu and Weikai Huang and Xiang Fan and Wei-Chuan Tsai and Shirui Chen and Yi Ru Wang and Shanli Xing and Jaemin Cho and Jae Sung Park and Ainaz Eftekhar and Peter Sushko and Karen Farley and Angad Wadhwa and Cole Harrison and Winson Han and Ying-Chun Lee and Eli VanderBilt and Rose Hendrix and Suveen Ellawela and Lucas Ngoo and Joyce Chai and Zhongzheng Ren and Ali Farhadi and Dieter Fox and Ranjay Krishna},
year = {2026},
eprint = {2605.02881},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2605.02881},
}
Unitree's open VLA for humanoid manipulation: a VLM further pretrained on robot data, fine-tuned on 12 open Unitree G1 datasets; the README reports one policy handling 12 task categories on a real G1.宇树开源的人形机器人操作 VLA：在 UnifoLM-VLM 基础上用机器人操作数据继续预训练，再在 12 个公开的 Unitree G1 数据集上微调；README 称单一策略可在真实 G1 上完成 12 类复杂操作任务，另提供 LIBERO 微调权重。
Open spatiotemporal embodied foundation models (2B, 8B, 30B-A3B MoE) for egocentric understanding, localization, grounded reasoning and planning, plus post-trained Nav, Plan, VLA and CoP variants.开源时空具身基础模型（2B、8B 与 30B-A3B MoE），覆盖第一视角理解、时空定位、物理落地推理与规划，并提供导航、规划、VLA、CoP 等后训练版本。
@misc{dang2026rynnbrain,
title = {{RynnBrain: Open Embodied Foundation Models}},
author = {Ronghao Dang and Jiayan Guo and Bohan Hou and Sicong Leng and Kehan Li and Xin Li and Jiangpin Liu and Yunxuan Mao and Zhikai Wang and Yuqian Yuan and Minghao Zhu and Xiao Lin and Yang Bai and Qian Jiang and Yaxi Zhao and Minghua Zeng and Junlong Gao and Yuming Jiang and Jun Cen and Siteng Huang and Liuyi Wang and Wenqiao Zhang and Chengju Liu and Jianfei Yang and Shijian Lu and Deli Zhao},
year = {2026},
eprint = {2602.14979},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.14979},
}
Practical warning: fine-tuning VLA weights in bf16 without an fp32 master copy rounds away AdamW updates below 1/512 of a weight; finds this recipe in 9 of 12 open VLA codebases and 79%+ of updates lost on LIBERO.实用警示：VLA 权重以 bf16 微调且不保留 fp32 主副本时，小于权重 1/512 的 AdamW 更新会被舍入抹掉；报告 12 个开源 VLA 代码库中 9 个的默认或文档配方如此训练，LIBERO 实验中至少 79% 的更新被抹除。
Six frontier general-purpose models drive a LIBERO-10 arm through tool calls (end-effector poses, gripper), no robot training; best reported success is 44%. More reasoning effort usually helps but costs more.让六个前沿通用大模型通过工具调用（指定末端位姿与夹爪开合）直接控制 LIBERO-10 机械臂，无任何机器人专门训练；报告的最高成功率为 44%。提高推理强度通常有帮助，但推理成本更高。
Harness where a frontier LLM reviews a frozen π0.5 policy less often as experience builds up and focuses on recovery after errors; reports ~57% fewer reviewer tokens and higher success on four sim tasks.让前沿大模型监督冻结的 π0.5 策略的执行框架：随着经验积累逐步减少审查频率，并把监督集中在出错后的恢复上；报告在四个仿真任务上审查 token 平均减少约 57%，成功率也更高。
@misc{lin2026fastharness,
title = {{FastHarness: Experience-Driven Review Scheduling for Vision-Language-Action Policies}},
author = {Tao Lin and Yuxin Du and Haoxiao Wang and Zile Zhong and Bo Zhao},
year = {2026},
eprint = {2610.31427},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.31427},
}
Between direct action prediction and full video prediction: VLDA learns a compact latent 'Delta' of task-relevant state change and uses it to seed action denoising. Built on Baidu's open LoongForge.介于直接预测动作与完整预测未来画面之间：VLDA 学习一个紧凑的隐变量 Delta 表示与任务相关的状态变化，并用它引导动作去噪；实现于百度开源训练框架 LoongForge。
@misc{zhang2026vision,
title = {{Vision-Language-Delta-Action: Learning Compact Deltas as Actionable Transition Interfaces for Robotic Manipulation}},
author = {Lingsen Zhang and Wei Li and Jie He and Henghua Zhang and Kai Mo and Shiyong Li and Yanpeng Wang and Rui Shao},
year = {2026},
eprint = {2610.25874},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25874},
}
Deployment knob for flow-matching VLAs: fewer solver steps hurt under fast replanning but help with long open-loop chunks. Tuning both together, training-free, reports higher success and lower latency for π0/π0.5/GR00T.流匹配 VLA 的部署调参发现：求解步数少在频繁重规划时会降低精度，但在长时开环执行动作块时反而有利。报告在 π0、π0.5、GR00T 上联合调整两者，无需训练即可提高成功率并降低推理延迟。
@misc{yang2026coupling,
title = {{Coupling Execution Horizon and Sampling Depth in Flow-Matching VLA Policies}},
author = {Jiabing Yang and Yuan Xu and Yixiang Chen and Qisen Ma and Peiyan Li and Kai Wang and Chenhang Cui and Tao Yu and Zichen Wen and Wen Ye and Bowen Fang and Ziheng He and Xiangnan Wu and Yingda Li and Yan Huang and Liang Wang},
year = {2026},
eprint = {2610.32102},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.32102},
}
Report: GPT-6-Astra drives a robot on EBench via waypoints or by calling a VLA/WAM policy. Direct control leads on short tasks, trails on long-horizon and contact-rich ones; calling a policy cuts token cost.评测报告：在 EBench 上让 GPT-6-Astra 直接输出路径点控制机器人，或通过执行框架调用 VLA/WAM 策略。直接控制在短时任务上领先，但在长时程与接触密集任务上落后；调用策略可显著减少 token 开销。
@misc{yao2026profiling,
title = {{Profiling a Frontier Agent with EBench: From Direct Control to Harness with Policy}},
author = {Yifei Yao and Yude Zou and Guanbo Jiao and Xing Gao and Ning Gao and Yukai Wang and Jiantong Chen and Zihou Zhu and Xinyu Li and Ningyu Yan and Shuai Wang and Jianming Ma and Haoyu Guo and Weinan Zhang and Chunhua Shen},
year = {2026},
eprint = {2610.35832},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.35832},
}
Renders wrist-camera views from a 3D Gaussian Splatting scan of the real workspace to fine-tune π0.5 for precise reaching in e-waste disassembly; reports 80% vs. 58.7% with 150 teleop demos.用真实工作台的 3D 高斯泼溅重建渲染腕部相机训练视角，微调 π0.5 以完成电子垃圾拆解中的精确抵达；报告成功率 80%，而 150 条遥操作示教为 58.7%，人工采集不到 10 分钟。
Mixed-precision quantization for VLAs that sets each weight matrix's bit width by how much it affects actions; reports 92.7% LIBERO success at 3 bits (above AWQ and GPTQ) and 70.5% at 2.5 bits.面向 VLA 的混合精度量化：根据每个权重矩阵对输出动作的影响分配比特数；报告在 3 比特下 LIBERO 成功率 92.7%（高于 AWQ 与 GPTQ），2.5 比特下仍有 70.5%。
@misc{qiu2026agq,
title = {{AGQ: Action-Guided Ultra-Low-Bit Quantization for Vision-Language-Action Models}},
author = {Xuhong Qiu and Zhuoyu Wu and Qiawei Zheng and Wenhui Ou and Wenqi Fang and Raphaël C. -W. Phan},
year = {2026},
eprint = {2610.33074},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.33074},
}
Pretrains a 4B VLA on 3D-grounded latent actions mined from 30K hours of human, robot and sim video, then adapts to real robot actions. Reports 98.8% on LIBERO and 88.6% on RoboTwin 2.0.从 3 万小时人类、机器人与仿真视频中学习带 3D 约束的连续潜在动作（RynnLAM），用于预训练 4B 参数 VLA，再适配到具体机器人动作空间。报告 LIBERO 98.8%、RoboTwin 2.0 平均 88.6%。
@misc{zhao2026rynnvla,
title = {{RynnVLA-Latent: Scaling Robotic Manipulation Foundation Models with 3D-Grounded Latent Actions}},
author = {Haoyu Zhao and Xingyue Zhao and Zhian Su and Bohan Hou and Yinzhou Tang and Fengyi Wang and Kehan Li and Yuming Jiang and Siteng Huang and Xin Li and Zhenyu Zeng and Mingxiu Chen and Deli Zhao and Zhongyu Li},
year = {2026},
eprint = {2610.34356},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.34356},
}
Meta-agent that improves a robot's harness by exploring: keeps a lineage graph of harness versions and runs new rollouts to probe failures before editing. Reports more success with fewer rollouts than static reflection.通过主动探索改进机器人 harness 的元智能体：维护 harness 版本谱系图，先主动生成新回放探查失败与恢复行为再修改；报告在 LIBERO-Pro、RoboCasa、RoboTwin 上比静态反思用更少仿真回放取得更高成功率。
@misc{zhang2026vista,
title = {{VISTA: Active Exploration for Self-Improvement in Embodied Agents}},
author = {Zhang Zhang and Shengjun Zhang and Simin Huang and Min Chen and Chong Xia and Xin Fei and Yueqi Duan},
year = {2026},
eprint = {2610.35051},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.35051},
}
Teaches a VLA to know when it is done or failing: stop and failure signals are two extra action dimensions, denoised faster than the actions so they can guide them; no frame-level progress labels. Real-robot tests.让 VLA 自己判断任务完成或失败：把停止与失败信号作为动作块的两个额外维度，与动作一起去噪且去噪更快，从而引导后续动作生成；无需逐帧进度标注，在真机操作实验中验证。
@misc{liu2026echo,
title = {{Echo: Coupled Action and Task Status Signal Denoising for Vision Language Action Models}},
author = {Yufeng Liu and Bohao Fan and Juntu Zhao and Xinyi Wang and Hang Yu and Wenzhuan Luo and Yingzhi Wang and Qinghai Li and Yingdong Hu and Junliang Guo and Richard Wang and Dequan Wang and Yang Gao and Cheng Ma},
year = {2026},
eprint = {2610.31569},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.31569},
}
Latent action model that fuses synchronized camera views to resolve depth and occlusion. Reports 17% better 6D action decodability than single-view and +13.5% policy success (vs +6.2%); human-hand demos help on hardware.融合多路同步相机视角的潜在动作模型（LAM），用交叉注意力解决深度与遮挡问题；报告 6D 动作可解码性比单视角高 17%，下游策略成功率提升 13.5%（单视角为 6.2%），真机上用人手演示也比纯动作策略高 20%。
@misc{hausdorfer2026multi,
title = {{Multi-View Latent Action Pretraining for Cross-Embodiment Robot Learning from Videos}},
author = {Oliver Hausdörfer and Felix Minzenmay and Mani Majd and Maximilian P. Christof and Shiyuan Zhang and Angela P. Schoellig},
year = {2026},
eprint = {2610.27794},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.27794},
}
Splits the VLA vs WAM debate into backbone, future loss and inference machinery: a foresight loss predicts frozen future features, no video generation. Reports beating FastWAM on LIBERO-Plus (64.5% vs 55.3%).把 VLA 与世界动作模型（WAM）之争拆成骨干初始化、未来预测损失和推理时是否保留预测模块三个选择：用预测冻结未来特征的前瞻损失，训练与推理都不生成视频；报告在 LIBERO-Plus 上 64.5%，FastWAM 为 55.3%。
@misc{yu2026vision,
title = {{Vision-Language-Action or World-Action? Just Action Models}},
author = {Yonghao Yu and Lang Huang and Runyi Li and Toshihiko Yamasaki},
year = {2026},
eprint = {2610.37219},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.37219},
}
Lets a VLA pick, at each step, between fine and temporally coarser action decoders: redundant motion runs faster while precision phases stay fine-grained. Tested on 3 sim benchmarks and 6 real tasks.让 VLA 在每一步自行选择细粒度或时间上更粗的动作解码器：冗余动作阶段加速执行，精细接触阶段保持细粒度控制；在 3 个仿真基准和 6 个真实任务上验证效率提升且成功率不降。
@misc{jeon2026pacevla,
title = {{PaceVLA: Adaptive Temporal Action Quantization in Vision-Language-Action Models}},
author = {Hojin Jeon and Jimin Lee and Taekwan Kim and Nayoung Oh and Taeyoung Kim and Myungkyu Koo and Jinmo Koo and Joonwoo Ahn and Jinwoo Shin},
year = {2026},
eprint = {2610.25961},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25961},
}
Xiaomi's compact embodied agent that plans, tracks progress and recovers from failures across manipulation and navigation, distilled from stronger teachers with a rubric-driven self-improvement loop.小米具身智能团队的轻量统一具身智能体：在操作与导航中负责任务理解、规划、进度跟踪和失败恢复，由更强的教师模型多教师在线蒸馏而来，并配有基于评分细则的自我改进数据飞轮。
@misc{team2026omnibrain,
title = {{OmniBrain-0: A Unified Agentic Model for Embodied Intelligence}},
author = {Xiaomi Embodied Intelligence Team and Yuechen Luo and Chong He and Fang Li and Shaoqing Xu and Jingkai Wang and Menglun Su and Jiaxin Liu and Shengyin Jiang and Zhiyi Lai and Long Chen and Longlong Wang and Liang Hu and Haochen Liu and Yuncheng Jiang and Hongwei Xie and Jieming Cui and Hanbing Li and Junjie Huang and Zhixiang Duan and Hexin Zhang and Weixiang Liang and Ying Gong and Yuhan Wang and Yong Pan and Pengna Li and Kangyi Wu and Ken Deng and Kun Ma and Fuxi Wen and Zhi-xin Yang and Hangjun Ye},
year = {2026},
eprint = {2610.36013},
archivePrefix = {arXiv},
primaryClass = {cs.CL},
url = {https://arxiv.org/abs/2610.36013},
}
Test-time add-on for frozen VLAs: a GNN retrieves future motion primitives from a graph of training demos and gates them into the actions. Reports up to +12% in sim and +40% real (GR00T N1.7, π0.5, StableVLA).面向冻结 VLA 的测试时插件：把训练示范组织成运动基元图，用 GNN 定位当前执行并检索后续基元，再按维度门控融合到预测动作中；报告在 GR00T N1.7、π0.5 等上仿真最高 +12%、真机 +40%。
@misc{quang2026retrieve,
title = {{Retrieve to Act: Motion Primitive Graph Retrieval for Robust VLA Execution}},
author = {Pham Tri Quang and Bao-Ngoc Dao and Minh Duc Nguyen and Duc-Duy Nguyen and Hung T. Ho and Linh Dang Le and Trong-Bao Ho and Thien-Loc Ha and Tan Q. Nguyen and Tuan Dam and Vu N. Duong and Daniel Sonntag and Ngan Le and Khoa D. Doan and Jan Peters and An Thai Le and Duy M. H. Nguyen and Vien Anh Ngo},
year = {2026},
eprint = {2610.37215},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.37215},
}
Finds VLAs over-rely on the wrist camera; predicts future interaction intent from past actions to route between global and wrist views per stage. Reports +10.1 pp over π0.5 on RoboTwin 2.0 and 71.7% → 85.0% real.发现现有 VLA 过度依赖腕部相机；根据历史动作预测未来交互意图，按任务阶段在全局与腕部视角间动态路由；报告在 RoboTwin 2.0 上比 π0.5 高 10.1 个百分点，真机成功率由 71.7% 提升到 85.0%。
S² is a training-free offline metric of how well a model's representation aligns with physical motion. Reports it tracks robot task success across model families and can rank VLM backbones before VLA fine-tuning.S² 是一种无需训练的离线指标，衡量模型视觉表征与物理运动几何的对齐程度（把柏拉图表征假说扩展到机器人）；报告它与多类模型的真实任务成功率正相关，可在微调成 VLA 之前筛选 VLM 骨干。
@misc{xie2026sensorimotor,
title = {{Sensorimotor Alignment: A Platonic Proxy for Embodied Policy Success}},
author = {Yunfei Xie and Howard H. Qian and Shiyi Lan and Kaiyu Hang and Chen Wei},
year = {2026},
eprint = {2610.31670},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.31670},
}
PAL has a high-level VLM write future-relevant subtasks into an executable memory while watching, then hand the current one to a low-level VLA. Reports 56.05% vs 44.51% on RoboMME and 60.83% on 4 real tasks.PAL 让高层 VLM 在观察过程中把未来需要的子任务提前写入可执行记忆（subtask_list），执行时再挑出当前子任务交给底层 VLA；报告在 RoboMME 上成功率 56.05%（最强基线 44.51%），4 个真实任务总体 60.83%。
@misc{liu2026plan,
title = {{Plan First, Act Later: Streaming Reasoning for Executable Memory in Robot Control}},
author = {Qinghang Liu and Conglin Wang and Zongsheng Li and Boyong He and Han Gao and Pengcheng Wang and Tao Sun and Lige Liu and Chang Liu},
year = {2026},
eprint = {2610.29329},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.29329},
}
Assigns VLA bit widths from short closed-loop task-success tests, not calibration sensitivity: ~4-bit weights on LIBERO across four models. Honest caveat: allocations transfer to a real arm for X-VLA but fail for π0.5.用短时闭环任务成功率（而非校准数据敏感度）为 VLA 各组件分配权重/激活位宽，并考虑组件间相互影响；在四个 VLA 与 LIBERO 全部套件上约 4 比特权重仍保持较强表现。作者如实指出：仿真选出的方案在真机上对 X-VLA 无损，对 π0.5 却失败。
@misc{xu2026quantize,
title = {{Quantize by Doing: Task-Guided Mixed-Precision Quantization of Vision-Language-Action Models}},
author = {Jiuyi Xu and Xiao Hu and Meida Chen and Song Wang and Yang Ye and Yangming Shi},
year = {2026},
eprint = {2610.38661},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.38661},
}
Humanoid VLA family pre-trained on 2,000+ h of egocentric human and robot data, predicting latent motion tokens for a whole-body controller; uses pseudo-proprioception and arm re-rendering to bridge the embodiment gap.人形机器人通用 VLA 系列：在 2,000 多小时第一视角人类与机器人数据上预训练，输出潜在动作 token 驱动预训练的全身控制器并附手部信号；通过仿真回放构造伪本体感知、重渲染手臂外观缩小人-机差异，报告 HumanoidArena 最优，并在 Unitree G1 上验证人类数据增益。
@misc{liu2026huge,
title = {{HuGe: Scaling Humanoid Generalist Vision-Language-Action Models with Egocentric Experience}},
author = {Yang Liu and Jiacheng Liu and Hengtao Li and Yuheng Huang and Zijian Yang and Ji Dai and Maoxin Wang and Yi Gao and Qishen Wang and Baohua Yin and Peiyang Li and Shuanghao Bai and Mingyang Sun and Pengxiang Ding and Jianqin Yin and Lixin Yang and Donglin Wang},
year = {2026},
eprint = {2610.34186},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.34186},
}
Paired test of previously proposed inference and data-collection tricks on all 100 BEHAVIOR-1K tasks with a π0.5 based on the 2025 challenge winner (Q-score 0.39): the tricks do not transfer uniformly.在 BEHAVIOR-1K 全部 100 个任务上，以基于 2025 BEHAVIOR 挑战赛冠军方案微调的 π0.5（公开测试集 Q 分 0.39）为对象，对已有推理与数据采集技巧做配对评测；结论是这些技巧在新环境与任务上的效果并不一致，提醒小规模结论不宜直接外推。
@misc{hammele2026what,
title = {{What Actually Matters? A Paired Evaluation of Inference Settings for π0.5 on BEHAVIOR-1K}},
author = {Ruben Hammele and Samira Huber and Mohammad Nazeri and Amirreza Payandeh and Alexandyr T. Card and Aaron Nguyen and Anuj Pokhrel and Daeun Song and Xuesu Xiao and Sören Pirk},
year = {2026},
eprint = {2610.37917},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.37917},
}
Rates VLA and world-action models on five axes (capability, training and deployment cost, verified research value, reproducibility), not success rate alone; withholds overall scores when evidence is missing.多维度评测 VLA 与世界-动作模型（WAM）：除任务能力外，还衡量下游训练成本、部署成本、经本地验证的研究价值和复现证据，按不同算力预算给出 0–100 画像；证据缺失时明确标注、不给总分，并开源评分代码与机器可读证据。
@misc{ma2026roboseed,
title = {{RoboSeed: Multidimensional Evaluation of Vision-Language-Action and World-Action Models: Capability, Resource Affordability, and Research Evidence}},
author = {Xin Ma and Furong Chen and Xinlong Man and Bo Zhang and Yalong Guo and Haolan Lei and Chaochen Cui and Hongyuan Zhang and Mulin Chen and Yi Wang and Lap-pui Chau and Chuang Yang},
year = {2026},
eprint = {2610.40026},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.40026},
}
Tests whether agent architecture search, so far used on text agents, transfers to perceptual embodied agents; AgentCanvas makes perception, memory, planning and action wiring searchable.研究原用于文本智能体的智能体架构搜索（AAS）能否迁移到有感知的具身智能体；提出 AgentCanvas，把感知、记忆、规划与动作模块的组合方式变成可搜索的设计空间。
@misc{zhou2026automating,
title = {{Automating the Design of Embodied Agent Architectures}},
author = {Jian Zhou and Sihao Lin and Jin Li and Shuai Fu and Gengze Zhou and Qi Wu},
year = {2026},
eprint = {2606.30111},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.30111},
}
FiberTune: action-supervised VLA fine-tuning can collapse visual structure the action loss barely constrains; a training-time objective keeps those residuals, with no added inference cost.FiberTune：动作监督微调会压缩动作损失约束较弱的视觉结构（沿局部动作纤维的残差坍塌）；提出训练期目标函数，对齐冻结的视觉教师以保留这些残差，不增加推理开销。
@misc{lin2026fibertune,
title = {{FiberTune: Preserving Action-Fiber Visual Residuals in Vision-Language-Action Fine-Tuning}},
author = {Haihao Lin and Xiangsheng Huang and Xiao Yang and Weibang Zhou and Yiqi Zhang and Bo Yang and Simin Zeng and Jiawei Yang and Zhengyang Wang and Jiahui Du},
year = {2026},
eprint = {2606.08653},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.08653},
}
Given an empty instruction, all six tested VLA checkpoints still moved in every LIBERO-Goal episode; five often finished an unrequested task. A 'do not move' prefix stopped none; an action-similarity gate held five.指令为空时，六个 VLA 检查点在 LIBERO-Goal 的每个回合都仍会启动，其中五个常常完成了未被要求的任务。给指令加“无任务就别动”前缀对六个都无效；比较首个动作块与空指令动作块的门控在五个上能让机器人保持不动。
@misc{jang2026vla,
title = {{VLA Policies Do Not Know When Not to Act: Unintended Task Execution and Start Suppression}},
author = {Minsu Jang and Jae-Woo Choi and Youngwoo Yoon and Dohyung Kim},
year = {2026},
eprint = {2610.41609},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.41609},
}
Evolves a frozen teacher's harness by recursive self-improvement with paired statistical tests, then distills a 4B student into it. Reports RoboTwin 33.4% → 63.6% for the teacher, 62.8% for the student.先冻结开放权重教师模型，用递归自我改进演化其 harness（每次修改须通过配对统计检验），再把 4B 学生模型蒸馏进同一 harness。报告 RoboTwin 上教师成功率由 33.4% 升到 63.6%，学生达 62.8%，加速器时间约为教师的五分之一。
@misc{gong2026tinyastral,
title = {{TinyAstral: Transferring Self-Evolving Harnesses to Tiny Generalist Manipulation Agents}},
author = {Kehong Gong and Xiaoxu Zheng and Guanli Hou and Seunghun Lee and Mengqi Guo and Runjia Tan and Zhengyu Wen and Qi Wang and Xiaoyu He and Mingyuan Zhang and Michael Bi Mi},
year = {2026},
eprint = {2610.41216},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.41216},
}
Runtime pairing a slow brain with fast policies under one rule: no running policy waits on the brain. Reports 95.9% over 3,000 LIBERO-Pro episodes vs 94.7% for the best single baseline, with 17% less control time.让慢速“大脑”（基础模型）与快速“小脑”（VLA、全身控制）协同的统一运行时，核心规则是运行中的策略绝不等待大脑。用路由记忆与边界触发的判定门替代逐次求助；报告 LIBERO-Pro 3000 回合成功率 95.9%（最佳单体基线 94.7%），控制时间减少 17%。
@misc{zhou2026rea,
title = {{REA: A Routed Embodied Agent Runtime for Asynchronous Brain-Cerebellum Control}},
author = {Pengfei Zhou and Yixing Ma and Zhengfeiyang Zhang and Haoyuan Chen and Zhenglin Wan and Hexin Wang and Wangbo Zhao and Yang You},
year = {2026},
eprint = {2610.42053},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.42053},
}
Turns a small open VLM into a code-as-policy agent: SFT, then multi-turn RL with a physical critic. Reports Qwen3.5-4B going from 0% to 73.8% on held-out LIBERO; code, data and weights released.把小型开放 VLM 训练成“代码即策略”机器人智能体的开源流水线：先 SFT，再做多轮 RL，并用读取仿真特权状态的物理评论家为每次请求分配贡献。报告 Qwen3.5-4B 在留出的 LIBERO 上由 0% 升到 73.8%，并公开代码、数据和权重。
@misc{fan2026captain,
title = {{CaPtain: A Pipeline for Post-Training Code-as-Policy Robot Agents}},
author = {Yijia Fan and Xilin He and Yilei Jiang and Dongming Wu and Jiaxu Wang and Xiangyu Yue},
year = {2026},
eprint = {2610.40799},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.40799},
}
A coding model reads rollouts of a frozen policy and rewrites its harness, with no weight updates. Reports average success rising from 9.7% to 43.3% on unseen scenes and 12.7% to 40.2% on unseen tasks.让编码模型观察冻结策略的执行轨迹、诊断失败并反复改写外围 harness，不更新任何权重。报告在未见场景上平均成功率从 9.7% 升到 43.3%，未见任务上从 12.7% 升到 40.2%。
@misc{wu2026roboarchitect,
title = {{RoboArchitect: Automating the Design of Robotic Agents with Persistent Memory}},
author = {Shengguang Wu and Suning Huang and Cherie Ho and Ria Doshi and Hao Zhu and Serena Yeung-Levy},
year = {2026},
eprint = {2610.42141},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.42141},
}
Pools robot data across six controller types by showing the policy examples of commanded actions and the motions they caused. Reports 33 points better average success than naive pooling in real experiments.跨六类控制器共享机器人数据：把“指令动作与实际运动”的交互样例编码为上下文，让同一策略在线适应。真实实验报告平均成功率比直接混合数据高 33 个百分点。
@misc{yang2026cross,
title = {{Cross-Controller Transfer via In-Context Behavior Adaptation}},
author = {Jonathan Yang and Chelsea Finn and Dorsa Sadigh},
year = {2026},
eprint = {2610.39280},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.39280},
}
CoRAct turns a VLA's stated movement and waypoints into checkable signals, then steers action noise toward agreeing with them, so actions align with the reasoning. Tested on RoboCasa, VLABench and real tasks.CoRAct 把 VLA 思维链中的运动意图与路点变成可验证信号，再用一致性奖励优化初始动作噪声，让动作真正对齐推理，而不只是以推理为条件。在 RoboCasa、VLABench 与真实任务上测试。
@misc{zhang2026from,
title = {{From Reasoning to Action: Consequence-Grounded Alignment for Vision-Language-Action Models}},
author = {Yixue Zhang and Wenke Xia and Kun Wu and Wenbo Yu and Pei Ren and Junnan Nie and Di Hu and Zhengping Che and Chang Liu},
year = {2026},
eprint = {2610.39543},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.39543},
}
Trains VLA models on 1,616 underwater episodes and grades 1,260 pool rollouts, best model 93.3% including partial successes. In a field case study at a WWII wreck, single-task ACT beat generalist SmolVLA 47.9% to 7.9%.用 1616 条水下数据训练 VLA，并在水池评估 1260 次，最佳模型成功率 93.3%（含部分成功）。在二战沉船实地案例中，单任务 ACT 以 47.9% 对 7.9% 明显胜过通用的 SmolVLA。
@misc{perezhaiek2026embodied,
title = {{Embodied at Sea: Imitation Learning for Multi-Task Underwater Manipulation}},
author = {Greta Perez-Haiek and Luca Macesanu and Kevin Angers and Mohammed Imam and Timothy Gambin and Christopher Clark},
year = {2026},
eprint = {2610.38199},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.38199},
}
@misc{jiang2026robottt,
title = {{RoboTTT: Context Scaling for Robot Policies}},
author = {Yunfan Jiang and Yevgen Chebotar and Ruijie Zheng and Fengyuan Hu and Yunhao Ge and Jimmy Wu and Tianyuan Dai and Scott Reed and Li Fei-Fei and Yuke Zhu and Linxi "Jim" Fan},
year = {2026},
eprint = {2607.15275},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.15275},
}
CoRL 2026 Oral. A steerable generalist foundation model: follows diverse instructions in unseen scenes and reaches zero-shot cross-embodiment tasks such as folding laundry, with no task-specific finetuning.CoRL 2026 Oral。可引导的通用基础模型：在未见环境中执行多样指令，跨形态零样本完成叠衣等任务，无需任务专用微调。
@misc{intelligence2026steerable,
title = {{$π_{0.7}$: a Steerable Generalist Robotic Foundation Model with Emergent Capabilities}},
author = {Physical Intelligence and Bo Ai and Ali Amin and Raichelle Aniceto and Ashwin Balakrishna and Greg Balke and Kevin Black and George Bokinsky and Shihao Cao and Thomas Charbonnier and Vedant Choudhary and Foster Collins and Ken Conley and Grace Connors and James Darpinian and Karan Dhabalia and Maitrayee Dhaka and Jared DiCarlo and Danny Driess and Michael Equi and Adnan Esmail and Yunhao Fang and Chelsea Finn and Catherine Glossop and Thomas Godden and Ivan Goryachev and Lachlan Groom and Haroun Habeeb and Hunter Hancock and Karol Hausman and Gashon Hussein and Victor Hwang and Brian Ichter and Connor Jacobsen and Szymon Jakubczak and Rowan Jen and Tim Jones and Gregg Kammerer and Ben Katz and Liyiming Ke and Mairbek Khadikov and Chandra Kuchi and Marinda Lamb and Devin LeBlanc and Brendon LeCount and Sergey Levine and Xinyu Li and Adrian Li-Bell and Vladislav Lialin and Zhonglin Liang and others},
year = {2026},
eprint = {2604.15483},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2604.15483},
}
CoRL 2026 Oral. Argues VLAs fail at dynamic tasks because one observation brings motion ambiguity and state aliasing, and adds learned temporal context to resolve both.CoRL 2026 Oral。指出 VLA 难以应对动态任务，是因为单帧观测带来运动歧义与状态混淆，并通过学习时序上下文同时解决这两个问题。
CoRL 2026 Spotlight. Gives VLAs memory at several timescales: abstract long-term state (which recipe stages are done) plus short-term recall of recent events and occluded objects.CoRL 2026 Spotlight。为 VLA 提供多时间尺度记忆：长期记住抽象状态（如菜谱进行到哪一步），短期记住近期事件与被遮挡物体。
@misc{torne2026mem,
title = {{MEM: Multi-Scale Embodied Memory for Vision Language Action Models}},
author = {Marcel Torne and Karl Pertsch and Homer Walke and Kyle Vedder and Suraj Nair and Brian Ichter and Allen Z. Ren and Haohuan Wang and Jiaming Tang and Kyle Stachowicz and Karan Dhabalia and Michael Equi and Quan Vuong and Jost Tobias Springenberg and Sergey Levine and Chelsea Finn and Danny Driess},
year = {2026},
eprint = {2603.03596},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.03596},
}
CoRL 2026 Spotlight. Similar prompts can make VLAs behave very differently, so this framework interactively searches for language sequences that raise closed-loop task performance.CoRL 2026 Spotlight。语义相近的指令可能让 VLA 行为迥异，该框架通过交互搜索语言序列来提升闭环任务表现。
@misc{jeong2026learning,
title = {{Learning What to Say to Your VLA: Mostly Harmless Vision Language Action Model Steering}},
author = {Hyun Joe Jeong and Gokul Swamy and Andrea Bajcsy},
year = {2026},
eprint = {2606.12299},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.12299},
}
CoRL 2026 Spotlight. Training-free steering that makes Cartesian, embodiment-agnostic policies aware of the deployed robot's body, using joint-space guidance for constraints such as whole-body collision avoidance.CoRL 2026 Spotlight。无需训练的引导框架：用关节空间引导让与本体无关的笛卡尔策略感知实际机器人本体，处理全身避碰等约束。
CoRL 2026 Spotlight. VLAs can ignore language and follow vision shortcuts from dataset bias. Introduces LIBERO-CF, a counterfactual benchmark for language following.CoRL 2026 Spotlight。VLA 可能忽略语言、沿用数据偏差形成的视觉捷径。论文提出反事实基准 LIBERO-CF，用于评测语言遵循能力。
@misc{fang2026when,
title = {{When Vision Overrides Language: Evaluating and Mitigating Counterfactual Failures in VLAs}},
author = {Yu Fang and Yuchun Feng and Dong Jing and Jiaqi Liu and Yue Yang and Zhenyu Wei and Daniel Szafir and Mingyu Ding},
year = {2026},
eprint = {2602.17659},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2602.17659},
}
CoRL 2026 Spotlight. A controlled study of zero-shot VLA transfer that isolates embodiment change from tasks, scenes and protocols, and separates strict zero-shot from looser definitions.CoRL 2026 Spotlight。对 VLA 零样本跨形态迁移的对照研究：将本体变化与任务、场景、流程隔离，并把严格零样本与宽松定义区分开。
@misc{yan2026zeta,
title = {{ZETA: A Controlled Study of Zero-Shot Cross-Embodiment VLA Transfer for Tabletop Manipulation}},
author = {Mi Yan and Wenhao Zhang and Zhiqi Zhang and Yu Peng and Tangxinyu Wang and Lingfei Zhai and Jiayi Su and Shengliang Deng and Lin Peng and Yaowei Liu and Yuxing Chen and Zhiyuan Wei and Jilong Wang and Jiayi Chen and Jiangran Lyu and Zhizheng Zhang and He Wang},
year = {2026},
eprint = {2609.02546},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.02546},
}
CoRL 2026 Spotlight. Splits the generalization gap of sequential instruction-following policies into marginal, compositional and context-action shift, viewed through instruction-space coverage.CoRL 2026 Spotlight。从指令空间覆盖出发，把序列任务策略的泛化差距分解为边际偏移、组合偏移与上下文-动作偏移。
@misc{wang2026diagnosing,
title = {{Diagnosing Compositional Generalization in Sequential Robot Tasks}},
author = {Yixiao Wang and Cheng-En Wu and Lingfeng Sun and Pengcheng Wang and Xiang Ji and Boyuan Liang and Guojian Zhan and Masayoshi Tomizuka},
year = {2026},
eprint = {2607.29687},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.29687},
}
CoRL 2026 Spotlight. A multi-agent self-learning harness aimed at Variational Automation, industrial tasks with more object variation than fixed automation, which need sustained reliability.CoRL 2026 Spotlight。面向“可变自动化”的多智能体自学习框架：这类工业任务的物体变化大于固定自动化，且要求持续可靠。
@misc{chen2026gap,
title = {{GaP: A Graph-as-Policy Multi-Agent Self-Learning Harness For Variational Automation Tasks}},
author = {Kaiyuan Chen and Shuangyu Xie and Letian Fu and Justin Yu and William Pacini and Sandeep Bajamahal and Hudson Kim and Jaimyn Drake and Daehwa Kim and Haoru Xue and Jonathan Francis and Christian Juette and Peter Schaldenbrand and Muhammet Yunus Seker and Ruwan Wickramarachchi and Uksang Yoo and Guanzhi Wang and Adithyavairavan Murali and Balakumar Sundaralingam and S. Shankar Sastry and Yuke Zhu and Linxi "Jim" Fan and Spencer Huang and Ken Goldberg},
year = {2026},
eprint = {2607.05369},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.05369},
}
CoRL 2026 Spotlight. RATs, teams of embodied coding agents, propose exploratory tasks and learn skills through self-directed play before any downstream task arrives.CoRL 2026 Spotlight。具身编程智能体团队 RATs 在下游任务到来之前，自行提出探索任务，并通过自主玩耍学习技能。
@misc{zhang2026playful,
title = {{Playful Agentic Robot Learning}},
author = {Junyi Zhang and Jiaxin Ge and Hanjun Yoo and Letian Fu and Zihan Yang and Yaowei Liu and Raj Saravanan and Shaofeng Yin and Justin Yu and Dantong Niu and Zirui Wang and Roei Herzig and Ken Goldberg and Yutong Bai and David M. Chan and Ion Stoica and Angjoo Kanazawa and Jiahui Lei and Haiwen Feng and Trevor Darrell},
year = {2026},
eprint = {2606.19419},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.19419},
}
Adapts a frozen pre-trained policy with residual RL instead of new demos, then merges compatible residuals to cut stored parameters; tested on LIBERO and a real Franka.在冻结的预训练策略上用残差强化学习适配新任务（无需新演示），再合并相容的残差以减少参数；在 LIBERO 与真实 Franka 上验证。
@misc{zhu2026resmerge,
title = {{ResMerge: Residual Policy Learning and Merging for Continual Adaptation of Pre-trained Robot Policies}},
author = {Kefei Zhu and Lingyun Xu and Yue Xin and Xiaoqi Li and Caixin Guo and Xincheng Du and Yuanpei Chen and Yaodong Yang and Hao Dong},
year = {2026},
eprint = {2610.41663},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.41663},
}
Training-free tweak for pre-trained VLAs: an EMA over late-layer updates keeps object-aware directions near the action output, mixed less when proprioception shows fast dynamics; reports +4.0 pts success.面向预训练 VLA 的免训练方法：对后层残差更新做 EMA，让靠近动作输出的表征保留物体感知方向，并按近期本体感知状态调节混合强度（动态变化快时更保守）；报告平均任务成功率提升 4.0 个百分点。
@misc{zhao2026star,
title = {{STAR: Steering Action Representations with Proprioception for VLA Inference}},
author = {Guangyu Zhao and Yongchang Zhang and Haoxuan Ru and Kaichen He and Haoran Wen and Kun Zhan and Zeyu Zhang and Yin Li and Yiwu Zhong},
year = {2026},
eprint = {2610.41009},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.41009},
}
Jointly predicts short-term states, semantic milestones and goal states across horizons, then lets the action query retrieve only the relevant parts of that imagined future.跨时间尺度联合预测短期状态、语义里程碑与目标状态，再由动作查询从中检索相关的未来表征。
@misc{ding2026hifor,
title = {{HiFOR: Hierarchical Foresight with Action-Guided World Grounding for Vision-Language-Action Model}},
author = {Rongxing Ding and Hongyu Qu and Kehong Wu and Xudong Wang and Fan Zhang and Xinguang Xiang and Xiangbo Shu},
year = {2026},
eprint = {2610.40373},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.40373},
}
Keeps model weights fixed and adapts during execution: an agent diagnoses, intervenes and stores linked experience records, revising how the policy is used rather than the policy.不改模型权重，在执行中适配：智能体负责诊断、干预并保存关联的经验记录，修订的是策略的使用方式而非策略本身。
@misc{yu2026robodoctor,
title = {{RoboDoctor: Training-Free Recursive Self-Improvement for Robot Policies}},
author = {Heyang Yu and Haomin Li and Sihang Li and Jixin Yan and Shiyun Peng and Wanqi Zhong and Tianyu Xin and Yuan Yao and Xiaoyu Tian and Yiming Li},
year = {2026},
eprint = {2610.35604},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.35604},
}
uni-robot-agent: an open harness for robot-use agents on open-weight models. It reports a RoboDojo score of 31.39 with an off-the-shelf Qwen3.8 model, above the published GPT-6 Astra-RoboProbe result of 28.97.uni-robot-agent：面向开放权重模型的机器人操作智能体（Robot-use Agent）框架，便于继续训练。作者报告现成的 Qwen3.8 模型在 RoboDojo 上得 31.39 分，高于已发表的 GPT-6 Astra-RoboProbe 的 28.97 分。
@misc{yuan2026democratizing,
title = {{Democratizing Robot-Use Agents: A Unified Harness with Open Models}},
author = {Jiarui Yuan and Yuxin Zuo and Xiaoyu Huang and Wenbo Zhang and Kwanyoung Park and Junlin Yang and Tianxiang Zhang and Ruoqu Chen and Yutao Ouyang and Bingxiang He and Xingtai Lv and Hongjin Wu and Jiajun Liu and Jiahao Zhao and Tianwei Luo and Haozhan Li and Cheng Qian and Jiaxin Ge and Huan-ang Gao and Kaisen Yang and Jiaqi Deng and Qixin Xu and Siqiao Huang and Bowen Zhou and Mengdi Xu and Zhiyuan Liu and Pieter Abbeel and Ning Ding},
year = {2026},
eprint = {2611.01560},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.01560},
}
The LLM decides; the harness adds calibrated 3D measurements, trajectory previews and stop-on-divergence checks. Reports first place on RoboDojo with 57.7% success, over double the reference harness.任务决策交给大模型，框架提供标定后的三维测量、轨迹预览和偏差即停的执行检查。作者报告在 RoboDojo 榜单排名第一，成功率 57.7%，是同模型参考框架的两倍多。
@misc{wang2026auriga,
title = {{Auriga Technical Report: A Task-Agnostic Agent Harness for Zero-Shot Embodied Manipulation}},
author = {Chengcheng Wang and Yuheng Shi and Siyu Xu and Yutong Wang and Jinjing Zhao and Chenghao Xia and Jinqi Liao and Jiankun Peng and Ligong Bi and Yuechuan Ma and Jiaqi Liu and Wenbo Wang and Zijian Wang and Yuemin Wu and Minjing Dong and Xiu Su and Tao Huang and Jianyuan Guo and Chengbin Du and Yunke Wang and Chang Xu},
year = {2026},
eprint = {2611.01112},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.01112},
}
Benchmark of VLA latency sensitivity across tasks, object speeds and execution modes: 100 ms can nearly halve success on moving targets. An empirical model then predicts success from limited calibration data.评测 VLA 对推理延迟的敏感度，覆盖不同任务、物体速度与同步/异步执行：100 ms 延迟就可能让移动目标任务的成功率几乎减半。作者还给出经验模型，用少量标定数据预测成功率。代码尚待发布。
@misc{wen2026vlalagbench,
title = {{VLALagBench: Understanding When and Why VLA Inference Latency Matters}},
author = {Qi Wen and Borui Xu and Hussein Amro and Hongshi Tan and Hanmin Lin and Chaoyi Ruan and Shanghao Liu and Gang Yang and Yuejiang Liu and Jiafei Duan and Bingsheng He and Jason Clemons and Christos Kozyrakis and Wenqi Jiang},
year = {2026},
eprint = {2611.03103},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.03103},
}
Gives a VLA working memory by keeping task history as explicit text, updated only when events are detected. Tested on 10 real non-Markovian tasks, including ones that need context across episodes.用显式文本保存任务历史、只在检测到事件时更新，为 VLA 增加工作记忆。在 10 个真实的非马尔可夫任务上测试，其中包括需要跨回合保持上下文的任务。
@misc{kim2026helm,
title = {{HeLM: Hierarchical Explicit Language Memory for Non-Markovian Manipulation}},
author = {Gyeonghun Kim and Donghoon Kim and Minsoo Choi and Suyun Lee},
year = {2026},
eprint = {2611.02660},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.02660},
}
JEPA-style VLA pretraining: future frames are only a latent target and the student sees the current observation alone, so it avoids leakage and learns dynamics rather than pixel changes; then fine-tune an action head.JEPA 式 VLA 预训练：未来帧只作为潜空间监督目标，学生网络只看当前观测，避免信息泄漏，学动力学而非像素变化；之后再微调动作头。
@misc{sun2026vla,
title = {{VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Model}},
author = {Jingwen Sun and Wenyao Zhang and Zekun Qi and Shaojie Ren and Zezhi Liu and Hanxin Zhu and Guangzhong Sun and Xin Jin and Zhibo Chen},
year = {2026},
eprint = {2602.10098},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.10098},
}
Shows reaction time hinges on time to first action, then uses a schedule that spends denoising on near-term actions: one step for the immediate reaction, about 10× faster for π0.5 and X-VLA; demoed on table tennis.指出反应速度取决于首动作耗时，并让去噪优先照顾近期动作：即时反应只需一步，π0.5 和 X-VLA 上约快 10 倍；真机演示包括乒乓球。
@misc{lu2026faster,
title = {{FASTER: Rethinking Real-Time Flow VLAs}},
author = {Yuxiang Lu and Zhe Liu and Xianzhe Fan and Zhenya Yang and Jinghua Hou and Junyi Li and Kaixin Ding and Hengshuang Zhao},
year = {2026},
eprint = {2603.19199},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.19199},
}
Splits a large VLM that slowly updates a semantic memory from a light expert that acts at high rate, trained on staleness and delay; reports running the VLM at a tenth of its nominal rate with no significant loss.把低频更新语义记忆的大型 VLM 与高频出动作的轻量专家解耦，并针对陈旧和延迟训练；报告 VLM 降到标称频率十分之一时性能无明显损失。
Shows visual perturbations can change a reasoning VLA's action while its rationale stays word-for-word identical, tested on Alpamayo and InstructVLA, so monitors must check the action channel, not the text.展示视觉扰动可以改变推理型 VLA 的动作，而解释文本一字不变（在 Alpamayo 和 InstructVLA 上测试），因此监控必须直接检查动作通道，不能只看文本。
@misc{feng2026say,
title = {{Say the Same, Act Differently: Diagnosing and Steering Text-Action Divergence in Reasoning VLAs}},
author = {Zihao Feng and Qingzhao Zhang and Chunyu Xia and Bo Yu and Z. Morley Mao and Ramesh Govindan},
year = {2026},
eprint = {2611.04166},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2611.04166},
}
DySta splits visual input into static and dynamic tokens, keeping one copy of the static tokens and reusing their KV cache, to extend VLA context and speed up inference.DySta 把视觉输入分成静态和动态 token，静态部分只保留一份并复用 KV 缓存，从而延长 VLA 上下文并加快推理。
@misc{qiu2026efficient,
title = {{Efficient Long-Horizon Vision-Language-Action Models via Static-Dynamic Disentanglement}},
author = {Weikang Qiu and Huashuo Lei and Tinglin Huang and Rex Ying},
year = {2026},
eprint = {2602.03983},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.03983},
}
Moves routine spatial reasoning from images to text: a locally refreshed 3D symbolic scene the language model reads directly, with vision only on demand; reports better multi-step completion and far fewer image queries.把常规空间推理从图像转到文本：语言模型直接读取局部刷新的三维符号场景，仅在需要时看图；报告多步任务完成度提高，图像查询大幅减少。
@misc{liu2026do,
title = {{Do Agentic Policies Need to Look at Every Step? Geometric Abstraction for Reasoning and Action}},
author = {Yihang Liu and Yiming Chen and Mingzhe Zheng},
year = {2026},
eprint = {2611.05058},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05058},
}
Finds that fine-tuning erodes the semantic structure VLAs inherit, and that this tracks task failure. A training-only anchor restores it, reporting up to +21.5% on out-of-distribution tasks.发现微调会破坏 VLA 继承的语义结构，且这与任务失败同步。提出仅在训练时使用的语义锚定方法，报告分布外任务提升最高 21.5%。
@misc{xu2026semantic,
title = {{Semantic Anchoring for Robotic Action Representations}},
author = {Yuan Xu and Youheng Shi and Chengyang Li and Wentao Zhu and Yizhou Wang},
year = {2026},
eprint = {2607.13597},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.13597},
}
Reframes VLA pretraining as goal-conditioned RL: contrastive learning gives the model a sense of whether a language goal is reachable. Pretrained on 167B tokens; strongest on long-horizon tasks.将 VLA 预训练重构为目标条件强化学习：用对比学习让模型判断语言目标是否可达。基于 1670 亿 token 预训练，在长时序任务上优势最明显。
@misc{zhang2026prts,
title = {{PRTS: A Primitive Reasoning and Tasking System via Contrastive Representations}},
author = {Yang Zhang and Jiangyuan Zhao and Chenyou Fan and Fangzheng Yan and Tian Li and Haitong Tang and Sen Fu and Xuan'er Wu and Qizhen Weng and Weinan Zhang and Xiu Li and Chi Zhang and Chenjia Bai and Xuelong Li},
year = {2026},
eprint = {2604.27472},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2604.27472},
}
Constrained decoding for LLM planners: grounded 'meta-tokens' guarantee syntax and preconditions by construction and cut decoder memory from over 107 TB to under 2 GB.面向 LLM 规划的约束解码：用接地的“元 token”从结构上保证语法与前置条件，并将解码内存从 107 TB 以上降到 2 GB 以内。
@misc{yidouweng2026meta,
title = {{Meta-Ctrl: Guaranteed Plan Generation by Decoupling Syntactic and Semantic Constraints}},
author = {Gwen Yidou-Weng and Edward Sun and Tianyi Ma and Metin Alp Dogan and Benjie Wang and Allen Peng and Guy Van den Broeck and Yuchen Cui},
year = {2026},
eprint = {2608.22149},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.22149},
}
Treats clarifying questions as Bayesian active learning over Signal Temporal Logic specs, so the robot asks fewer questions and hands a formal planner a verifiable task.把澄清提问视为信号时序逻辑规约上的贝叶斯主动学习，机器人提问更少，并向形式化规划器交付可验证的任务。
@misc{li2026bayesian,
title = {{Bayesian Active Learning for Intent Disambiguation in Interactive Robot Planning}},
author = {Huao Li and Carson Sobolewski and Augustinos Saravanos and William Tan and John Karigiannis and Chuchu Fan},
year = {2026},
eprint = {2609.34270},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.34270},
}
An open foundation model for humanoid loco-manipulation that argues against co-training on mixed human and humanoid data because of their kinematic differences.面向人形移动操作的开放基础模型，主张因运动学差异，不宜直接混合人类与人形数据共同训练。
@misc{wei2026open,
title = {{$Ψ_0$: An Open Foundation Model Towards Universal Humanoid Loco-Manipulation}},
author = {Songlin Wei and Hongyi Jing and Boqian Li and Zhenyu Zhao and Jiageng Mao and Zhenhao Ni and Sicheng He and Jie Liu and Xiawei Liu and Kaidi Kang and Sheng Zang and Weiduo Yuan and Marco Pavone and Di Huang and Yue Wang},
year = {2026},
eprint = {2603.12263},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.12263},
}
Fourth-place MetaFine Challenge entry: a task router picks between adapted DINOv2-DiT and π0.5 policies behind one interface, a practical recipe for adapting pretrained policies to precise tasks.MetaFine 挑战赛第四名方案：任务路由在统一接口下选择微调后的 DINOv2-DiT 或 π0.5 策略，是把预训练策略适配到精细任务的实用做法。
@misc{zhang2026finecraft,
title = {{FineCraft: Enhancing VLA Models for Fine-Grained Manipulation}},
author = {Shuaike Zhang and Ziheng Quan and Junhao Ran and Muhe Ding and Shaokun Wang and Jianlong Wu},
year = {2026},
eprint = {2611.03163},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.03163},
}
RDT2 builds a robot foundation model on a 7B-parameter VLM and tests how far UMI data scales toward zero-shot deployment on new embodiments.RDT2 基于 70 亿参数 VLM 构建机器人基础模型，探索 UMI 数据规模能把零微调跨形态部署推到多远。
@misc{liu2026rdt,
title = {{RDT2: Exploring the Scaling Limit of UMI Data Towards Zero-Shot Cross-Embodiment Generalization}},
author = {Songming Liu and Bangguo Li and Kai Ma and Lingxuan Wu and Hengkai Tan and Xiao Ouyang and Hang Su and Jun Zhu},
year = {2026},
eprint = {2602.03310},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.03310},
}
ABot-M0 pairs a data curation pipeline with action manifold learning for one model across hardware, the "one brain, many forms" goal.ABot-M0 结合数据整理流程与动作流形学习，让一个模型跨不同硬件工作，即“一脑多形”。
@misc{yang2026abot,
title = {{ABot-M0: VLA Foundation Model for Robotic Manipulation with Action Manifold Learning}},
author = {Yandan Yang and Shuang Zeng and Tong Lin and Xinyuan Chang and Dekang Qi and Junjin Xiao and Haoyun Liu and Ronghan Chen and Yuzhi Chen and Dongjie Huo and Feng Xiong and Xing Wei and Zhiheng Ma and Mu Xu},
year = {2026},
eprint = {2602.11236},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2602.11236},
}
Qwen-VLA studies whether manipulation, navigation and other decision tasks can share one VLA model across tasks, environments and robots.Qwen-VLA 研究操作、导航等决策任务能否由同一 VLA 模型统一，跨任务、环境和机器人形态。
@misc{wang2026qwen,
title = {{Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments}},
author = {Qiuyue Wang and Mingsheng Li and Jian Guan and Jinhui Ye and Sicheng Xie and Yitao Liu and Junhao Chen and Zhixuan Liang and Jie Zhang and Xintong Hu and Xuhong Huang and Pei Lin and Junyang Lin and Dayiheng Liu and Shuai Bai and Jingren Zhou and Jiazhao Zhang and Haoqi Yuan and Gengze Zhou and Hang Yin and Ye Wang and Yiyang Huang and Zixing Lei and Wujian Peng and Delin Chen and Yingming Zheng and Jingyang Fan and Xianwei Zhuang and Xin Zhou and Haoyang Li and Anzhe Chen and Tong Zhang and Xuejing Liu and Yuchong Sun and Ruizhe Chen and Zhaohai Li and Chenxu Lü and Zhibo Yang and Tao Yu and Xionghui Chen},
year = {2026},
eprint = {2605.30280},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2605.30280},
}
CoRL 2026 workshop paper: a planner picks subtasks, a π0.5 policy acts, and a small fine-tuned monitor decides when to call the planner again; reports 79.13% on RoboMME with about 3.6 planner calls per episode.CoRL 2026 研讨会论文：大模型规划子任务，π0.5 策略执行，小型微调监视器判断何时再次调用规划器；报告在 RoboMME 上成功率 79.13%，每回合约调用规划器 3.6 次。
Coding agents keep real-robot experience as memory and reusable code skills, refining them through physical trials; reports 91.1% success on eight real tasks and 84.4% on 42 RoboDojo tasks.让编码智能体把真机交互经验存为记忆和可复用的代码技能，并在实物试验中持续改进；报告八个真实任务成功率 91.1%，RoboDojo 的 42 个任务成功率 84.4%。
@misc{wang2026pyrua,
title = {{PyRUA-REAL: Robot Exploration and Adaptive Learning through RSI in the Real World}},
author = {Qiang Wang and Jianxin Bi and Shulong Jiang and Ruiyang Si and Wenbo Huang and Kaiqi Chen and Jiaxin Li and Liyun Li and Xuanyi Li and Daquan Zhou},
year = {2026},
eprint = {2611.06768},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.06768},
}
World-language-action model: one autoregressive backbone predicts text subtasks, subgoal images and actions; world prediction can be off at inference or on for test-time scaling. Reports 92.94% on RoboTwin 2.0 Clean.世界-语言-动作模型：单一自回归主干同时预测文本子任务、子目标图像与动作；推理时可关闭世界预测，也可开启做测试时扩展。报告在 RoboTwin 2.0 Clean 上成功率 92.94%。
@misc{yang2026world,
title = {{World-Language-Action Model for Unified World Modeling, Language Reasoning, and Action Synthesis}},
author = {Yi Yang and Zhihong Liu and Siqi Kou and Yiyang Chen and Yanzhe Hu and Jianbo Zhou and Boyuan Zhao and Zhijie Wei and Xiao Xia and Xueqi Li and Pengfei Liu and Zhijie Deng},
year = {2026},
eprint = {2606.05979},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.05979},
}
Memory for VLAs that keeps only sparse keyframes the policy itself predicts will matter later. Adds the RoboTwin-MeM benchmark; reports +40% average success over memory-augmented VLAs across 17 sim and 4 real tasks.为 VLA 加入稀疏记忆：由策略自己预测哪些关键帧之后会有用并只保存这些帧。同时提出 RoboTwin-MeM 基准；报告在 17 个仿真和 4 个真机任务上平均成功率比带记忆的 VLA 高 40%。
@misc{yang2026eventvla,
title = {{EventVLA: Event-Driven Visual Evidence Memory for Long-Horizon Vision-Language-Action Policies}},
author = {Ganlin Yang and Zhangzheng Tu and Yuqiang Yang and Sitong Mao and Junyi Dong and Tianxing Chen and Jiaqi Peng and Jing Xiong and Jiafei Cao and Jifeng Dai and Wengang Zhou and Yao Mu and Tai Wang},
year = {2026},
eprint = {2606.20092},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.20092},
}
Converts human videos from five sources into robot-aligned observations and actions (about 630K episodes). Pretraining a VLA on more of it raises real-task completion from 51.5% to 80.3% and OOD from 34.9% to 72.2%.把五个来源的人类视频转换成与机器人对齐的观测和动作（约 63 万条轨迹）。报告用更多此类数据预训练 VLA，真机任务完成率从 51.5% 升到 80.3%，分布外从 34.9% 升到 72.2%。
@misc{jeong2026huro,
title = {{HuRo: Robotizing Human Videos for Scalable VLA Pretraining}},
author = {Jinho Jeong and Se June Joo and Jaehyun Kang and Dongyun Kim and Yena Kim and Hanjung Kim and Seon Joo Kim},
year = {2026},
eprint = {2609.10706},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.10706},
}
Joint VLA pretraining on 4.53K h of robot/sim data plus 1.48K h of egocentric human video with pseudo-actions, camera-space actions and a reliability-weighted loss. Reports SOTA on RoboCasa GR1 TableTop and RoboTwin 2.0.用 4530 小时机器人/仿真数据加 1480 小时带伪动作的第一视角人类视频联合预训练 VLA，采用相机空间动作与按可靠性加权的损失。报告在 RoboCasa GR1 TableTop 和 RoboTwin 2.0 上达到最优。
@misc{li2026ace,
title = {{ACE-Ego-0: Unifying Egocentric Human and Robotic Data for VLA Pretraining}},
author = {Hao Li and Ganlong Zhao and Yufei Liu and Haotian Hou and Guoquan Ye and Tongyan Fang and Chunxiao Liu and Siyuan Huang and Jianbo Liu and Xiaogang Wang and Hongsheng Li},
year = {2026},
eprint = {2606.17200},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.17200},
}
Pretrain the action expert first as a language-free vision-action prior, then inject language through gated fusion. Targets VLAs that ignore unseen instructions; works with π-style and GR00T-style designs.先把动作专家预训练为不含语言的视觉-动作先验，再通过门控融合注入语言。针对 VLA 对未见指令泛化差的问题，适用于 π 系和 GR00T 系架构。
PAINT keeps action chunks consistent under async execution by picking the initial noise (backward Euler inversion), not by guiding the flow. No training; reports RTC-level task success with better prefix consistency.PAINT 通过选择初始噪声（后向欧拉反演）而非引导流场，让异步执行时的动作块首尾一致；无需训练，报告任务成功率与 RTC 相当、前缀一致性更好。
@misc{ho2026start,
title = {{Start Right, Arrive Right: Asynchronous Execution via Initial Noise Selection}},
author = {Trong-Bao Ho and Quang-Tan Nguyen and Thien-Loc Ha and Gia-Binh Nguyen and Viet-Thanh Nguyen and Long Dinh and Minh N. Vu and Duy M. H. Nguyen and An Thai Le and Ngo Anh Vien},
year = {2026},
eprint = {2606.19774},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.19774},
}
Latent actions with polar structure in hyperbolic space: radius encodes how much changes, direction encodes what kind. Reports gains over latent-action baselines and pretrained VLAs.在双曲空间中构造极坐标结构的潜动作：半径表示变化多大，方向表示变化类型。报告优于其他潜动作方法和预训练 VLA。
@misc{jeong2026polar,
title = {{PoLAR: Factorizing Extent and Mode in Latent Actions for Robot Policy Learning}},
author = {Youngjoon Jeong and Jihwan Yu and Minsoo Jo and Junha Chun and Taesup Kim},
year = {2026},
eprint = {2606.21139},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.21139},
}
Continual pre-training of π0.5 that aligns intermediate visual features with a frozen 3D perception model across many robot datasets. The extra parts are training-only, so inference and weights stay π0.5-compatible.在多个机器人数据集上对 π0.5 做持续预训练，把中间视觉特征与冻结的 3D 感知模型对齐。附加组件只在训练时使用，推理结构和权重与 π0.5 保持兼容。
@misc{li2026sla,
title = {{SLA: Scaling Spatial-Language-Action Model via Continual Pre-training}},
author = {Fuhao Li and Wenxuan Song and Qizhou Huang and Ziyang Zhou and Jingbo Wang and Junwu E and Pengwei Wang and Yan Wang and Shunbo Zhou and Haoang Li},
year = {2026},
eprint = {2611.08373},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08373},
}
Puts continuous actions inside one diffusion VLM: each timestep is a latent token, no separate action expert (0.21% of params for action I/O). Reports 98.15% on LIBERO and 85.9% on LIBERO-Plus.把连续动作直接放进同一个扩散 VLM：每个控制时间步是一个潜动作 token，没有独立动作专家（动作接口仅占 0.21% 参数）。报告 LIBERO 98.15%、LIBERO-Plus 85.9%。
@misc{wang2026concertina,
title = {{Concertina: Embodying Manipulation and Cognition in One Diffusion Vision-Language Model}},
author = {Chengcheng Wang and Wei He and Wenhao Li and Hongguang Li and Kai Han and Jianyuan Guo and Chang Xu},
year = {2026},
eprint = {2611.07513},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.07513},
}
Lightweight probe on a VLA's latent layers that flags failures early with calibrated miss and false-alarm bounds and names the failure mode. Releases 3,180 annotated manipulation episodes.在 VLA 隐层上加轻量探针，提前预警失败并给出校准的漏报与误报上界，还能判断失败类型。并发布 3180 条带标注的操作轨迹。
@misc{podolinsky2026failure,
title = {{Failure Prediction and Classification in Vision-Language-Action Models Via Calibrated Latent Probing}},
author = {Michal P. Podolinsky and Yunhao Yang and Daniel Little and Hong Lu and Rohan Siva and Olin Goudey and Neel P. Bhatt and Matthias Scheutz and Ufuk Topcu},
year = {2026},
eprint = {2611.08425},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08425},
}
Weights each action chunk by how much it shortens the path to success on a reachability graph built over demos, rollouts and interventions. Reports 91% on precise tasks after a few offline rounds; no reward model.在示教、rollout 与人工干预构成的可达图上，按每个动作块缩短到成功的步数来加权行为克隆。报告经几轮离线迭代在精细任务上达到 91%，无需奖励模型。
@misc{li2026graph,
title = {{Graph-Reweighted Behavior Cloning with Human Interventions for Precise Robotic Manipulation}},
author = {Xiaoman Li and Zhiqiang Yang and Jiayu Hu and Jipeng Kong and Yang Zhang and Jingwei Zhang and Fangzheng Yan and Chunhua Shen and Chenjia Bai},
year = {2026},
eprint = {2610.20504},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.20504},
}
@misc{kim2026disentangling,
title = {{Disentangling Spurious Correlations in Vision-Language-Action Models via Predicting Domain-Invariant Latent Lookahead}},
author = {Junghyun Kim and Ngseo Kim and ChungWoo Lee and Seoyeon Lee and Woo-Jeong Baek and Adam Zhou and Chip Huyen and Jun-Ki Lee and Gi-Cheon Kang and Byoung-Tak Zhang},
year = {2026},
eprint = {2609.37165},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.37165},
}
Keeps the last denoising step of a diffusion VLA as a fast feedback layer, so each action is corrected with the newest image. Reports dynamic sim tasks 27.5% to 85.0% and real tasks 51% to 73%.把扩散 VLA 的最后一步去噪保留为快速反馈层，用最新图像修正每个动作。报告动态仿真任务从 27.5% 升到 85.0%，真机任务从 51% 升到 73%。
@misc{ji2026catch,
title = {{Catch Me If You Can: Real-Time Feedback Denoising for Responsive VLAs}},
author = {Yiheng Ji and Xingru Zhou and Luis Sentis and Mingyo Seo},
year = {2026},
eprint = {2609.21022},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.21022},
}
VLA that lifts VLM features into 3D scene tokens, predicts how scene points will move, and conditions a 3D flow action denoiser on that motion. Reports SOTA on CALVIN without robot-action pretraining.把 VLM 特征提升为 3D 场景 token，预测场景点将如何运动，并以此为条件驱动 3D 流匹配动作去噪器。报告在无机器人动作预训练下 CALVIN 达到最优。
@misc{xiong2026geomvla,
title = {{GeomVLA: Unifying Scene, Motion, and Action in 3D}},
author = {Ziyin Xiong and Nikolaos Gkanatsios and Moritz Reuss and Katerina Fragkiadaki},
year = {2026},
eprint = {2609.13812},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.13812},
}
A task-manager VLM re-predicts the remaining subtasks and a 2D keypoint trace at every step; the VLA executor just follows the trace. Failed steps stay in the plan, so replanning comes for free. Sim + real Franka.由任务管理 VLM 每一步重新预测剩余子任务和一条 2D 关键点轨迹，VLA 执行器只需沿轨迹执行。失败的步骤会留在计划里，因此无需额外逻辑即可自动重规划。仿真与真实 Franka 实验。
@misc{liu2026long,
title = {{Long-Horizon Manipulation via Trace-Conditioned VLA Planning}},
author = {Isabella Liu and An-Chieh Cheng and Rui Yan and Geng Chen and Ri-Zhao Qiu and Xueyan Zou and Sha Yi and Hongxu Yin and Xiaolong Wang and Sifei Liu},
year = {2026},
eprint = {2604.21924},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2604.21924},
}
Fights shortcut learning when fine-tuning VLAs: predicted masks of task-relevant objects and the robot serve as an auxiliary grounding loss; nothing changes at inference. Tested on four foundation models and π0.5.针对 VLA 微调中的捷径学习：预测任务相关物体与机器人的掩码，作为辅助对齐损失引导策略关注正确区域，推理时不增加任何开销。在四个基础模型和真机 π0.5 上验证。
@misc{sun2026artificial,
title = {{Artificial Foveated Perception for Mitigating Shortcut Learning in Robotic Foundation Models}},
author = {Xiatao Sun and Yuan Zhuang and Mateo Sanchez Lopez Negrete and Matei-Victor Coldea and Chen Liang and Haoyang Zhang and Che Liu and Ziyao Zeng and Shawn Li and Qian Wang and Fei Miao and Daniel Rakita},
year = {2026},
eprint = {2607.10655},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.10655},
}
Shows VLAs can do implicit human-robot collaboration, but action chunks that cross task transitions make the robot help too early; an inference-time steering fix is tested in a 16-person assembly study.表明 VLA 可用于隐式人机协作，但跨越任务切换点的动作块会让机器人过早提供帮助；提出推理时引导方法缓解，并在 16 人装配用户研究中验证。
@misc{xu2026learning,
title = {{Learning to Assist: Collaborative VLAs for Implicit Human-Robot Collaboration}},
author = {Leo Xu and Letian Li and Alex Cuellar and Michael Hagenow},
year = {2026},
eprint = {2606.12475},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.12475},
}
Action tokenizer whose first quantization level is trained as a semantic 'action language' while deeper levels keep detail; pretrained on 2.4M trajectories. Reports gains over FAST on grounding and long-horizon tasks.动作 tokenizer：残差量化的第一层被训练成表达粗粒度运动意图的语义“动作语言”，更深层保留细节；在 240 万条轨迹上预训练。报告在多模态对齐和长程任务上优于 FAST。
@misc{kang2026x,
title = {{X-Tokenizer: A Multimodal Action Tokenizer for Vision-Language-Action Pretraining}},
author = {Miracle Kang and Lights Shi and Lucy Liang and Roy Gan and Dongxiu Liu and Pushi Zhang and Sylas Chen and Shawn Qin and Yinan Zheng and Jinliang Zheng and Hao Wang and Xianyuan Zhan and Hang Su},
year = {2026},
eprint = {2606.14752},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.14752},
}
Trains the VLA executor with refined subtask language plus an explicit budget on how much visual evidence it may use. Reports mean subtask success up from 54.2% to 79.0% over π0.5 on eight real tasks.用更细的子任务语言重标注轨迹，并对执行器可使用的视觉证据施加显式预算来训练 VLA。报告在八个真机任务上平均子任务成功率比 π0.5 从 54.2% 提升到 79.0%。
Aligns robot visual features with human ones using paired human-robot demos, then pretrains a latent-action VLA on human and robot videos. Reports 4.481 average length on CALVIN ABC→D.利用成对的人-机演示把机器人视觉表征与人类表征对齐，再用人类和机器人视频预训练潜在动作 VLA。报告 CALVIN ABC→D 平均长度 4.481。
One VLA backbone runs as an independent copy on each robot, conditioned on its own view and a robot ID prompt, for decentralized team tasks. Reports a 64-point gain over from-scratch decentralized models.同一个 VLA 主干在每台机器人上独立运行，只依赖本机观测和机器人身份提示，实现去中心化多机协作。报告比从零训练的去中心化模型提升 64 个百分点。
@misc{doshi2026chorus,
title = {{CHORUS: Decentralized Multi-Embodiment Collaboration with One VLA Policy}},
author = {Ria Doshi and Tian Gao and Annie Chen and Chelsea Finn and Jeannette Bohg},
year = {2026},
eprint = {2606.12352},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.12352},
}
Discrete VLA that generates actions autoregressively across temporal blocks and with masked diffusion inside each block, gaining speed and supporting real-time chunking via temporal in-painting.离散 token VLA：块与块之间自回归、块内做掩码离散扩散，兼顾时间连贯性与推理速度，并可通过时间补全支持实时分块执行。
Turns guidance from 3D foundation models, VLMs or human pose into differentiable 3D attractor/repeller energies that steer VLA action sampling. Reports gains for π0.5 and GR00T N1.6 in sim and real.把 3D 基础模型、VLM、人体姿态等指导信号统一表示为 3D 空间中可微的吸引/排斥能量函数，引导 VLA 动作采样。报告提升 π0.5 和 GR00T N1.6 在仿真与真机上的表现。
@misc{song2026omniguide,
title = {{OmniGuide: Universal Spatial Guidance Fields for Enhancing Generalist Robot Policies}},
author = {Yunzhou Song and Long Le and Yong-Hyun Park and Jie Wang and Junyao Shi and Lingjie Liu and Jiatao Gu and Eric Eaton and Dinesh Jayaraman and Kostas Daniilidis},
year = {2026},
eprint = {2603.10052},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.10052},
}
VLM agent that plans, monitors and recovers by treating a VLA or world-action model as an interruptible tool it can stop and steer mid-rollout. Comes with RoboVoLo, a long-horizon open-vocabulary benchmark.VLM 智能体把 VLA/世界动作模型当作可随时中断的工具，在执行中途调度和纠正，从而完成规划、监控与恢复。配套发布开放词汇长程操作基准 RoboVoLo。
@misc{chen2026volo,
title = {{VoLo: A Physical Orchestrator for Open-Vocabulary Long-Horizon Manipulation}},
author = {Siyi Chen and Hugo Hadfield and Alex Zook and Mikaela Angelina Uy and Chan Hee Song and Erwin Coumans and Xuning Yang and Faisal Ladhak and Qing Qu and Stan Birchfield and Jonathan Tremblay and Valts Blukis},
year = {2026},
eprint = {2606.07723},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.07723},
}
Steers a flow-matching VLA in real time from a keyboard by writing the user input into the action expert's initial noise; no retraining. Fine-tuning on steered runs also improves the autonomous policy.把键盘等简单用户输入写入流匹配动作专家的初始噪声，即可实时引导 VLA，无需重训；用引导后的轨迹微调还能提升自主策略。
@misc{kao2026flow,
title = {{Flow Control: Steering Vision-Language-Action Models with Simple Real-Time Inputs}},
author = {Jonathan C. Kao and Jason Chan and Andy Wang},
year = {2026},
eprint = {2606.10180},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.10180},
}
Turns a VLM into a VLA without architecture changes by writing a short language action description before the numeric action tokens. 2B model reports 90.8% on LIBERO after 4–5 hours of training; 0.8B–4B weights.不改结构直接把 VLM 变成 VLA：在数值动作 token 之前先生成一句自然语言动作描述。2B 模型训练 4–5 小时，报告 LIBERO 90.8%；发布 0.8B–4B 开放权重。
@misc{ishitoya2026clap,
title = {{CLAP: Direct VLM-to-VLA Adaptation via Language-Action Grounding}},
author = {Yuri Ishitoya and Jeremy Siburian and Masashi Hamaya and Haruto Suzuki and Kuniaki Saito and Toshihiko Fukushima and Cristian C. Beltran-Hernandez and Mai Nishimura},
year = {2026},
eprint = {2607.08974},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.08974},
}
Steers autoregressive VLAs at inference by injecting low-dimensional user input into the action-token space. Reports success from 10.0% to 72.5% and 16.7% to 93.8% on two household tasks.推理时把低维用户输入注入自回归 VLA 的动作 token 空间来引导轨迹，无需训练。报告两个家务任务成功率分别从 10.0% 升到 72.5%、从 16.7% 升到 93.8%。
Runtime wrapper for frozen VLAs: conformalized quantile regression ranks sampled actions, and a Mahalanobis anomaly score on latent states warns about abnormal execution. No retraining.冻结 VLA 的运行时外壳：用共形分位数回归为多个候选动作打分择优，并在潜在状态上计算马氏距离异常分，预警异常执行。无需重训。
@misc{lyu2026sentryvla,
title = {{SentryVLA: Uncertainty-Guided Action Selection and State Monitoring for Vision-Language-Action Robotic Control}},
author = {Zongyao Lyu and Lingling Chen and William J. Beksi},
year = {2026},
eprint = {2604.16677},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2604.16677},
}
Lets each modality update at its own sensor rate through latent buffers read by the action head, adding fast signals via gated cross-attention. Reports 95.2% vs 40.95% on seven contact-rich real tasks at 100 Hz.各模态按各自传感器频率刷新潜在缓存，供动作头持续读取，高频信号通过门控交叉注意力接入。报告在七个富接触真机任务上成功率 95.2%，对比基线 40.95%，控制频率 100 Hz。
@misc{vanjani2026dam,
title = {{DAM-VLA: Decoupled Asynchronous Multimodal Vision Language Action model}},
author = {Pankhuri Vanjani and Zhuoyue Li and Jakub Suliga and Moritz Reuss and Gianluca Geraci and Xinkai Jiang and Rudolf Lioutikov},
year = {2026},
eprint = {2606.12105},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.12105},
}
Serving one robot policy to a fleet from a remote GPU, posed as scheduling. Naive batching fails when robots consume action chunks at different rates; their scheduler reports up to 18% higher real-world throughput.把从远程 GPU 为多台机器人提供策略推理建模为调度问题。当各机器人消耗动作块的速度不同时，简单批处理失效；所提调度算法在真机上报告吞吐量最高提升 18%。
@misc{bansal2026action,
title = {{Action Chunk Scheduling for Batched Robot Policy Serving}},
author = {Rohan Bansal and David He and Nadun Ranawaka Arachchige and Zhenyang Chen and Soobum Kim and Kexin Rong and Danfei Xu},
year = {2026},
eprint = {2608.00337},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.00337},
}
Trains sparse autoencoders on VLA activations and finds features for motion primitives and concepts that transfer across episodes and can be amplified to steer behaviour. LIBERO + real DROID tests.在 VLA 隐层激活上训练稀疏自编码器，发现对应运动基元和语义概念、可跨回合迁移的特征，放大这些特征即可引导行为。在 LIBERO 和真实 DROID 平台验证。
@misc{swann2026sparse,
title = {{Sparse Autoencoders Reveal Interpretable and Steerable Features in VLA Models}},
author = {Aiden Swann and Lachlain McGranahan and Hugo Buurmeijer and Monroe Kennedy and Mac Schwager},
year = {2026},
eprint = {2603.19183},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.19183},
}
Shows a prompt that still reads like the intended task can redirect a VLA's rollout to an attacker's target; an on-policy prompt search finds such near-benign prompts. Sim + hardware.揭示 VLA 的轨迹级漏洞：看起来仍在描述原任务的指令，可以把机器人执行结果引向攻击者指定的目标；作者用在线提示搜索找到这类近乎无害的提示。仿真与真机验证。
Turns expert demos into negative supervision by pairing actions with mismatched instructions, training an instruction-grounded reward for offline RL of VLAs. More robust on LIBERO-PRO and a real AgiBot-G2-type robot.把专家动作与不匹配的指令配对生成反事实样本，训练与指令对齐的奖励模型，用于 VLA 离线强化学习，无需额外采集。在 LIBERO-PRO 和真机上提升鲁棒性。
@misc{chen2026geoalign,
title = {{GeoAlign: Beyond Semantics with State-Guided Spatial Alignment in VLA Models}},
author = {Yizhi Chen and Zhanxiang Cao and Xinyi Peng and Yixiao Zheng and Xiaxi Si and Yiheng Li and Liyun Yan and Keqi Zhu and Xueyun Chen and Shengcheng Fu and Tianyue Zhan and Yufei Jia and Jinming Yao and Yan Xie and Kun Wang and Cewu Lu and Yue Gao},
year = {2026},
eprint = {2606.03240},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.03240},
}
Two cheap failure detectors for fine-tuned VLAs that need no failure data: last-layer Mahalanobis distance and consistency between overlapping action chunks, plus a metric that also scores detection time.为微调后的 VLA 提出两种无需失败数据的轻量失败检测器：末层特征马氏距离与相邻动作块一致性，并提出同时衡量检测时间的指标。
Adds a training-only head that predicts 3D point tracks from VLA features, teaching world dynamics with no inference cost. Reports +10% on LIBERO-Long and +40% on RoboCasa.训练时为 VLA 加一个预测 3D 点轨迹的轻量头，让模型学到场景动力学，推理时去掉、零额外开销。报告 LIBERO-Long 提升 10%，RoboCasa 提升 40%。
@misc{kim2026pri,
title = {{Pri4R: Learning World Dynamics for Vision-Language-Action Models with Privileged 4D Representation}},
author = {Jisoo Kim and Jungbin Cho and Sanghyeok Chu and Ananya Bal and Jinhyung Kim and Gunhee Lee and Sihaeng Lee and Seung Hwan Kim and Bohyung Han and Hyunmin Lee and Laszlo A. Jeni and Seungryong Kim},
year = {2026},
eprint = {2603.01549},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2603.01549},
}
Transfers a pretrained π0 to aerial pick-and-place: payload-aware guidance in flow sampling handles flight dynamics and 3DGS data covers navigation. Reports pick-and-place success up from 23% to 50% in 460 real trials.把预训练 π0 迁移到空中抓放：在流采样中注入载荷约束以应对飞行动力学，并用高斯泼溅合成导航数据。在 460 次真机实验中报告抓放成功率从 23% 提升到 50%。
@misc{tucker2026but,
title = {{$π$, But Make It Fly: Physics-Guided Transfer of VLA Models to Aerial Manipulation}},
author = {Johnathan Tucker and Denis Liu and Aiden Swann and Allen Ren and Javier Yu and Jiankai Sun and Brandon Kim and Lachlain McGranahan and Quan Vuong and Mac Schwager},
year = {2026},
eprint = {2603.25038},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.25038},
}
Steers a pretrained flow-matching VLA with a spatial cue from a grounding model: a small adapter predicts an affine transform of the action chunk toward the cued target. Up to ~2x task success for open-world delivery.用视觉定位模型给出的空间提示引导预训练流匹配 VLA：轻量适配器预测作用在动作块上的仿射变换，使动作对准目标。开放世界配送任务成功率最高提升近 2 倍。
@misc{wang2026cue,
title = {{Cue the Flow: Steering Flow-Matching Policies for Open-World Delivery Manipulation}},
author = {Haoxuan Wang and Griffin Galimi and Junhua Huang and Selina Song and Wayne Wu and Yan Yan and Bolei Zhou},
year = {2026},
eprint = {2609.38989},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.38989},
}
Residual RL on top of a frozen VLA using only object poses, so it transfers from sim to real; a sim twin of the VLA is trained by replaying the same teleop demos. Reports real FR3 success from 42% to 76% zero-shot.在冻结的 VLA 之上做只依赖物体位姿的残差强化学习，从而可直接从仿真迁移到真机；通过回放同一批遥操作数据训练 VLA 的仿真副本。报告真实 FR3 上成功率从 42% 提升到 76%。
@misc{kim2026object,
title = {{Object-Centric Residual RL for Zero-Shot Sim-to-Real VLA Enhancement}},
author = {Kinam Kim and Namiko Saito and Heecheol Kim and Katsushi Ikeuchi and Jaegul Choo and Yasuyuki Matsushita},
year = {2026},
eprint = {2606.18953},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.18953},
}
Lifts 2D inputs to 3D with per-pixel vertex maps and expresses camera views and actions in one shared bird's-eye-view frame, aligning data across robots, cameras and datasets.用逐像素顶点图把 2D 输入提升到 3D，并把各相机视角与机器人动作统一表示在同一鸟瞰（BEV）坐标系中，对齐不同机器人、相机与数据集的数据。
@misc{zhou2026dexterity,
title = {{Dexterity-BEV: Aligning 3D World and Actions for Generalizable Robot Policies Learning}},
author = {Huayi Zhou and Wei Gao and Dekun Lu and Ruiji Liu and Zhanqi Zhang and Ziyang Zhang and Jian Chen and Wenlve Zhou and Sheng Xu and Shumin Li and Kangyi Guo and Shichen Xu and Zixin Huang and Yongyi Su and Kui Jia},
year = {2026},
eprint = {2606.02274},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.02274},
}
Plan-then-execute VLA that first predicts a short sequence of coarse action tokens summarising the intended trajectory, then the executable tokens conditioned on it. Largest gains on long multi-stage tasks.先规划后执行的 VLA：先预测概括未来轨迹的少量粗粒度动作 token，再以此为条件生成可执行动作 token；规划与执行共享同一离散动作词表。长程多阶段任务提升最大。
@misc{wu2026coarse,
title = {{Coarse-to-Control: Action-Token Planning for Vision-Language-Action Models}},
author = {Jinhao Wu and Shiduo Zhang and Yicheng Liu and Xiaopeng Yu and Sixian Li and Siyin Wang and Hang Zhao and Jing Huo and Yang Gao and Jingjing Gong and Xipeng Qiu and Yu-Gang Jiang},
year = {2026},
eprint = {2606.07107},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.07107},
}
Action tokenizer adapted from neural audio codecs (multi-scale RVQGAN with a Vocos-style decoder). Reports higher success than binning, FAST and earlier VQ tokenizers on LIBERO-10, RoboMimic and real tasks.借鉴神经音频编解码器的动作 tokenizer（多尺度 RVQGAN + Vocos 式解码器）。报告在 LIBERO-10、RoboMimic 与真机任务上成功率高于分箱、FAST 和已有 VQ 方法。
Gives a VLA memory as motion rather than frames: a short past video becomes continuous trajectory-field tokens that current visual tokens query. Smoother, more direct long-horizon execution.让 VLA 记住“运动”而非“帧”：把一小段历史视频转成时间连续的轨迹场 token，供当前视觉 token 查询。长程操作更平滑、更直接。
@misc{yuan2026motionvla,
title = {{MotionVLA: Injecting Geometric Motion into Vision-Language-Action Model}},
author = {Shanglin Yuan and Weiheng Zhao and Xianda Guo and Wei Sui and Li Yu and Wenyu Liu and Xinggang Wang},
year = {2026},
eprint = {2606.08288},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.08288},
}
RL post-training of VLAs with an online-updated world model whose measured reliability decides how much imagined data and how long rollouts to use. Reports 79.7% on perturbed LIBERO-PRO.VLA 强化学习后训练：在线更新的世界模型与奖励头作为模拟器，按其可信度动态决定想象数据比例和推演长度。报告扰动版 LIBERO-PRO 平均成功率 79.7%。
@misc{li2026trust,
title = {{Trust-Calibrated VLA Policy Post-Training with Evolving Imagination}},
author = {Wenzhan Li and Yiran Qin and Heng Zhou and Huirui Wang and Yulan Guo and Ruimao Zhang},
year = {2026},
eprint = {2610.16517},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.16517},
}
VLA with sparse experts: sensor availability gates modality experts, so it degrades gracefully when depth is missing, and task intent routes action experts. Strong on LIBERO, RoboCasa GR1 and real dual-arm tasks.稀疏专家 VLA：传感器是否可用决定模态专家的开关，缺少深度时也能平稳退化；任务意图再路由到动作专家。在 LIBERO、RoboCasa GR1 和真实双臂任务上表现强。
@misc{zhang2026core,
title = {{CoRE-VLA: Towards Scalable and Robust Vision-Language-Action Modeling via Conditional Routing of Experts}},
author = {Haozhe Zhang and Sixian Li and Yifei Zhang and Zezheng Huai and Hao Chen and Chunhua Shen and Jingjing Gong and Xipeng Qiu},
year = {2026},
eprint = {2607.03693},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.03693},
}
System for stable, sample-efficient RL fine-tuning of pretrained VLAs; reports 30/30 success on every task (e.g. plugging in string lights, pool shots) after ~19 minutes of robot data on average.稳定且样本高效的 VLA 强化学习微调系统；报告在插灯串插头、台球击球入袋等所有任务上均 30/30 成功，平均只需约 19 分钟在线机器人数据。
Always-on dual-arm robot: a π0.5 backbone with three light async modules that watch the live stream, track embodied state and recall its own past actions. Reports 45% vs 28% on the combined task.“永远在线”的双臂机器人：在 π0.5 主干上加三个轻量异步模块，分别监看实时流、跟踪本体状态、回忆自己过去的动作，两臂可并行执行。报告综合任务成功率 45%，最强基线 28%。
@misc{yi2026watch,
title = {{Watch, Recall, Act: Always-On Robots in Concurrent Embodied Streams}},
author = {Ding Yi and Peiwen Sun and Chenchu Rong and Jianan Wang and Xili Dai and Xiangyu Yue and Xi Lin},
year = {2026},
eprint = {2609.28429},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.28429},
}
@misc{dong2026actioncodec,
title = {{ActionCodec: What Makes for Good Action Tokenizers}},
author = {Zibin Dong and Yicheng Liu and Shiduo Zhang and Baijun Ye and Yifu Yuan and Fei Ni and Jingjing Gong and Xipeng Qiu and Hang Zhao and Yinchuan Li and Jianye Hao},
year = {2026},
eprint = {2602.15397},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.15397},
}
Turns robot state into VLM-vocabulary tokens that, with the instruction, choose which image patches the VLM processes. Keeping ~12% of visual tokens beats the full-token baseline on CALVIN with lower latency.把机器人状态离散成 VLM 词表 token，与指令一起决定哪些图像块进入 VLM 计算。仅保留约 12% 的视觉 token 就在 CALVIN 上超过全 token 基线，同时降低延迟。
@misc{wang2026think,
title = {{Think Proprioceptively: State-Grounded Visual Token Selection for VLA Policies}},
author = {Fangyuan Wang and Peng Zhou and Jiaming Qi and Shipeng Lyu and Chengyang He and David Navarro-Alarcon and Guodong Guo},
year = {2026},
eprint = {2602.06575},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.06575},
}
Interleaves text and image reasoning in one model: a forward chain predicts the next subgoal image, an inverse chain reasons about the actions to reach it, then actions are generated. Biggest gains on long tasks.在同一模型中交替进行文本与图像推理：正向推理预测下一个子目标图像，逆向推理据此分析如何到达，再生成动作。长程任务提升最大。
@misc{lu2026thinkingvla,
title = {{ThinkingVLA: Interleaved Vision and Language Reasoning for Robotic Manipulation}},
author = {Tianyi Lu and Hui Zhang and Zijie Diao and Junke Wang and Shengqi Xu and Xingyao Lin and Guojin Zhong and Ziyi Ye and Peng Wang and Zuxuan Wu and Yu-Gang Jiang},
year = {2026},
eprint = {2606.17937},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.17937},
}
Training-only 3D flow heads (point-level and gripper-level) teach a 2D VLA how the scene moves; no extra sensors or inference cost. Improves success, zero-shot robustness and data efficiency.仅在训练时加入 3D 流预测头（点级与末端级双重监督），让 2D VLA 学到场景动力学；推理时无需额外传感器或计算。提升成功率、零样本鲁棒性与数据效率。
@misc{go2026flowvla,
title = {{FlowVLA: Enabling World Dynamics Understanding through Pluggable 3D Flow Learning}},
author = {Junhyeong Go and Mingyu Jung and Woobin Im and Jongbin Ryu},
year = {2026},
eprint = {2610.16160},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.16160},
}
Lifelong VLA learning with short-term and long-term LoRA adapters mixed by a task-aware gate, plus cheap stochastic replay. Learns new skills while keeping old ones; tested on a real xArm.终身学习 VLA：短期与长期两条 LoRA 适配通路由任务感知门控融合，再配合低存储的随机回放，兼顾学新与保旧。在真实 xArm 上验证。
@misc{he2026towards,
title = {{Towards Human-like Physical Intelligence: Lifelong Vision-Language-Action Learning for Robotic Manipulation}},
author = {Yao He and Gan Sun and Wenqi Liang and Fazeng Li and Yang Cong},
year = {2026},
eprint = {2607.14852},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.14852},
}
Learns, with RL, a meta-policy that picks per state how many denoising steps (sequential) and candidate samples (parallel) a frozen diffusion/flow policy gets, trading success against compute.用强化学习训练一个元策略，按状态决定冻结的扩散/流策略在每步分配多少去噪步数（串行）和多少候选采样（并行），在成功率与计算量之间权衡。
@misc{li2026elastic,
title = {{ELASTIC: Efficiently Learning to Adaptively Scale Test-Time Compute for Generative Control Policies}},
author = {Andrew Zou Li and Gokul Swamy and Yonatan Bisk and Andrea Bajcsy},
year = {2026},
eprint = {2606.31132},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.31132},
}
Keeps a streaming buffer of action chunks at different noise levels with chunk-wise causal attention, so each inference step emits one executable chunk and asynchronous execution stays smooth.维护一个包含多个不同噪声水平动作块的流式缓冲区，并使用块级因果注意力，使每次推理产出一个可执行动作块，异步执行时动作保持连贯。
@misc{li2026flashvla,
title = {{FlashVLA: Streaming Action Decoding for Fast and Asynchronous VLA Inference}},
author = {Zekai Li and Jiaming Tang and Zhijian Liu},
year = {2026},
eprint = {2608.27384},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.27384},
}
Masks the end-effector in the wrist camera (rendered from known geometry, plus mask augmentation). A VLA trained on one parallel gripper then transfers zero-shot to another gripper, another arm and a five-finger hand.在腕部相机画面中遮住末端执行器（依据已知几何实时渲染遮罩，并做遮罩增强）。只用单一平行夹爪数据训练的 VLA 可零样本迁移到其他夹爪、其他机械臂和五指手。
@misc{piseno2026cloak,
title = {{Cloak: Zero-Shot Cross-Embodiment Manipulation by Masking the End-Effector from the VLA}},
author = {Michael Piseno and Guy Tevet and C. Karen Liu},
year = {2026},
eprint = {2606.22836},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.22836},
}
Splits a VLA into a key-action model (next subgoal) and a motion model (dense actions toward it) in a shared camera-centric action space across datasets and robots. Reports 91% on SimplerEnv Google Robot visual matching.把 VLA 拆为关键动作模型（预测下一个子目标）与运动规划模型（生成通往它的稠密动作），并统一到以相机为中心的动作空间。在 SimplerEnv Google Robot 视觉匹配上报告 91%。
@misc{peng2026h,
title = {{H-VLA: Hierarchical Vision-Language-Action Model with Key-Action Reasoning and Motion Planning in a Unified Action Space}},
author = {Xiongfeng Peng and Lu Xu and Yandong Wang and Jiaqian Yu and Zirui Zheng and Yamin Mao and Weiming Li and Inseop Chung and Hyun-woong Cho and Jaewook Yoo and Dongwook Lee and Daehyun Ji and Chao Zhang},
year = {2026},
eprint = {2609.22895},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.22895},
}
Feeds a grounded 3D target point through a two-layer MLP straight into a VLA's action head. On LIBERO-PRO it lifts GR00T-N1.6 from 31.2 to 77.5 under task perturbation; π0.5 gains similarly.把一个已定位的 3D 目标点经两层 MLP 直接注入 VLA 的动作头。在 LIBERO-PRO 任务扰动下，GR00T-N1.6 成功率从 31.2 提升到 77.5，π0.5 也有类似提升。
@misc{tsai2026direct,
title = {{Direct Action-Head Injection of A Grounded 3D Point Unlocks Spatial and Task Generalization}},
author = {Shiang-Feng Tsai and Jin-Cheng Jhang and Yen-Ling Tai and Jia-Hong Lai and Shih-Yun Wong and Kang-Tung Hsu and Yi-Ting Chen and Min Sun},
year = {2026},
eprint = {2606.27663},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.27663},
}
Adds an auxiliary branch that predicts future 3D tracks of robot keypoints, computed for free by forward kinematics, with gradient isolation from the VLM. The branch is dropped at inference.增加一个辅助分支，预测机器人关键点未来的 3D 轨迹（由正运动学免费得到），并与视觉语言主干做梯度隔离。推理时丢弃该分支，策略接口不变。
@misc{he2026elan,
title = {{ELAN4D: Embodiment-Centric 4D Supervision for Vision-Language-Action Models via Plug-and-Play Adaptation}},
author = {Zeyuan He and Bowen Yang and Zhirui Fang and Keru Zhou and Lei Jiang and Jingjing Qian and Fan Mo and Junchi Yan and Philip Torr and Xiu Li and Li Jiang and Jialin Yu},
year = {2026},
eprint = {2605.30484},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2605.30484},
}
A router picks a VLM planner per subtask from scene context (CoT depth, model size, memory). On a Franka DROID setup it matches a stronger model's success at up to 65% lower average latency.路由器根据场景上下文为每个子任务选择 VLM 规划器（思维链深度、模型规模、记忆长度）。在 Franka DROID 平台上以最多降低 65% 的平均延迟达到更强模型的成功率。
@misc{dao2026direct,
title = {{DIRECT: When and Where Should You Allocate Test-Time Compute in Embodied Planners?}},
author = {Jadelynn Dao and Milan Ganai and Yasmina Abukhadra and Ajay Sridhar and Mozhgan Nasr Azadani and Katie Luo and Clark Barrett and Jiajun Wu and Chelsea Finn and Marco Pavone},
year = {2026},
eprint = {2606.12402},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.12402},
}
Info-VLA fights forgetting in continual VLA learning by keeping cross-modal structure: replay anchors from a frozen teacher for contrastive alignment, plus a vision-language mutual-information term.Info-VLA 通过保持跨模态信息结构来缓解 VLA 持续学习中的遗忘：以冻结教师模型构造回放锚点做对比对齐，并加入视觉-语言互信息约束。
Shows autoregressive policies can run in real time by tuning the tokenization horizon and constrained decoding under strict latency bounds; reports they beat an equivalent flow-matching policy.说明通过调整分词时域并在严格延迟上限下做约束解码，自回归策略也能实时执行；报告其表现优于同级别的流匹配策略。
@misc{lee2026real,
title = {{Real-Time Execution with Autoregressive Policies}},
author = {Sangkyu Lee and Seohyeon Park and Tackgeun You and Avi Caciularu and Idan Szpektor and Hwasup Lim and Youngjae Yu},
year = {2026},
eprint = {2606.13355},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.13355},
}
A VLA conditioned on one demo video of an unseen task predicts visibility-aware future end-effector traces for precise grounding. Releases RoboCasa-DC; reports +9.5 pts real-world success from human demos.以未见任务的一段示教视频为条件的 VLA，通过预测考虑可见性的未来末端轨迹实现精确定位。发布 RoboCasa-DC 基准；在以人类示教为条件的真机实验中成功率提升 9.5 个百分点。
@misc{son2026seetraceact,
title = {{SeeTraceAct: Visibility-Aware Latent Planning from Cross-Embodiment Demonstration Videos}},
author = {Jaehyeon Son and Junhyun Kim and Kyle Kam and Jeremiah Coholich and Seok Joon Kim and Jinhoo Kim and Chris Dongjoo Kim and Jaemin Cho and Dieter Fox and Zsolt Kira},
year = {2026},
eprint = {2606.02745},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.02745},
}
FreeBar: training-free plugin that keeps consecutive action chunks of flow-based VLAs consistent with a CBF-style constraint solved in closed form, with little extra inference cost.FreeBar：无需训练的推理插件，用闭式求解的类 CBF 一致性约束让流式 VLA 的相邻动作块与已执行动作衔接，额外推理开销很小。
@misc{yu2026barrier,
title = {{Barrier-Constrained Guidance For Real-time Asynchronous VLA Execution}},
author = {Wenbo Yu and Jifan Li and Wenke Xia and Yuchen Li and Yizhuo Zhang and Pei Ren and Zhengping Che and Di hu},
year = {2026},
eprint = {2610.14997},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.14997},
}
Adds skill-conditioned LoRA mixture-of-experts adapters to the LLM backbone and turns a predicted affordance heatmap into a spatial prior for the action head. Reports Avg. Len 4.65 on CALVIN ABC→D.在大模型主干中加入以技能为条件的 LoRA 混合专家适配器，并把预测的可供性热图转为动作头的空间先验。在 CALVIN ABC→D 上报告平均完成长度 4.65。
@misc{yang2026mast,
title = {{MAST-VLA: Skill-Conditioned Adaptation with Trajectory-Affordance Priors for Multi-Skill Robot Manipulation}},
author = {Jiachi Yang and Zhiyi Xu and Hongquan Zhang and Xin Tan and Zhizhong Zhang and Jingyu Gong and Yuan Xie},
year = {2026},
eprint = {2610.32312},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.32312},
}
Aligning VLA features point-wise with 3D foundation models causes rank collapse; aligning pairwise relations instead adds 3D awareness without losing expressivity. Adds a new spatial reasoning benchmark.将 VLA 特征逐点对齐到 3D 基础模型会导致秩坍塌；改为对齐成对关系，可在不损失表达力的前提下注入 3D 空间感知。并提出新的空间推理基准。
@misc{zhu2026relational,
title = {{Relational Spatial Alignment for Vision-Language-Action Models}},
author = {Chaoran Zhu and Hengyi Wang and Chenglin Cui and Yazhe Wan and Taein Kwon and Changjae Oh},
year = {2026},
eprint = {2610.24705},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.24705},
}
B2FF: before execution the VLA imagines a bank of future milestone images; after a deviation a selector picks a reachable one as a fixed visual goal. Failure-injected LIBERO: 56.3% → 74.0%, no fine-tuning.B2FF：执行前让 VLA 想象一组未来里程碑图像；偏离轨迹后，由选择器挑出可恢复的那个作为固定视觉目标。在注入失败的 LIBERO 上成功率从 56.3% 升至 74.0%，无需微调。
@misc{shin2026back,
title = {{Back to the Familiar Future: Failure Recovery for VLA Policies via Pre-Imagined Milestone Selection}},
author = {Suyeon Shin and Juwon Kim and Hyeonbin Park and Hyunseo Kim and Hyundo Lee and Hyung-Sin Kim and Byoung-Tak Zhang},
year = {2026},
eprint = {2606.09258},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.09258},
}
Continuous (not vector-quantized) latent actions learned from 23.3K hours of egocentric human video serve as pre-training labels for a VLA; reports consistent scaling gains.从 2.33 万小时第一人称人类视频中学习连续（非向量量化）的潜动作，结合感知、语义与光流对齐信号，作为 VLA 预训练的动作监督；报告随规模持续提升。
@misc{xiang2026lapt,
title = {{LAPT-Scale: Scaling Embodied Policies with Latent Action Pre-Training on Human Videos}},
author = {Zhiyuan Xiang and Peidong Liu and Mingyang Li and Jiawei Li and Wenhow Li and Wenbo Ding and Xiang Deng},
year = {2026},
eprint = {2610.13732},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.13732},
}
Advantage-conditioned BC for VLAs: predicts pairwise conditions directly and injects them into the DiT action expert via AdaRMS. On real retry-heavy data: 90.0% vs 56.7% (bowl stacking), 56.7% vs 33.3% (key hanging).面向 VLA 的优势条件行为克隆：直接预测成对条件，并通过 AdaRMS 注入 DiT 动作专家。在含大量重试的真机数据上，叠碗成功率 90.0%（基线 56.7%），挂钥匙 56.7%（基线 33.3%）。
@misc{wang2026align,
title = {{ALIGN: Aligning Condition Estimation and Expression in Advantage-Conditioned Behavior Cloning}},
author = {Zechen Wang and Zhenqi Qiu and Houlong Xiong and Suohang Zhang and Yiyu Ren and Wanting Xu and Hongfei Niu and Chengyang He and Ge Sun and Ran Cheng and Qian Zhu},
year = {2026},
eprint = {2610.26455},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.26455},
}
Measures which instruction factors policies over-rely on: across six foundation policies, color ≥ object ≥ spatial ≥ verb ≥ size. Collecting more data for under-grounded factors wins with half the demos.量化策略过度依赖哪些指令要素：六个基础策略普遍呈现颜色 ≥ 物体 ≥ 空间 ≥ 动词 ≥ 尺寸的顺序。把采集预算倾斜给欠落地的要素，只用一半示教即可胜出。
@misc{qi2026scale,
title = {{Scale Up Strategically: Learning Compositional Generalization via Bias-Aware Evaluation and Data Collection for Robotic Manipulation}},
author = {Yu Qi and Zhang Ye and Xinyi Xu and Yuxuan Lu and Amitoj Sandhu and Boce Hu and Haojie Huang and Jonathan Tremblay and Lawson L. S. Wong},
year = {2026},
eprint = {2607.21582},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.21582},
}
FLARE, a physical spotlight attack, drops VLA success to zero. Naive augmentation defenses make models ignore color (≤47.5% on color tasks); ChromaGuard keeps color and reaches 97.5% and 92.5% on a real arm.FLARE 物理聚光灯攻击可让 VLA 成功率降为零。朴素的增强防御会让模型忽略颜色（颜色相关任务至多 47.5%）；ChromaGuard 保留色彩信息，在真机上达到 97.5% 与 92.5%。
DLS finds where VLA success breaks via on-policy digital-twin rollouts, locates the crossing with privileged progress signals, and shapes the flow toward success-producing denoising directions.DLS 通过数字孪生中的在策略推演大规模发现 VLA 成功与失败的边界，用特权进度信号定位越界位置，并据此强化导向成功的去噪方向、抑制导向失败的方向。
AHEAD wraps a frozen VLA with a small latent world model that forecasts future patch tokens from optical-flow velocity and feeds them to the action decoder. +4.9M params on a 7B OpenVLA; 79-97% on 20 dynamic sim scenes.AHEAD 为冻结的 VLA 加上一个小型潜空间世界模型，依据光流速度预测未来图像块 token 并交给动作解码器，以应对运动物体。在 7B OpenVLA 上只增加 490 万参数，20 个动态仿真场景成功率 79–97%。
@misc{syed2026intercepting,
title = {{Intercepting the Future: Latent-Space Predictive World Model for Dynamic VLA Manipulation}},
author = {Shahram Najam Syed and Arthur Jakobsson and Haoran Hao and Jeffrey Ichnowski},
year = {2026},
eprint = {2606.02486},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.02486},
}
Predicts task-relevant future states in a pretrained prediction-oriented feature space (no pixel reconstruction) and conditions actions on them via causal attention. +11.8 pts over reactive policies on RoboTwin.在预训练的面向预测的特征空间中预测与任务相关的未来状态（不做像素重建），并通过结构化因果注意力让动作生成以其为条件。在 RoboTwin 困难长时任务上比反应式策略高 11.8 个百分点。
@misc{liu2026plaw,
title = {{PLaW-VLA: Predictive Latent World Modeling for Vision-Language-Action Policies}},
author = {Yu Liu and Hetian Guo and Tianlv Huang and Ziyi Cai and Wudi Chen and Hantang Wang and Qiutong Liu and Yingzhi Peng and Wei Han and Peijun Tang and Jianan Wang and Zipei Fan and Zhiyuan Zha and Xuan Song},
year = {2026},
eprint = {2610.12285},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.12285},
}
FR-VLA uses learnable force query tokens to let vision-language context decide how much measured force should enter the action pathway, rather than feeding force in through a fixed path. Five real tasks, two robots.FR-VLA 通过可学习的力查询 token，由视觉语言上下文判断测得的力应在多大程度上进入动作生成通路，而不是走固定的注入路径。在两种机器人的五个接触密集任务上验证。
@misc{he2026fr,
title = {{FR-VLA: Implicit Force Reasoning VLA for Contact-Rich Manipulation}},
author = {Jiaju He and Zhiyuan Yang and Ruyang Han and Jiahong Tang and Zhiping Lin and Haiyue Zhu},
year = {2026},
eprint = {2610.25981},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25981},
}
Makes UMI data usable for VLAs: a fisheye wrist-view VQA dataset aligns the VLM to distorted views, and a physics validation pipeline scores trajectories for kinematic limits, collisions and controller bandwidth.让 UMI 数据可用于训练 VLA：构建腕部鱼眼视角的 VQA 数据集使 VLM 适应畸变画面，并用物理校验流程对轨迹的运动学极限、碰撞与控制带宽进行评分筛选。
@misc{yang2026vista,
title = {{VISTA: Vision-Grounded and Physics-Validated Adaptation of UMI data for VLA Training}},
author = {Siyuan Yang and Linzheng Guo and Ouyang Lu and Zhaxizhuoma and Daoran Zhang and Xinmiao Wang and Ting Xiao and Fangzheng Yan and Zhijun Chen and Yan Ding and Chao Yu and Chenjia Bai and Xuelong Li},
year = {2026},
eprint = {2606.04708},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.04708},
}
Gaze recorded in VR teleop is drawn as crosshairs on training images; at test time a small predictor supplies the gaze. With π0: 26.3% → 56.0% on six real bimanual tasks. Releases GazeMani (1,200 trajectories).把 VR 遥操作中记录的注视点画成十字准星叠加在训练图像上，部署时由轻量预测器估计注视点。基于 π0，六个真实双臂任务平均成功率从 26.3% 升至 56.0%；发布 GazeMani 数据集（1200 条轨迹）。
@misc{zhou2026gaze,
title = {{Gaze Prompts: Temporally Dense Human Attention for Vision-Language-Action Fine-Tuning}},
author = {Yihan Zhou and Rui Yan and Mingcong Li and Zheyuan Huang and Xu Yang and Xueyang Guo and Yilin Mo},
year = {2026},
eprint = {2609.34550},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.34550},
}
Online safe RL for VLAs inside an interactive video world model: a success classifier and a contact-probability safety head are balanced with Lagrangian-constrained GRPO. +8.25 pts safe success on SafeLIBERO Level I.在可交互视频世界模型中对 VLA 做在线安全强化学习：成功分类器与预测接触概率的安全头，经拉格朗日约束的 GRPO 平衡。在 SafeLIBERO Level I 上安全成功率提升 8.25 个百分点。
@misc{tang2026safedojo,
title = {{SafeDojo: Safe Reinforcement Learning for VLA via Interactive World Model}},
author = {Kai Tang and Peidong Jia and Zhong Chu and Jiajun Cao and Jixian Wu and Bohua Zhang and Rui Ma and Fangyuan Zhao and Sixiang Chen and Yichen Guo and Xiaowei Chi and Chun-Kai Fan and Kevin Zhang and Jinchang Xu and Fubing Yang and Weishi Mi and Xiaozhu Ju and Jian Tang and Shanghang Zhang},
year = {2026},
eprint = {2606.20698},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.20698},
}
RL-Co: SFT on mixed real and sim demos, then RL in simulation with an auxiliary supervised loss on real data to avoid forgetting. Tested on four real tabletop tasks with OpenVLA and π0.5.RL-Co：先在真实与仿真混合示教上做监督微调，再在仿真中强化学习，同时加入真实数据的辅助监督损失以防遗忘。用 OpenVLA 和 π0.5 在四个真实桌面任务上验证。
@misc{shi2026beyond,
title = {{Beyond Imitation: Reinforcement Learning-Based Sim-Real Co-Training for VLA Models}},
author = {Liangzhi Shi and Shuaihang Chen and Feng Gao and Yinuo Chen and Kang Chen and Tonghe Zhang and Hongzhi Zang and Jiakai Zhou and Weinan Zhang and Chao Yu and Yu Wang},
year = {2026},
eprint = {2602.12628},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.12628},
}
Generated robot videos keep geometry but not control, so GRA uses them only to supervise visual perception (aligned to extracted geometry) and leaves action learning to real demos.生成的机器人视频保留了几何轨迹却丢失了控制信号，因此 GRA 只用它们监督视觉感知（与提取的几何对齐），动作学习仍交给真实示教。
@misc{chen2026supervise,
title = {{Supervise What Survives: Geometry-Guided VLA Adaptation from Synthetic Robot Videos}},
author = {Danze Chen and Yanzhe Chen and Qiming Huang and Zhijun Cao and Chen Gao and Mike Zheng Shou},
year = {2026},
eprint = {2606.24448},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.24448},
}
AdvDex trains a dexterous VLA on human and robot demos: OmniShare human data with touch, a joint-aligned action space (wrist pose + 15 finger joints) shared by hands and grippers, and adversarial removal of body cues.AdvDex 用人类与机器人示教训练灵巧操作 VLA：含触觉的人类示教数据集 OmniShare、人手/灵巧手/夹爪共用的关节对齐动作空间（腕部位姿 + 15 个手指关节），并以对抗学习去除本体相关特征。
@misc{zhao2026advdex,
title = {{AdvDex: Learning Dexterous Manipulation from Human Demonstrations via Joint-Aligned Actions and Adversarial Learning}},
author = {Zhiyue Zhao and Jingyi Wu and Hairuo Liu and Mingyu Liu and Liyang Li and Hengdi Zhang and Tong He and Zhengxue Cheng},
year = {2026},
eprint = {2608.14028},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.14028},
}
PACER fine-tunes a VLA for electronics disassembly on its own rollouts plus human corrections, reweighting action chunks by progress, source and outcome; 87/100 vs 54/100 for plain SFT on motherboard parts.PACER 用策略自身的执行轨迹和人工纠正数据微调用于电子产品拆解的 VLA，按进度、来源和结果为动作块加权；在主板部件拆解上成功 87/100，普通 SFT 为 54/100。
@misc{liu2026pacer,
title = {{PACER: Process-Aware VLA Policy Learning from Human Corrections for Robotic Disassembly}},
author = {Chang Liu and Sibo Tian and Litian Gong and Zhiwen Fan and Jiachen Li and Xiao Liang and Minghui Zheng},
year = {2026},
eprint = {2610.19248},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.19248},
}
CompVLA adds a compliance expert to a VLA that predicts time-varying stiffness and virtual displacement alongside motion, executed with geometric impedance control for contact-rich tasks.CompVLA 为 VLA 增加柔顺专家模块，在预测运动的同时输出时变刚度与虚拟位移，并通过几何阻抗控制执行，用于接触密集任务。
@misc{kim2026compvla,
title = {{CompVLA: A Variable Compliance Vision-Language-Action Model for Contact-rich Manipulation}},
author = {Jongmin Kim and Junsu Ha and Che-Sang Park and Minchang Song and Hyeokju Jeong and Himchan Hwang and Jianlong Fu and Frank C. Park},
year = {2026},
eprint = {2609.23614},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.23614},
}
HAF-VLA adapts a flow-matching VLA to humanoid loco-manipulation by denoising in stages, locomotion and head first, then waist, then full body, linked by a cross-stage KV cache; six real tasks.HAF-VLA 将流匹配 VLA 适配到人形移动操作：分阶段去噪，先生成移动与头部指令，再细化腰部，最后输出全身操作动作，各阶段通过跨阶段 KV 缓存衔接；在六个真实任务上验证。
@misc{gu2026haf,
title = {{HAF-VLA: Hierarchical Action Flow Framework for Humanoid Vision-Language-Action Control}},
author = {Langzhe Gu and Chengkai Hou and Meng Li and Xinhua Wang and Xinyuan Lv and Shuanghao Bai and Guangrun Li and Jingyang He and Ziluo Ding and Zhiyuan Xu and Ning Ding and Kuan Cheng and Jian Tang and Shanghang Zhang and Zhengping Che},
year = {2026},
eprint = {2610.26584},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.26584},
}
Studies adapting closed-weight Gemini Robotics On-Device on a humanoid via a managed fine-tuning API; CLIFT turns deployment reward feedback into SFT data, reaching near-perfect success after two cycles.研究通过托管微调 API 在人形机器人上适配闭源权重的 Gemini Robotics On-Device；CLIFT 把部署时的奖励反馈转为监督微调数据，两轮迭代后接近完美成功率。
@misc{chen2026clift,
title = {{CLIFT: Turning Gemini Robotics On-Device into Humanoid Specialists via Non-Invasive Closed-Loop Iterative Fine-Tuning}},
author = {Yuxin Chen and Hari Srikanth and Nathan Jew and Menglin Wu and Pengcheng Wang and Junli Ren and Masayoshi Tomizuka and Peng Xu and Jinyu Xie and Thomas Tian},
year = {2026},
eprint = {2607.29172},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.29172},
}
Stores each skill as a frozen low-rank action-expert adapter and routes per action chunk via a VLM-conditioned router, so a VLA keeps old skills while learning composite tasks; tested on RoboCasa365 and a PIPER arm.把每类技能存为冻结的低秩动作专家适配器，由 VLM 条件路由器按动作块选择组合，使 VLA 学习组合任务时不遗忘旧技能；在 RoboCasa365 和 PIPER 真机上验证。
@misc{tao2026prism,
title = {{PRISM: Primitive Routing via In-context Skill Mixing for Retentive VLA Adaptation}},
author = {Jiachen Tao and Gengyu Zhang and Haoxuan Wang and Bing Liu and Yan Yan},
year = {2026},
eprint = {2611.08917},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08917},
}
Renders embodied tasks as on-screen software so an unmodified computer-use agent acts through screenshots, mouse and keys; one design placed 1st, 2nd, 3rd and 6th across four ECCV workshop competitions.把具身任务渲染成屏幕上的软件，让未经改动的计算机使用智能体通过截图和键鼠完成；同一设计在四项 ECCV 研讨会竞赛中分获第一、二、三和第六名。
VLA for language-guided person following with mask and box grounding supervision and target-conditioned history compression; the 0.8B model reports 77.9% on EVT-Bench, plus a 126-episode quadruped tracking dataset.面向语言引导行人跟随的 VLA：加入掩码与检测框定位监督，并按目标压缩历史视觉信息；0.8B 模型在 EVT-Bench 上平均成功率 77.9%，另发布 126 段四足机器人跟踪数据。
@misc{deng2026dobottrack,
title = {{DobotTrack: Target-Grounded Vision-Language-Action Learning for Embodied Person Tracking}},
author = {Shengjun Deng and Yinghan Zhang and Chengrui Wei and Liu Liang and Songyuan Liu and Pengkun Zheng and Yajun Zheng and Ying Zhang and Yuan Tian and Xi Li and Teng Wang},
year = {2026},
eprint = {2611.09041},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09041},
}
SPARK-VLA distils a frozen 3D expert into a LoRA visual branch beside the original one, and a small router passes only action-relevant tokens to a 0.5B LLM. No 3D input at inference; ~53M trainable parameters.SPARK-VLA 把冻结的 3D 专家蒸馏到与原视觉通路并行的 LoRA 分支，并用轻量路由器只把与动作相关的视觉 token 送入 0.5B 语言模型。推理时无需 3D 输入，可训练参数约 5300 万。
SERA turns the high-level VLM into an agent with an editable harness that tracks task progress for a fixed VLA executor, and revises the harness from rollouts; reports 51.4% vs 44.4% overall success on RoboMME.SERA 把高层 VLM 规划器做成带可编辑执行框架的智能体，为固定的底层 VLA 记录任务进度与执行反馈，并利用环境回放自动改进框架；报告在 RoboMME 上总体成功率 51.4%，未自我演化时为 44.4%。
@misc{du2026from,
title = {{From VLM Planners to Robotic Agents: Self-Evolving Harnesses for History-Dependent Manipulation}},
author = {Tingwen Du and Haoyu Guo and Hanxue Zhang and Xuanxu Wang and Junting Dong and Chao Gao and Mingfei Han and Tao Lu and Mulin Yu and Chunhua Shen and Xiaojun Chang and Zhihui Li},
year = {2026},
eprint = {2610.29560},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.29560},
}
Coding-agent search over robot programs that adds physical understanding at object, scene and interaction levels: a knowledge base, scene grounding, and self-designed experiments that test competing hypotheses.让编码智能体搜索机器人程序时具备物体、场景和交互三层物理理解：知识库、场景落地，以及用自行设计的实验检验相互竞争的假设。
@misc{liu2026roboauto,
title = {{RoboAuto: Autonomous Robot Program Search through Multi-Level Physical Understanding}},
author = {Buzhao Liu and Jianing Ye and Xiu Yuan and Yangtong Feng and Chengsong Huang and Tyler Dinh and Yevgeniy Vorobeychik and Ning Zhang and Chongjie Zhang},
year = {2026},
eprint = {2610.36533},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.36533},
}
Adapts a frozen diffusion or flow policy at test time: a VLM writes differentiable rewards that steer denoising when obstacles or clutter shift the scene. Reports +31% on CALVIN and +13% on LIBERO-PRO; no retraining.在测试时适配冻结的扩散或流匹配策略：当障碍或杂物改变场景时，由 VLM 生成可微奖励来引导去噪。报告 CALVIN 提升 31%、LIBERO-PRO 提升 13%，无需重新训练。
Dexterous visuotactile VLA on a mixture-of-transformers: contact-aware gating decides when touch matters; the model jointly imagines future visual and tactile latents. Reports 71% average success on contact-rich tasks.基于混合 Transformer 的灵巧视触觉 VLA：接触感知门控决定何时引入触觉，并联合想象未来的视觉与触觉潜变量。报告在富接触任务上平均成功率 71%。
@misc{fu2026decal,
title = {{DeCAL: Towards Physically-Grounded Dexterous Vision-Language-Action Models via Contact-Aware Latent Co-Imagination}},
author = {Yankai Fu and Ning Chen and Junkai Zhao and Heng Zhang and Guocai Yao and Pengwei Wang and Zhongyuan Wang and Shanghang Zhang},
year = {2026},
eprint = {2609.09119},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.09119},
}
Mobile-manipulation agent with two loops: a VLM plans and composes flow-matching skills at run time, and an outer loop mines rollouts to find and fine-tune skills. Reports +22.5 pts over π0.5 on BEHAVIOR-1K.双循环移动操作智能体：内循环由 VLM 规划并组合流匹配技能，外循环从部署数据中自动发现并微调技能。报告在 BEHAVIOR-1K 上比 π0.5 高 22.5 个百分点。
@misc{liu2026mobiagent,
title = {{MobiAgent: Dual-Loop Recursive Policy Self-Improvement for Long-Horizon Mobile Manipulation}},
author = {Chenzhi Liu and Yue Zhang and Jiehong Lin and Jianan Wang and Bo Wang and Zhongrui Wang and Xiaojuan Qi},
year = {2026},
eprint = {2610.03476},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.03476},
}
Harness that lets a VLM drive a humanoid: it points at image locations for 3D positions and issues pelvis, torso, hand, walk and gaze commands. After fine-tuning, a real G1 carries a box with no real data.为 VLM 控制人形机器人提供“挽具”接口：在图像上点选即得 3D 位置，并下达骨盆、躯干、手、行走和视线指令。用教师轨迹微调后，真实 G1 无需真机数据即可搬箱子。
@misc{lu2026hitch,
title = {{HITCH: Teaching a Vision-Language Model to Operate a Humanoid through a Whole-Body Harness}},
author = {Qingzhou Lu and Chenxin Liu and Xinming Wang and Liu Cao and Yanjiang Guo and Mengdi Xu and Jianyu Chen},
year = {2026},
eprint = {2611.08039},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08039},
}
World action model that brings scene-level 3D flow (tracked 3D points) inside the policy rather than as a separate module or extra head, so 4D geometry and visual semantics are modelled together with actions.世界动作模型：把场景级三维流（跟踪的三维点）纳入策略内部，而不是作为独立模块或额外输出头，让 4D 几何、视觉语义与动作一起建模。
@misc{tian2026uniflowwam,
title = {{UniFlowWAM: A Unified World-Action Model with Visual Semantics and 4D Geometry}},
author = {Jialin Tian and Zhuofan Li and Xinkai Wang and Bokai Lin and Chenyi Wang and Xinyu Zhan and Cewu Lu and Lixin Yang},
year = {2026},
eprint = {2610.30991},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.30991},
}
A 16B model interleaving video prediction and language reasoning for embodied understanding and long-horizon visual planning, plus a world-action extension for control; reports 51.5% on 20 RoboCasa365 tasks.16B 统一具身模型，交替进行视频预测与语言推理，兼顾具身理解与长程视觉规划，并扩展出用于机器人控制的世界-动作版本；另提出 Rynn-VP、Rynn-GEU 评测，报告在 20 个 RoboCasa365 组合任务上成功率 51.5%。
@misc{zhu2026rynnunify,
title = {{RynnUnify: Towards Unified Embodied Brain}},
author = {Minghao Zhu and Bohan Hou and Kun Xiang and Zixuan Wang and Yinzhou Tang and Yifei Yang and Zhikai Wang and Tianyi Zhang and Kehan Li and Haoyu Zhao and Siteng Huang and Leigang Qu and Sicong Leng and Dongchi Huang and Mingxiu Chen and Shijian Lu and Yue Wang and Xiaodan Liang and Zhenyu Zeng and Deli Zhao and Chengju Liu and Jianfei Yang and Xin Li},
year = {2026},
eprint = {2610.32965},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.32965},
}
Auxiliary training task for world action models: an EMA teacher sees the clean frame, the online model sees a perturbed one and aligns its action features. Reports better robustness for four WAMs.为世界-动作模型加一个自监督辅助任务：EMA 教师看原始画面，在线模型看外观扰动后的画面并对齐其动作表征；报告在 RoboTwin 2.0、LIBERO-Plus 等基准上提升四种 WAM 对视觉扰动的鲁棒性，且不损失原有性能。
@misc{zou2026pica,
title = {{PICA-WAM: Visual Perturbation Invariance via Cross-Appearance Self-Alignment in World Action Models}},
author = {Yude Zou and Yifei Yao and Guanbo Jiao and Jianming Ma and Guanjie Zheng and Xing Gao},
year = {2026},
eprint = {2610.26813},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.26813},
}
World action model that acts in one diffusion step: 'imagination tokens' replace random-noise initialisation as an action prior. Reports ~3× speedup over direct-policy WAMs, 40 ms latency on an RTX 5090.单步扩散即可生成动作的世界动作模型：用“想象 token”代替随机噪声初始化，为动作提供先验。报告比直接策略式 WAM 快约 3 倍，编译后在 RTX 5090 上延迟低至 40 毫秒，覆盖自动驾驶与具身控制。
@misc{zhang2026oswam,
title = {{OSWAM: An Effective World Action Model with One-Step Action Generation}},
author = {Guowen Zhang and Liyi Chen and Xu Hu and Pengfei Wang and Jingfan Chen and Chenhang He and Lei Zhang},
year = {2026},
eprint = {2610.35850},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.35850},
}
World action model without video generation: predicts future features in a vision foundation model's latent space as policy subgoals. Reports 98.6% on LIBERO and up to 24× lower latency than pixel WAMs.不生成视频的世界动作模型：在视觉基础模型的特征空间中训练潜动作模型，预测未来观测特征作为“潜在视觉子目标”来条件化动作生成。报告 LIBERO 98.6%、RoboTwin 91.22%，每个动作块推理 187 ms，比像素级 WAM 延迟低最多 24 倍。
@misc{chen2026lawam,
title = {{LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies}},
author = {Jialei Chen and Kai Wang and Kang Chen and Shuaihang Chen and Feng Gao and Wenhao Tang and Zhiyuan Li and Weilin Liu and Zhuyu Yao and Boxun Li and Yuanbo Xu and Chao Yu},
year = {2026},
eprint = {2606.15768},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.15768},
}
A latent world-action model that adds future-aware latent queries to a VLA-style policy and trains them with a future-informed branch, without generating future frames.潜空间世界动作模型，在 VLA 式策略中加入面向未来的潜查询，并用利用未来信息的分支训练，全程不生成未来帧。
@misc{luo2026being,
title = {{Being-H0.7: A Latent World-Action Model from Egocentric Videos}},
author = {Hao Luo and Wanpeng Zhang and Yicheng Feng and Sipeng Zheng and Haiweng Xu and Chaoyi Xu and Ziheng Xi and Yuhui Fu and Zongqing Lu},
year = {2026},
eprint = {2605.00078},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2605.00078},
}
A robustness study that compares world action models against VLAs under perturbation, to test the claim that explicit future prediction gives better generalization.在扰动下对比世界动作模型与 VLA 的鲁棒性研究，用于检验显式未来预测能带来更好泛化这一说法。
@misc{zhang2026do,
title = {{Do World Action Models Generalize Better than VLAs? A Robustness Study}},
author = {Zhanguang Zhang and Zhiyuan Li and Behnam Rahmati and Rui Heng Yang and Yintao Ma and Amir Rasouli and Sajjad Pakdamansavoji and Yangzheng Wu and Lingfeng Zhang and Tongtong Cao and Feng Wen and Xinyu Wang and Xingyue Quan and Yingxue Zhang},
year = {2026},
eprint = {2603.22078},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.22078},
}
Builds a robot foundation model on video instead of language: an open video planner trained on internet-scale human video makes zero-shot plans that are turned into robot actions. Model and dataset are released.以视频而非语言为主干构建机器人基础模型：在互联网规模人类视频上训练开源视频规划器，零样本生成视频计划再转为机器人动作；模型与数据集均已开源。
@misc{chen2025large,
title = {{Large Video Planner Enables Generalizable Robot Control}},
author = {Boyuan Chen and Tianyuan Zhang and Haoran Geng and Caiyi Zhang and Peihao Li and Kiwhan Song and William T. Freeman and Jitendra Malik and Pieter Abbeel and Russ Tedrake and Vincent Sitzmann and Yilun Du},
year = {2025},
eprint = {2512.15840},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2512.15840},
}
Small world action model trained from scratch on frozen V-JEPA 2.1 features, predicting future features only in training. Reports SOTA on RoboTwin 2.0 among policies without embodied pretraining.在冻结的 V-JEPA 2.1 特征上从零训练的紧凑世界动作模型，仅在训练时预测未来特征。报告在无具身预训练的策略中 RoboTwin 2.0 最优，并修正了 LIBERO-Plus 的指令。
@misc{clemente2026la,
title = {{La-JEPA: Compact Robot Policies with Latent World Modeling}},
author = {Mateo Clémente and George Z. Xue and Amin Abyaneh and Liheng Ma and Rui Heng Yang and Tongtong Cao and Zhanguang Zhang and Yingxue Zhang},
year = {2026},
eprint = {2611.08156},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08156},
}
Finds a VLA at 80% on atomic skills drops to 19% when composed into mobile manipulation, diagnoses why, and builds an agentic system around it.发现 VLA 在原子技能上 80% 成功率、组合成移动操作后仅 19%，分析失败原因并据此构建智能体式系统。
@misc{wang2026cracking,
title = {{Cracking Mobile Manipulation: From Failure Diagnosis to Agentic Robot System}},
author = {Zixuan Wang and Jia-Ning Li and Yinzhou Tang and Yanzhao Yu and Biao Wu and Meng Fang and Xin Li and Yiwu Zhong},
year = {2026},
eprint = {2610.19600},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.19600},
}
Adapts a language-guided navigation policy to a new building from one iPhone scan, compiled into a training environment with synthetic instructed episodes. Reports 85% real success vs 49% for a demo-fine-tuned baseline.只需用 iPhone 扫描一次新建筑：扫描被编译成训练环境并自动合成带语言指令的导航回合，再用无评论家的闭式 Pareto 效用适配策略；报告真机成功率 85%，用人类示教后训练的基线为 49%。
@misc{yu2026one,
title = {{One Scan is Enough: Demonstration-Free Adaptation for Language-Guided Navigation}},
author = {Zhuoyuan Yu and Yuxing Long and Du Wang and Junyang Wang and Hongwei Fan and Hao Dong},
year = {2026},
eprint = {2610.33794},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.33794},
}
Argues VLN success metrics hide a 'last-3-meter grounding gap': agents stop nearby without facing the target. Adds instance-level metrics and REALM, a refinement module for any backbone.指出 VLN 成功率掩盖了“最后 3 米定位缺口”：智能体停在附近却未对准目标。提出实例级指标和可接入任意主干的修正模块 REALM。
@misc{shi2026from,
title = {{From Region Arrival to Instance-Level Grounding in Vision-and-Language Navigation}},
author = {Xiangyu Shi and Ruoxi Yang and Wei Tao and Jiwen Zhang and Yanyuan Qiao and Qi Wu},
year = {2026},
eprint = {2607.03792},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.03792},
}
Fine-tunes Qwen3.5-9B on ~1.5M grounding samples auto-built from robot recordings plus other visual data; reports a 75.41 average over seven grounding benchmarks vs 73.68 for GPT-6 Astra.用自动标注的机器人操作录像加多来源视觉数据构建约 150 万条可验证的定位监督，微调 Qwen3.5-9B；报告在七个公开视觉定位基准上平均 75.41 分，GPT-6 Astra 为 73.68。
@misc{wang2026ace,
title = {{ACE-Brain-0.6: Scaling Spatial Grounding for Physical Foundation Models}},
author = {Yinhe Wang and Tao Tao and Zexin He and Shuang Luo and Wei Mao and Yao Mu and Feng Pan and Hanming Deng and Lewei Lu and Zhi Hou and Xingang Pan and Dahua Lin and Dacheng Tao and Xiaogang Wang},
year = {2026},
eprint = {2610.27040},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.27040},
}
Continual off-policy RL for VLAs — small actor/critic networks carry over between tasks, distilled on a few kept episodes to avoid forgetting; reports 31–68.5% less real-robot time on new assembly tasks.面向 VLA 的持续离策略强化学习：轻量 actor/critic 在任务间延续，用少量保留的旧任务轨迹做蒸馏以防遗忘。论文报告新的接触式装配任务所需真机训练时间减少 31–68.5%，旧任务几乎不遗忘。
@misc{kim2026accrue,
title = {{ACCRUE: Continual Off-Policy Reinforcement Learning for Robot Foundation Models}},
author = {Changyeon Kim and Kimin Lee and Yashraj Narang and Yijie Guo},
year = {2026},
eprint = {2610.18596},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.18596},
}
Post-trains an open 9B robot planner (SFT + multi-turn RL) across harnesses that call VLA skills or write control code; reports LIBERO-PRO success rising from 18.1% to 71.0% without external memory.像训练编程智能体一样对 9B 开源机器人规划器做后训练（SFT + 多轮强化学习），适配调用 VLA 技能和生成控制代码两类执行框架；报告在 LIBERO-PRO 上无外部记忆成功率从 18.1% 提升到 71.0%。
@misc{liu2026agenticroborl,
title = {{AgenticRoboRL: Bringing Coding-Agent Post-Training to Agentic Robotics}},
author = {Xinxin Liu and Bingxin Xu and Wenrui Bao and Ming Li and Wentao Bao and Chen Chen and Yuzhang Shang},
year = {2026},
eprint = {2610.28589},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.28589},
}
Distills a pretrained VLA teacher into a policy without action pretraining via on-policy supervision plus teacher-run recoveries, triggered by a reward model (AutoDAgger); reports up to 4.1x over BC on 3 real tasks.把预训练具身基础模型（教师）蒸馏到未做动作预训练的策略：在学生访问的状态上做在线监督，并由奖励模型监控进度、自动触发教师接管采集纠错轨迹（AutoDAgger，无需人工看护）；报告在 3 个真机任务上比行为克隆最高提升 4.1 倍。
@misc{lu2026tilde,
title = {{TILDE: Teacher-in-the-Loop Distillation from Embodied Foundation Models}},
author = {Yuxiang Lu and Yantong Liu and Jinghua Hou and Xueli Sun and Ruihua Han and Zhe Liu and Hengshuang Zhao},
year = {2026},
eprint = {2610.25376},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25376},
}
Co-trains a VLA on robot demos and noisy egocentric-video actions by weighting each source by reliability and distilling from clean future frames; reports steeper scaling than the baseline up to 30k hours of data.将机器人示教与动作标签噪声较大的第一视角视频联合训练 VLA：按数据源可靠性分配监督权重，并从干净的未来画面蒸馏表征；报告在最多 3 万小时预训练数据范围内扩展斜率高于基线。
@misc{zhang2026egoforesight,
title = {{EgoForesight: Effectively Scaling Policy Learning with Noisy Egocentric Actions}},
author = {Chubin Zhang and Changyuan Wang and Yirui Wang and Ao Li and Shilin Ma and Yue Su and Cai Zhou and Xiuwei Xu and Tianle Zhang and Yansong Tang},
year = {2026},
eprint = {2610.31845},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.31845},
}
Packages VLA policies, perception models and know-how as 250 reusable "embodied skills" and measures how much they help LLM agents on physical tasks.将 VLA 策略、感知模型与经验知识封装成 250 个可复用的“具身技能”，评测它们对大模型智能体完成物理任务的帮助。
@misc{han2026embodied,
title = {{Embodied Skill Bench: Benchmarking How Well Agents with Embodied Skills Perform on Physical Tasks}},
author = {Zhi Han and Zihan Guo and Chenxi Zeng and Zicai Cui and Yue Xin and Ming Zhou and Dongzhan Zhou and Weinan Zhang and Yang Li},
year = {2026},
eprint = {2610.16238},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.16238},
}
VLA robustness benchmark on RoboCasa365: paired clean/perturbed rollouts differing in one shift, plus stage-level diagnostics of where long-horizon failures start; evaluated on GR00T N1.5.基于 RoboCasa365 的 VLA 鲁棒性诊断基准：成对比较只差一个分布偏移因素的干净/扰动执行，并定位长时序任务中失败从哪个阶段开始，而不只看总成功率；在 GR00T N1.5 上评测。
@misc{wang2026robocasa,
title = {{RoboCasa-Shift: Stage-Aware Robustness Evaluation of Vision-Language-Action Policies in Long-Horizon Household Manipulation}},
author = {Haocheng Wang and Youquan Liu and Weiyu Guo and Aokai Zhang and Donghang Wu and Hongyu Liu and Sha Zhang and Weidong Yang and Ben Fei},
year = {2026},
eprint = {2610.27861},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.27861},
}
Benchmark of 410 long-horizon task environments (manipulation, puzzles, household) for foundation-model robot agents, with a curated EP-100 split and a shared protocol checking goals, rules and physical validity.面向基础模型机器人智能体的长程任务基准：410 个任务环境（基础操作、序列谜题、多阶段家务），精选 EP-100 子集，统一协议同时检查目标完成、任务规则与物理有效性。
@misc{chen2026embodied,
title = {{Embodied Playground: A Long-Horizon Agentic Robotics Benchmark for Physical, Strategic, and Spatial Reasoning}},
author = {Feng Chen and Pei Zhou and Lixuan He and Yunchao Yao and Li Sun and Shenghua Gao and Mingyu Ding and Kaifeng Zhang and Kaichen Zhou and Yi Ma},
year = {2026},
eprint = {2610.27039},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.27039},
}
BEHAVIOR-Spatial, built on BEHAVIOR-1K: spatial QA, point grounding and robot action tests. A baseline policy drops from 41.6% on held-out layouts to 16.1% when the requested target changes; oracle target cues help most.基于 BEHAVIOR-1K 的 BEHAVIOR-Spatial 基准：包含空间问答、指点定位与机器人动作评测。基线策略在未见布局上成功率 41.6%，但在同一布局中更换目标物体指令时降至 16.1%；提供真值目标提示带来的提升最大。
@misc{chen2026behavior,
title = {{Behavior Spatial Benchmark: Evaluating Spatial Understanding for Robotics}},
author = {Sirui Chen and Huang Huang and Qineng Wang and Wenlong Huang and Manling Li and Ruohan Zhang and Li Fei-Fei and Jiajun Wu},
year = {2026},
eprint = {2610.34051},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.34051},
}
Benchmark where a robot must run physics experiments to find an unknown answer: 30 experiments, 5 disciplines, parameters resampled per instance. Reports even the best frontier agent succeeds under half the time.“为求知而行动”的基准：机器人在物理仿真实验室中自行操作仪器、读数，测出每次随机重采样的未知参数。共 30 个实验、5 个学科；报告最强的前沿模型智能体在原生控制下成功率也不到一半，失败多源于与仪器交互。
@misc{gao2026embodiedlab,
title = {{EmbodiedLab: Benchmarking Embodied Agents That Act to Know}},
author = {Hongcheng Gao and Zelin Zheng and Alex Zhou and Jay Zhu and Shijia Ge and Yazhe Wang and Xiping Li and Jianshu Zeng and Di Wu and Xuan Shangguan and Yexing Wan and Lingyu He and Yongli Wang and Xiaohan Yu and Zhiqi Jia and Yue Liu and Bryan Hooi and Yinpeng Dong and Wentao Zhang and Sihang Wu and Xiao He},
year = {2026},
eprint = {2610.35385},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.35385},
}
@misc{mao2026robotagentbench,
title = {{RobotAgentBench: Towards Open-world Robot Autonomy}},
author = {Lingjun Mao and Jinglin Cao and Lukun He and Wenpeng Xu and Yuchen Yan and Hengyu Jin and Zijun Zhang and Jinzhou Tang and Songyao Jin and Ziche Liu and Yifei Shao and Junbo Huang and Fang Nan and Ruobing Han and Ziqiao Xi and Ziming Xu and Shuang Liang and Sibo Zhu and Wenyi Wu and Xinyue Wang and Yiming Huang and Xinkai Zou and Gao Zhu and Kun Zhou and Biwei Huang},
year = {2026},
eprint = {2610.34649},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.34649},
}
Argues VLA progress now depends more on data engines and evaluation than on architectures; finds a fidelity-versus-cost trade-off in datasets and gaps in compositional and long-horizon benchmarks.认为 VLA 的进展更取决于数据引擎和评测，而不是模型架构；指出数据集存在保真度与成本的权衡，评测基准在组合任务和长时程任务上仍有缺口。
@misc{wang2026vision,
title = {{Vision-Language-Action in Robotics: A Survey of Datasets, Benchmarks, and Data Engines}},
author = {Ziyao Wang and Bingying Wang and Hanrong Zhang and Tingting Du and Tianyang Chen and Guoheng Sun and Yexiao He and Zheyu Shen and Wanghao Ye and Ang Li},
year = {2026},
eprint = {2604.23001},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2604.23001},
}
Wraps each model once and each benchmark once, then schedules simulator workers and policy replicas as separate pools; reports evaluation 4.45x to 6.16x faster than the official StarVLA evaluator on one GPU.模型和基准各封装一次，把仿真 worker 与策略副本分成两个池调度；报告单卡评测比官方 StarVLA 评测器快 4.45 到 6.16 倍。
@misc{wu2026roboscope,
title = {{RoboScope Kit: Unified Interfaces for Fast and Scalable Embodied-AI Evaluation}},
author = {Deng-Ke Wu and Yifan Jin and Yida Yang and Lei Qian and Zhaoye Zhou and Hongru Zhang and Kai Ye and Shuyuan Hu and Yiran Mao and Zewei Ye and Bo Zhao},
year = {2026},
eprint = {2611.03850},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.03850},
}
A simulation benchmark for RL fine-tuning of VLAs without frequent resets; finds naive fine-tuning fails from object and end-effector shifts, and that targeted reset allocation and bias fixes mitigate it.面向无频繁重置场景下 VLA 强化学习微调的仿真基准；发现直接微调会因物体和末端位姿偏移而失败，定向分配重置并处理预训练偏差可以缓解。
Benchmark showing VLA success often hides weak instruction following: scenes allow several valid tasks so language matters. Nine VLA and WAM policies drop sharply beyond the base level.揭示 VLA 的高成功率常掩盖薄弱的指令遵循：场景支持多种合法任务，必须依赖语言。九个 VLA 与 WAM 策略在基础级以外的测试中明显下降。
@misc{guo2026robofollow,
title = {{RoboFollow: Unveiling the Instruction Following Mirage in Embodied Agents}},
author = {Chang Guo and Yukun Xie and Bohan Tan and Zheng Chang and Zhaokai Yin and Qianli Ma and Yingqiao Wang and Chao Liang and Zhipeng Zhang},
year = {2026},
eprint = {2609.25636},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.25636},
}
Compresses action chunks with a DCT-based tokenizer so autoregressive VLAs can learn high-frequency, dexterous control.用基于 DCT 的分词器压缩动作块，使自回归 VLA 能学习高频灵巧控制。
@misc{pertsch2025fast,
title = {{FAST: Efficient Action Tokenization for Vision-Language-Action Models}},
author = {Karl Pertsch and Kyle Stachowicz and Brian Ichter and Danny Driess and Suraj Nair and Quan Vuong and Oier Mees and Chelsea Finn and Sergey Levine},
year = {2025},
eprint = {2501.09747},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2501.09747},
}
@misc{team2025gemini,
title = {{Gemini Robotics: Bringing AI into the Physical World}},
author = {Gemini Robotics Team and Saminda Abeyruwan and Joshua Ainslie and Jean-Baptiste Alayrac and Montserrat Gonzalez Arenas and Travis Armstrong and Ashwin Balakrishna and Robert Baruch and Maria Bauza and Michiel Blokzijl and Steven Bohez and Konstantinos Bousmalis and Anthony Brohan and Thomas Buschmann and Arunkumar Byravan and Serkan Cabi and Ken Caluwaerts and Federico Casarini and Oscar Chang and Jose Enrique Chen and Xi Chen and Hao-Tien Lewis Chiang and Krzysztof Choromanski and David D'Ambrosio and Sudeep Dasari and Todor Davchev and Coline Devin and Norman Di Palo and Tianli Ding and Adil Dostmohamed and Danny Driess and Yilun Du and Debidatta Dwibedi and Michael Elabd and Claudio Fantacci and Cody Fong and Erik Frey and Chuyuan Fu and Marissa Giustina and Keerthana Gopalakrishnan and Laura Graesser and Leonard Hasenclever and Nicolas Heess and Brandon Hernaez and Alexander Herzog and R. Alex Hofer and Jan Humplik and Atil Iscen and Mithun George Jacob and Deepali Jain and others},
year = {2025},
eprint = {2503.20020},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2503.20020},
}
A high-level VLM reads open-ended prompts and mid-task corrections and turns them into short language commands for a low-level π0 policy.高层 VLM 理解开放式指令与执行中的纠正，并将其转化为简短语言指令交给底层 π0 策略执行。
@misc{shi2025hi,
title = {{Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Models}},
author = {Lucy Xiaoyang Shi and Brian Ichter and Michael Equi and Liyiming Ke and Karl Pertsch and Quan Vuong and James Tanner and Anna Walling and Haohuan Wang and Niccolo Fusai and Adrian Li-Bell and Danny Driess and Lachy Groom and Sergey Levine and Chelsea Finn},
year = {2025},
eprint = {2502.19417},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.19417},
}
ByteDance Seed's generalist VLA: robot imitation data co-trained with web vision-language data, adapted quickly with VR-collected human trajectories; shown on long-horizon bimanual mobile tasks.字节跳动 Seed 的通用 VLA——机器人数据模仿学习并与网络视觉-语言数据联合训练，再用 VR 采集的人类轨迹快速适配；在 ByteMini 双臂移动机器人上展示长时程与衣物操作任务。
@misc{cheang2025gr,
title = {{GR-3 Technical Report}},
author = {Chilam Cheang and Sijin Chen and Zhongren Cui and Yingdong Hu and Liqun Huang and Tao Kong and Hang Li and Yifeng Li and Yuxiao Liu and Xiao Ma and Hao Niu and Wenxuan Ou and Wanli Peng and Zeyu Ren and Haixin Shi and Jiawen Tian and Hongtao Wu and Xin Xiao and Yuyang Xiao and Jiafeng Xu and Yichu Yang},
year = {2025},
eprint = {2507.15493},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2507.15493},
}
Adds 3D-aware position encoding to visual tokens and adaptive action grids, pretrained on 1.1M real robot episodes; new robot setups are handled by re-discretizing the grids.为视觉 token 加入 3D 位置编码，并用自适应动作网格表示动作，在 110 万条真机轨迹上预训练；迁移到新机器人时只需重新划分动作网格。
@misc{qu2025spatialvla,
title = {{SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Model}},
author = {Delin Qu and Haoming Song and Qizhi Chen and Yuanqi Yao and Xinyi Ye and Yan Ding and Zhigang Wang and JiaYuan Gu and Bin Zhao and Dong Wang and Xuelong Li},
year = {2025},
eprint = {2501.15830},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2501.15830},
}
An open 7B VLA that first outputs depth-aware perception tokens, then an editable 2D trajectory, then actions; reports 70.5% zero-shot on SimplerEnv and 86.6% on LIBERO, with weights, code and data released.开源 7B VLA，先输出深度感知 token，再生成可编辑的二维轨迹，最后输出动作；论文报告 SimplerEnv 零样本 70.5%、LIBERO 86.6%，并开放权重、代码和数据。
@misc{lee2025molmoact,
title = {{MolmoAct: Action Reasoning Models that can Reason in Space}},
author = {Jason Lee and Jiafei Duan and Haoquan Fang and Yuquan Deng and Shuo Liu and Boyang Li and Bohan Fang and Jieyu Zhang and Yi Ru Wang and Sangho Lee and Winson Han and Wilbert Pumacay and Angelica Wu and Rose Hendrix and Karen Farley and Eli VanderBilt and Ali Farhadi and Dieter Fox and Ranjay Krishna},
year = {2025},
eprint = {2508.07917},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2508.07917},
}
Pretrains a grasping VLA almost only on SynGrasp-1B, a billion-frame photorealistic sim dataset, co-trained with web data; reports zero-shot sim-to-real open-vocabulary grasping and few-shot adaptation.几乎只用十亿帧级的照片级仿真抓取数据集 SynGrasp-1B（并联合网络语义数据）预训练抓取 VLA；论文报告可零样本从仿真迁移到真实开放词汇抓取，并支持少样本适配。
@misc{deng2025graspvla,
title = {{GraspVLA: a Grasping Foundation Model Pre-trained on Billion-scale Synthetic Action Data}},
author = {Shengliang Deng and Mi Yan and Songlin Wei and Haixin Ma and Yuxin Yang and Jiayi Chen and Zhiqi Zhang and Taoyu Yang and Xuheng Zhang and Wenhao Zhang and Heming Cui and Zhizheng Zhang and He Wang},
year = {2025},
eprint = {2505.03233},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.03233},
}
Gives each data source its own small set of learnable soft-prompt embeddings, so one plain-Transformer flow-matching policy (0.9B) trains across embodiments; evaluated on 6 sim benchmarks and 3 real robots.为每个数据来源配一组可学习的软提示嵌入，使一个基于标准 Transformer 的流匹配策略（0.9B）能跨本体联合训练；在 6 个仿真基准和 3 台真实机器人上评测。
@misc{zheng2025x,
title = {{X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Model}},
author = {Jinliang Zheng and Jianxiong Li and Zhihao Wang and Dongxiu Liu and Xirui Kang and Yuchun Feng and Yinan Zheng and Jiayin Zou and Yilun Chen and Jia Zeng and Ya-Qin Zhang and Jiangmiao Pang and Jingjing Liu and Tai Wang and Xianyuan Zhan},
year = {2025},
eprint = {2510.10274},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.10274},
}
Pretrains spatial grounding on 2.3M samples ('where to act'), then post-trains actions with spatial prompts ('how to act'); reports +14.6% on SimplerEnv Google Robot over the same model without spatial guidance.先在 230 万条空间推理样本上做空间定位预训练（"在哪里做"），再用空间提示做动作后训练（"怎么做"）；论文报告在 SimplerEnv Google Robot 上比无空间引导的同一模型高 14.6%。
@misc{chen2025internvla,
title = {{InternVLA-M1: A Spatially Guided Vision-Language-Action Framework for Generalist Robot Policy}},
author = {Xinyi Chen and Yilun Chen and Yanwei Fu and Ning Gao and Jiaya Jia and Weiyang Jin and Hao Li and Yao Mu and Jiangmiao Pang and Yu Qiao and Yang Tian and Bin Wang and Bolun Wang and Fangjing Wang and Hanqing Wang and Tai Wang and Ziqin Wang and Xueyuan Wei and Chao Wu and Shuai Yang and Jinhui Ye and Junqiu Yu and Jia Zeng and Jingjing Zhang and Jinyu Zhang and Shi Zhang and Feng Zheng and Bowen Zhou and Yangkun Zhu},
year = {2025},
eprint = {2510.13778},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.13778},
}
A dexterous VLA pretrained on human hand motion from mocap, VR and RGB video ('physical instruction tuning'), with part-level motion tokens, then aligned to 3D space and post-trained for real robot hands.灵巧手 VLA：先在动捕、VR 与普通视频中的人手动作上预训练（"物理指令微调"），使用部件级动作 token，再做三维空间对齐，并针对真实机器人灵巧手后训练。
@misc{luo2025being,
title = {{Being-H0: Vision-Language-Action Pretraining from Large-Scale Human Videos}},
author = {Hao Luo and Yicheng Feng and Wanpeng Zhang and Sipeng Zheng and Ye Wang and Haoqi Yuan and Jiazheng Liu and Chaoyi Xu and Qin Jin and Zongqing Lu},
year = {2025},
eprint = {2507.15597},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2507.15597},
}
A hierarchical VLA in which a fine-tuned VLM draws a coarse 2D path for the end effector on the image and a low-level 3D policy follows it, so the VLM can be trained on cheap off-domain data such as sim or videos.分层 VLA：微调后的 VLM 在图像上画出末端执行器的粗略二维路径，再由底层三维策略跟随执行；因此 VLM 可以用仿真、视频等低成本的非本域数据训练。
@misc{li2025hamster,
title = {{HAMSTER: Hierarchical Action Models for Open-World Robot Manipulation}},
author = {Yi Li and Yuquan Deng and Jesse Zhang and Joel Jang and Marius Memmel and Raymond Yu and Caelan Reed Garrett and Fabio Ramos and Dieter Fox and Anqi Li and Abhishek Gupta and Ankit Goyal},
year = {2025},
eprint = {2502.05485},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.05485},
}
One model that decides when to think and when to act: it reasons at key moments (subtask done, error, human input) and otherwise acts on its latest reasoning; co-trained with synthetic reasoning-rich VL data.单一模型自主决定何时“想”、何时“做”：在子任务完成、出错或人类插话等关键时刻进行推理，其余时间依据最新推理结果生成动作；并与合成的推理型视觉-语言数据共同训练。
@misc{lin2025onetwovla,
title = {{OneTwoVLA: A Unified Vision-Language-Action Model with Adaptive Reasoning}},
author = {Fanqi Lin and Ruiqian Nai and Yingdong Hu and Jiacheng You and Junming Zhao and Yang Gao},
year = {2025},
eprint = {2505.11917},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.11917},
}
Pairs a 2B Qwen2-VL with a ~1B plug-in diffusion action expert with one head per robot, trained in three stages (cross-embodiment, alignment, task); reports shirt folding without task-specific tuning and 97.3 on LIBERO.将 2B 的 Qwen2-VL 与约 1B 参数、按机器人分头输出的可插拔扩散动作专家结合，分三阶段训练（跨本体预训练、本体对齐、任务适配）；报告无需任务专门微调即可叠衬衫，LIBERO 平均 97.3。
@misc{wen2025dexvla,
title = {{DexVLA: Vision-Language Model with Plug-In Diffusion Expert for General Robot Control}},
author = {Junjie Wen and Yichen Zhu and Jinming Li and Zhibin Tang and Chaomin Shen and Feifei Feng},
year = {2025},
eprint = {2502.05855},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.05855},
}
Keeps a VLA able to chat: phased alignment training plus a mixture-of-experts split to fight forgetting and task interference; reports strong VQA/MMMU scores and beats OpenVLA on 25 real tasks.让 VLA 保留“对话”能力：通过分阶段对齐训练和混合专家结构缓解遗忘与任务干扰；论文报告在保持 VQA/MMMU 等多模态理解成绩的同时，在 25 个真实操作任务上优于 OpenVLA。
@misc{zhou2025chatvla,
title = {{ChatVLA: Unified Multimodal Understanding and Robot Control with Vision-Language-Action Model}},
author = {Zhongyi Zhou and Yichen Zhu and Minjie Zhu and Junjie Wen and Ning Liu and Zhiyuan Xu and Weibin Meng and Ran Cheng and Yaxin Peng and Chaomin Shen and Feifei Feng},
year = {2025},
eprint = {2502.14420},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.14420},
}
Dual-system VLA: a value-query head scores sampled action-chunk candidates (System 2, ~4 Hz) and a light cascaded denoiser refines it (System 1, ~90 Hz); reports gains over π0 on LIBERO, SIMPLER and real robots.双系统 VLA：价值查询头对多组候选动作块打分并择优（System 2，约 4 Hz），轻量级级联去噪策略再将其细化为流畅动作（System 1，约 90 Hz）；论文报告在 LIBERO、SIMPLER 和真机上均优于 π0。
@misc{song2025hume,
title = {{Hume: Introducing System-2 Thinking in Visual-Language-Action Model}},
author = {Haoming Song and Delin Qu and Yuanqi Yao and Qizhi Chen and Qi Lv and Yiwen Tang and Modi Shi and Guanghui Ren and Maoqing Yao and Bin Zhao and Dong Wang and Xuelong Li},
year = {2025},
eprint = {2505.21432},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.21432},
}
Builds a bimanual VLA by composing two copies of a pretrained single-arm VLA instead of pretraining on bimanual data; reports beating a similar-size RDT-1B and narrowing the gap to π0 on real and sim tasks.不做双臂预训练，而是把两份预训练单臂 VLA 组合成协同的双臂 VLA；论文报告在真实与仿真双臂任务上优于同规模 RDT-1B，并缩小了与依赖大量双臂数据的 π0 的差距。
@misc{im2025twinvla,
title = {{TwinVLA: Data-Efficient Bimanual Manipulation with Twin Single-Arm Vision-Language-Action Models}},
author = {Hokyun Im and Euijin Jeong and Andrey Kolobov and Jianlong Fu and Youngwoon Lee},
year = {2025},
eprint = {2511.05275},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2511.05275},
}
@misc{wang2025vq,
title = {{VQ-VLA: Improving Vision-Language-Action Models via Scaling Vector-Quantized Action Tokenizers}},
author = {Yating Wang and Haoyi Zhu and Mingyu Liu and Jiange Yang and Hao-Shu Fang and Tong He},
year = {2025},
eprint = {2507.01016},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2507.01016},
}
~500 h / 100K trajectories on one mobile bimanual robot (R1 Lite) across 50 real scenes, plus G0: a VLM planner over a flow-matching VLA; reports that single-embodiment pretraining matters most for few-shot transfer.在单一移动双臂机器人（R1 Lite）上采集约 500 小时、10 万条轨迹、覆盖 50 个真实场景，并提出 G0：VLM 规划器 + 流匹配 VLA 执行器的双系统模型；论文报告单本体预训练对少样本迁移最关键。
@misc{jiang2025galaxea,
title = {{Galaxea Open-World Dataset and G0 Dual-System VLA Model}},
author = {Tao Jiang and Tianyuan Yuan and Yicheng Liu and Chenhao Lu and Jianning Cui and Xiao Liu and Shuiqi Cheng and Jiyang Gao and Huazhe Xu and Hang Zhao},
year = {2025},
eprint = {2509.00576},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2509.00576},
}
Open-source 'brain–cerebellum' stack: a RoboBrain MLLM plans and assigns tasks across heterogeneous robots, per-robot skill libraries execute, and a shared spatial/temporal/robot memory keeps them in sync.开源的“大脑–小脑”分层系统：RoboBrain 多模态大模型负责跨异构机器人的任务规划与分配，各机器人上的技能库负责执行，共享的空间/时间/本体记忆保持多机协同。
@misc{tan2025roboos,
title = {{RoboOS: A Hierarchical Embodied Framework for Cross-Embodiment and Multi-Agent Collaboration}},
author = {Huajie Tan and Xiaoshuai Hao and Cheng Chi and Minglan Lin and Yaoxu Lyu and Mingyu Cao and Dong Liang and Zhuo Chen and Mengsi Lyu and Cheng Peng and Chenrui He and Yulong Ao and Yonghua Lin and Pengwei Wang and Zhongyuan Wang and Shanghang Zhang},
year = {2025},
eprint = {2505.03673},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.03673},
}
CoRL 2026 Spotlight. A fully native 1-bit VLA with ternary weights, built on BitNet b1.58 2B4T, designed for edge devices instead of relying on post-hoc compression.CoRL 2026 Spotlight。原生 1-bit VLA，权重三值化，基于 BitNet b1.58 2B4T，面向边缘设备设计，而非依赖事后压缩。
@misc{wang2025bitvla,
title = {{BitVLA: 1-bit Vision-Language-Action Models for Robotics Manipulation}},
author = {Hongyu Wang and Chuyan Xiong and Ruiping Wang and Xilin Chen},
year = {2025},
eprint = {2506.07530},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.07530},
}
Tech report on a multi-embodiment VLA plus the ER 1.5 reasoning model: Motion Transfer learns across varied robot data, and actions interleave with natural-language reasoning so the robot can think before acting.技术报告：跨本体 VLA 与具身推理模型 ER 1.5。Motion Transfer 让模型从不同机器人的数据中学习，动作与自然语言推理交替进行，使机器人先想后做。
@misc{team2025gemini,
title = {{Gemini Robotics 1.5: Pushing the Frontier of Generalist Robots with Advanced Embodied Reasoning, Thinking, and Motion Transfer}},
author = {Gemini Robotics Team and Abbas Abdolmaleki and Saminda Abeyruwan and Joshua Ainslie and Jean-Baptiste Alayrac and Montserrat Gonzalez Arenas and Ashwin Balakrishna and Nathan Batchelor and Alex Bewley and Jeff Bingham and Michael Bloesch and Konstantinos Bousmalis and Philemon Brakel and Anthony Brohan and Thomas Buschmann and Arunkumar Byravan and Serkan Cabi and Ken Caluwaerts and Federico Casarini and Christine Chan and Oscar Chang and London Chappellet-Volpini and Jose Enrique Chen and Xi Chen and Hao-Tien Lewis Chiang and Krzysztof Choromanski and Adrian Collister and David B. D'Ambrosio and Sudeep Dasari and Todor Davchev and Meet Kirankumar Dave and Coline Devin and Norman Di Palo and Tianli Ding and Carl Doersch and Adil Dostmohamed and Yilun Du and Debidatta Dwibedi and Sathish Thoppay Egambaram and Michael Elabd and Tom Erez and Xiaolin Fang and Claudio Fantacci and Cody Fong and Erik Frey and Chuyuan Fu and Ruiqi Gao and Marissa Giustina and Keerthana Gopalakrishnan and Laura Graesser and others},
year = {2025},
eprint = {2510.03342},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.03342},
}
Survey that reads every VLA as one pipeline producing ever more actionable tokens, and reviews eight token types from language and code to trajectories, latents, raw actions and reasoning.综述：把所有 VLA 看作生成越来越可执行 token 的同一流程，并梳理语言、代码、可供性、轨迹、目标状态、潜表示、原始动作和推理八类 token。
@misc{zhong2025survey,
title = {{A Survey on Vision-Language-Action Models: An Action Tokenization Perspective}},
author = {Yifan Zhong and Fengshuo Bai and Shaofei Cai and Xuchuan Huang and Zhang Chen and Xiaowei Zhang and Yuanfei Wang and Shaoyang Guo and Tianrui Guan and Ka Nam Lui and Zhiquan Qi and Yitao Liang and Yuanpei Chen and Yaodong Yang},
year = {2025},
eprint = {2507.01925},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2507.01925},
}
Survey of over three hundred studies with a taxonomy of autoregression, diffusion, reinforcement, hybrid and specialized VLAs, plus the datasets, benchmarks and simulators used to test them.综述三百多项研究，把 VLA 分为自回归、扩散、强化学习、混合和专用几类，并介绍常用数据集、评测基准和仿真平台。
@misc{zhang2025pure,
title = {{Pure Vision Language Action (VLA) Models: A Comprehensive Survey}},
author = {Dapeng Zhang and Jing Sun and Chenghui Hu and Xiaoyan Wu and Zhenlong Yuan and Rui Zhou and Fei Shen and Qingguo Zhou},
year = {2025},
eprint = {2509.19012},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2509.19012},
}
Speeds up VLA inference on edge devices by dropping the autoregressive end-effector prediction, reported as a 7x speedup, and by using small language models.通过去掉末端位姿的自回归预测（报告提速 7 倍）并使用小型语言模型，加快 VLA 在边缘设备上的推理。
@misc{budzianowski2025edgevla,
title = {{EdgeVLA: Efficient Vision-Language-Action Models}},
author = {Paweł Budzianowski and Wesley Maa and Matthew Freed and Jingxiang Mo and Winston Hsiao and Aaron Xie and Tomasz Młoduchowski and Viraj Tipnis and Benjamin Bolte},
year = {2025},
eprint = {2507.14049},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2507.14049},
}
Adds point cloud input to pretrained VLAs without retraining: freezes the action expert and injects 3D features only into blocks that a skip-block analysis finds less useful.无需重训即可为预训练 VLA 加入点云输入：冻结动作专家，只向跳块分析认定作用较小的模块注入 3D 特征。
@misc{li2025pointvla,
title = {{PointVLA: Injecting the 3D World into Vision-Language-Action Models}},
author = {Chengmeng Li and Junjie Wen and Yan Peng and Yaxin Peng and Feifei Feng and Yichen Zhu},
year = {2025},
eprint = {2503.07511},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2503.07511},
}
A survey of efficient VLAs across the model, training and data pipeline, with a unified taxonomy of work on real-time inference, training cost and scalable data collection.综述高效 VLA，贯穿模型、训练和数据全流程，用统一分类梳理实时推理、训练成本和可扩展数据采集方面的工作。
@misc{yu2025survey,
title = {{A Survey on Efficient Vision-Language-Action Models}},
author = {Zhaoshu Yu and Bo Wang and Pengpeng Zeng and Haonan Zhang and Ji Zhang and Zheng Wang and Lianli Gao and Jingkuan Song and Nicu Sebe and Heng Tao Shen},
year = {2025},
eprint = {2510.24795},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2510.24795},
}
Trains a multimodal language model for task planning, object affordance and trajectory prediction, using ShareRobot, a dataset labelled for all three.训练多模态语言模型完成任务规划、物体可供性和轨迹预测，并配套标注了这三类信息的 ShareRobot 数据集。
@misc{ji2025robobrain,
title = {{RoboBrain: A Unified Brain Model for Robotic Manipulation from Abstract to Concrete}},
author = {Yuheng Ji and Huajie Tan and Jiayu Shi and Xiaoshuai Hao and Yuan Zhang and Hengyuan Zhang and Pengwei Wang and Mengdi Zhao and Yao Mu and Pengju An and Xinda Xue and Qinghang Su and Huaihai Lyu and Xiaolong Zheng and Jiaming Liu and Zhongyuan Wang and Shanghang Zhang},
year = {2025},
eprint = {2502.21257},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.21257},
}
Being-0: a humanoid agent that coordinates a foundation model, a vision-language model and modular skills to limit compounding errors on long tasks.Being-0：协调基础模型、视觉语言模型与模块化技能的人形智能体，以减少长任务中的误差累积。
@misc{yuan2025being,
title = {{Being-0: A Humanoid Robotic Agent with Vision-Language Models and Modular Skills}},
author = {Haoqi Yuan and Yu Bai and Yuhui Fu and Bohan Zhou and Yicheng Feng and Xinrun Xu and Yi Zhan and Börje F. Karlsson and Zongqing Lu},
year = {2025},
eprint = {2503.12533},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2503.12533},
}
@misc{xue2025leverb,
title = {{LeVERB: Humanoid Whole-Body Control with Latent Vision-Language Instruction}},
author = {Haoru Xue and Xiaoyu Huang and Dantong Niu and Qiayuan Liao and Thomas Kragerud and Jan Tommy Gravdahl and Xue Bin Peng and Guanya Shi and Trevor Darrell and Koushil Sreenath and Shankar Sastry},
year = {2025},
eprint = {2506.13751},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.13751},
}
Trains VLAs on egocentric human videos, predicting wrist and hand actions, because human data scales beyond robot data and is richer in scenes and tasks.用第一视角人类视频训练 VLA，预测手腕与手部动作；人类数据规模更大，场景和任务也更丰富。
@misc{yang2025egovla,
title = {{EgoVLA: Learning Vision-Language-Action Models from Egocentric Human Videos}},
author = {Ruihan Yang and Qinxi Yu and Yecheng Wu and Rui Yan and Borui Li and An-Chieh Cheng and Xueyan Zou and Yunhao Fang and Xuxin Cheng and Ri-Zhao Qiu and Hongxu Yin and Sifei Liu and Song Han and Yao Lu and Xiaolong Wang},
year = {2025},
eprint = {2507.12440},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2507.12440},
}
VITRA pretrains VLA models on unscripted in-the-wild human videos, treating the human hand as a dexterous end-effector, with no annotations needed.VITRA 用无脚本的真实人类活动视频预训练 VLA，把人手视为灵巧末端执行器，无需人工标注。
@misc{li2025scalable,
title = {{Scalable Vision-Language-Action Model Pretraining for Robotic Manipulation with Real-Life Human Activity Videos}},
author = {Qixiu Li and Yu Deng and Yaobo Liang and Lin Luo and Lei Zhou and Chengtang Yao and Lingqi Zeng and Zhiyuan Feng and Huizhi Liang and Sicheng Xu and Yizhong Zhang and Xi Chen and Hao Chen and Lily Sun and Dong Chen and Jiaolong Yang and Baining Guo},
year = {2025},
eprint = {2510.21571},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.21571},
}
Empirical study of whether RL fine-tuning generalises better than supervised fine-tuning, which suffers compounding errors under distribution shift.实证研究强化学习微调是否比监督微调泛化更好；后者在分布偏移下易累积误差。
@misc{liu2025what,
title = {{What Can RL Bring to VLA Generalization? An Empirical Study}},
author = {Jijia Liu and Feng Gao and Bingwen Wei and Xinlei Chen and Qingmin Liao and Yi Wu and Chao Yu and Yu Wang},
year = {2025},
eprint = {2505.19789},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2505.19789},
}
EMMA: a diffusion-transformer video editor (DreamTransfer) changes foreground, background and lighting with multi-view consistency, and AdaMix weights real and generated samples by policy performance.EMMA：扩散 Transformer 视频编辑器 DreamTransfer 在保持多视角一致的前提下修改前景、背景与光照，AdaMix 再按策略表现为真实与生成样本自适应加权。
@misc{dong2025emma,
title = {{EMMA: Enhancing Robotic Control Robustness through a Generative Visual Model}},
author = {Zhehao Dong and Xiaofeng Wang and Zheng Zhu and Yirui Wang and Yang Wang and Yukun Zhou and Boyuan Wang and Chaojun Ni and Runqi Ouyang and Wenkang Qin and Xinze Chen and Yun Ye and Guan Huang and Zhen Lu and Yue Yang},
year = {2025},
eprint = {2509.22407},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2509.22407},
}
No end-to-end training: a VLM coding agent writes robot programs from a library of perception, planning and control tools in a plan-act-replan loop; reports matching or beating VLAs zero-shot on tabletop tasks.不训练端到端策略，而让 VLM 作为编程智能体，调用感知、规划与控制工具库编写机器人程序，按“规划-执行-重规划”循环运行；在桌面任务上零样本达到或超过 VLA 模型，并可完成移动操作。
@misc{shi2025maestro,
title = {{MAESTRO: Orchestrating Robotics Modules with Coding Agents for Zero-Shot Generalist Robots}},
author = {Junyao Shi and Rujia Yang and Kaitian Chao and Selina Bingqing Wan and Yifei Shao and Jiahui Lei and Jianing Qian and Long Le and Pratik Chaudhari and Kostas Daniilidis and Chuan Wen and Dinesh Jayaraman},
year = {2025},
eprint = {2511.00917},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2511.00917},
}
A pretrained VLM plans and a diffusion controller grasps with a dexterous hand, working on domain-invariant features from foundation models; reports 90%+ grasp success across thousands of unseen cluttered scenes.由预训练 VLM 做高层规划、扩散控制器驱动灵巧手抓取，底层输入为基础模型提取的域不变特征；报告在数千个未见过的杂乱场景中抓取成功率超过 90%。
@misc{zhong2025dexgraspvla,
title = {{DexGraspVLA: A Vision-Language-Action Framework Towards General Dexterous Grasping}},
author = {Yifan Zhong and Xuchuan Huang and Ruochong Li and Ceyao Zhang and Zhang Chen and Tianrui Guan and Fanlian Zeng and Ka Num Lui and Yuyao Ye and Yitao Liang and Yaodong Yang and Yuanpei Chen},
year = {2025},
eprint = {2502.20900},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2502.20900},
}
Uses egocentric human demonstrations as cross-embodiment data for humanoid manipulation policies, narrowing the human-robot gap in data and model design.把第一视角人类示教作为人形操作策略的跨形态数据，从数据和模型两方面缩小人与机器人的差距。
@misc{qiu2025humanoid,
title = {{Humanoid Policy ~ Human Policy}},
author = {Ri-Zhao Qiu and Shiqi Yang and Xuxin Cheng and Chaitanya Chawla and Jialong Li and Tairan He and Ge Yan and David J. Yoon and Ryan Hoque and Lars Paulsen and Ge Yang and Jian Zhang and Sha Yi and Guanya Shi and Xiaolong Wang},
year = {2025},
eprint = {2503.13441},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2503.13441},
}
H-RDT pretrains a diffusion transformer on human manipulation data to work around the morphology mismatch of cross-robot datasets.H-RDT 用人类操作数据预训练扩散 Transformer，规避跨机器人数据集的形态差异问题。
@misc{bi2025h,
title = {{H-RDT: Human Manipulation Enhanced Bimanual Robotic Manipulation}},
author = {Hongzhe Bi and Lingxuan Wu and Tianwei Lin and Hengkai Tan and Zhizhong Su and Hang Su and Jun Zhu},
year = {2025},
eprint = {2507.23523},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2507.23523},
}
Latent VLA that learns from cheap action-free egocentric video, paired with an RL controller for advancing, turning and squatting; reports a 21.3% improvement over a baseline on the AgiBot X2.潜空间 VLA 从低成本、无动作标注的第一视角视频学习，再配合面向前进、转向、下蹲的强化学习控制器；在智元 X2 上报告比基线提升 21.3%。
@misc{jiang2025wholebodyvla,
title = {{WholeBodyVLA: Towards Unified Latent VLA for Whole-Body Loco-Manipulation Control}},
author = {Haoran Jiang and Jin Chen and Qingwen Bu and Li Chen and Modi Shi and Yanjie Zhang and Delong Li and Chuanzhe Suo and Chuang Wang and Zhihui Peng and Hongyang Li},
year = {2025},
eprint = {2512.11047},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2512.11047},
}
Adds a few tokens to a VLA-style diffusion transformer and aligns them with latents of future observations, so the policy anticipates the future without a separate world model.只在 VLA 式扩散 Transformer 中加入少量 token，并让它们对齐未来观测的潜表示，使策略无需独立世界模型就能预判未来。
@misc{zheng2025flare,
title = {{FLARE: Robot Learning with Implicit World Modeling}},
author = {Ruijie Zheng and Jing Wang and Scott Reed and Johan Bjorck and Yu Fang and Fengyuan Hu and Joel Jang and Kaushil Kundalia and Zongyu Lin and Loic Magne and Avnish Narayan and You Liang Tan and Guanzhi Wang and Qi Wang and Jiannan Xiang and Yinzhen Xu and Seonghyeon Ye and Jan Kautz and Furong Huang and Yuke Zhu and Linxi Fan},
year = {2025},
eprint = {2505.15659},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.15659},
}
Report on an online system for testing many robot policies on many real tasks, plus a study of recent VLAs on its first benchmark, Table30.介绍一套在线系统，用于在真机上大规模评测多种机器人策略，并报告近期 VLA 在首个基准 Table30 上的表现。
@misc{yakefu2025robochallenge,
title = {{RoboChallenge: Large-scale Real-robot Evaluation of Embodied Policies}},
author = {Adina Yakefu and Bin Xie and Chongyang Xu and Enwen Zhang and Erjin Zhou and Fan Jia and Haitao Yang and Haoqiang Fan and Haowei Zhang and Hongyang Peng and Jing Tan and Junwen Huang and Kai Liu and Kaixin Liu and Kefan Gu and Qinglun Zhang and Ruitao Zhang and Saike Huang and Shen Cheng and Shuaicheng Liu and Tiancai Wang and Tiezhen Wang and Wei Sun and Wenbin Tang and Yajun Wei and Yang Chen and Youqiang Gui and Yucheng Zhao and Yunchao Ma and Yunfei Wei and Yunhuan Yang and Yutong Guo and Ze Chen and Zhengyuan Du and Ziheng Zhang and Ziming Liu and Ziwei Yan},
year = {2025},
eprint = {2510.17950},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.17950},
}
Instead of emitting action tokens from the VLM, a separate diffusion action transformer is conditioned on the VLM's "cognition" feature; reports clear gains over OpenVLA in sim and on real robots.不再让 VLM 直接输出动作 token，而是用独立的扩散动作 Transformer 以 VLM 的“认知”特征为条件生成动作；报告在仿真和真机上明显优于 OpenVLA。
@misc{li2024cogact,
title = {{CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation}},
author = {Qixiu Li and Yaobo Liang and Zeyu Wang and Lin Luo and Xi Chen and Mozheng Liao and Fangyun Wei and Yu Deng and Sicheng Xu and Yizhong Zhang and Xiaofan Wang and Bei Liu and Jianlong Fu and Jianmin Bao and Dong Chen and Yuanchun Shi and Jiaolong Yang and Baining Guo},
year = {2024},
eprint = {2411.19650},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2411.19650},
}
An early survey of VLA models, sorting them into component work (vision, language, world models), low-level control policies and high-level task planners, plus datasets, simulators and benchmarks; companion paper list.较早的 VLA 综述：把相关工作分为组件研究（视觉、语言、世界模型等）、底层控制策略与高层任务规划器，并汇总数据集、仿真器与基准；附配套论文列表。
@misc{ma2024survey,
title = {{A Survey on Vision-Language-Action Models for Embodied AI}},
author = {Yueen Ma and Zixing Song and Yuzheng Zhuang and Jianye Hao and Irwin King},
year = {2024},
eprint = {2405.14093},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2405.14093},
}
Broad embodied-AI survey — robots, simulators, embodied perception, interaction, agents and sim-to-real — with the role of multimodal and world models; backed by the widely used HCPLab paper list.覆盖面很广的具身智能综述：机器人本体、仿真器、具身感知、交互、具身智能体与 sim-to-real，并讨论多模态大模型与世界模型的作用；配套中山大学 HCPLab 维护的常用论文列表。
@misc{liu2024aligning,
title = {{Aligning Cyber Space with Physical World: A Comprehensive Survey on Embodied AI}},
author = {Yang Liu and Weixing Chen and Yongjie Bai and Xiaodan Liang and Guanbin Li and Wen Gao and Liang Lin},
year = {2024},
eprint = {2407.06886},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2407.06886},
}
HPT: per-robot stems turn proprioception and vision into a few tokens for one shared transformer trunk, pretrained on 52 datasets (robot, sim, human video); reports 20%+ gains on unseen tasks.HPT：为每种机器人设计“输入头”，把本体感知与视觉映射成短 token 序列，送入共享的 Transformer 主干；在 52 个数据集（真机、仿真、人类视频）上预训练，报告在未见任务上微调提升 20% 以上。
LAPA pretrains a VLA without action labels: it learns discrete latent actions between video frames, predicts them from image + text, then maps them to robot actions; reports beating OpenVLA on real tasks.LAPA 无需动作标注即可预训练 VLA：先从视频相邻帧间学习离散隐动作，再训练模型由图像和指令预测隐动作，最后用少量机器人数据映射到真实动作；报告在真机任务上优于 OpenVLA。
@misc{ye2024latent,
title = {{Latent Action Pretraining from Videos}},
author = {Seonghyeon Ye and Joel Jang and Byeongguk Jeon and Sejune Joo and Jianwei Yang and Baolin Peng and Ajay Mandlekar and Reuben Tan and Yu-Wei Chao and Bill Yuchen Lin and Lars Liden and Kimin Lee and Jianfeng Gao and Luke Zettlemoyer and Dieter Fox and Minjoon Seo},
year = {2024},
eprint = {2410.11758},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2410.11758},
}
A 3D-aware LLM with scene, object and action tokens, aligned with diffusion models so it can imagine goal images and point clouds before planning actions; trained on ~2M 3D-language-action pairs the authors assembled.具备三维感知的大语言模型，引入场景、物体与动作交互 token，并与扩散模型对齐，可在规划动作前“想象”目标图像和点云；训练数据为作者整理的约 200 万条三维-语言-动作样本。
@misc{zhen2024d,
title = {{3D-VLA: A 3D Vision-Language-Action Generative World Model}},
author = {Haoyu Zhen and Xiaowen Qiu and Peihao Chen and Jincheng Yang and Xin Yan and Yilun Du and Yining Hong and Chuang Gan},
year = {2024},
eprint = {2403.09631},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2403.09631},
}
Pretrained to generate future video on 38M clips, then fine-tuned to predict frames and actions together; reports 97.7% average success over 100+ real tabletop tasks and roughly double GR-1's success in unseen scenes.先在 3800 万段视频上预训练未来帧生成，再在机器人数据上微调，同时预测视频帧与动作轨迹；报告在 100 多个真机桌面任务上平均成功率 97.7%，在未见场景中成功率约为 GR-1 的两倍。
@misc{cheang2024gr,
title = {{GR-2: A Generative Video-Language-Action Model with Web-Scale Knowledge for Robot Manipulation}},
author = {Chi-Lam Cheang and Guangzeng Chen and Ya Jing and Tao Kong and Hang Li and Yifeng Li and Yuxiao Liu and Hongtao Wu and Jiafeng Xu and Yichu Yang and Hanbo Zhang and Minzhao Zhu},
year = {2024},
eprint = {2410.06158},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2410.06158},
}
Draws the robot's recent motion traces on the input image ('visual trace prompting') and fine-tunes OpenVLA on 150K trajectories; reports ~10% gains on SimplerEnv and 3.5x on real WidowX tasks; 4B variant too.把机器人近期的运动轨迹画在输入图像上（“视觉轨迹提示”），并在 15 万条轨迹上微调 OpenVLA；报告在 SimplerEnv 上提升约 10%、真实 WidowX 任务上提升 3.5 倍，另有基于 Phi-3-Vision 的 4B 精简版。
@misc{zheng2024tracevla,
title = {{TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policies}},
author = {Ruijie Zheng and Yongyuan Liang and Shuaiyi Huang and Jianfeng Gao and Hal Daumé and Andrey Kolobov and Furong Huang and Jianwei Yang},
year = {2024},
eprint = {2412.10345},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.10345},
}
A compact VLA built from a small multimodal model (up to 1.4B) with LoRA and a diffusion policy head, without large-scale robot pretraining; reports higher real-robot success than OpenVLA and ~20x faster inference.由小型多模态模型（至多 1.4B）加 LoRA 微调与扩散策略头构成的紧凑 VLA，无需大规模机器人预训练；报告真机成功率高于 OpenVLA，推理速度约快 20 倍。
@misc{wen2024tinyvla,
title = {{TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulation}},
author = {Junjie Wen and Yichen Zhu and Jinming Li and Minjie Zhu and Kun Wu and Zhiyuan Xu and Ning Liu and Ran Cheng and Chaomin Shen and Yaxin Peng and Feifei Feng and Jian Tang},
year = {2024},
eprint = {2409.12514},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2409.12514},
}
Multi-exit VLA that stops at the smallest model fit for the moment, with budgets for power, latency and GPU memory; on CALVIN it reports 5.2–6.5× lower LLM compute and 2–6× less memory without hurting performance.多出口 VLA：按当前情境停在够用的最小模型，并按功耗、时延和显存预算设定退出条件；在 CALVIN 上报告 LLM 计算量降低 5.2–6.5 倍、显存降低 2–6 倍，性能不降。
@misc{yue2024deer,
title = {{DeeR-VLA: Dynamic Inference of Multimodal Large Language Models for Efficient Robot Execution}},
author = {Yang Yue and Yulin Wang and Bingyi Kang and Yizeng Han and Shenzhi Wang and Shiji Song and Jiashi Feng and Gao Huang},
year = {2024},
eprint = {2411.02359},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2411.02359},
}
Trains VLAs to reason through plans, subtasks, motions and visual features before acting; synthetic reasoning data gives OpenVLA a reported 28% success gain on hard generalization tasks with no extra robot data.训练 VLA 在动作前依次推理计划、子任务、动作和视觉特征；用合成的推理数据，OpenVLA 在困难泛化任务上成功率提升 28%，无需额外机器人数据。
@misc{zawalski2024robotic,
title = {{Robotic Control via Embodied Chain-of-Thought Reasoning}},
author = {Michał Zawalski and William Chen and Karl Pertsch and Oier Mees and Chelsea Finn and Sergey Levine},
year = {2024},
eprint = {2407.08693},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2407.08693},
}
An end-to-end VLA built on Mamba state space models, aiming for both robotic reasoning and cheap fine-tuning and inference.基于 Mamba 状态空间模型的端到端 VLA，兼顾机器人推理能力和较低的微调与推理成本。
@misc{liu2024robomamba,
title = {{RoboMamba: Efficient Vision-Language-Action Model for Robotic Reasoning and Manipulation}},
author = {Jiaming Liu and Mengzhen Liu and Zhenyu Wang and Pengju An and Xiaoqi Li and Kaichen Zhou and Senqiao Yang and Renrui Zhang and Yandong Guo and Shanghang Zhang},
year = {2024},
eprint = {2406.04339},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2406.04339},
}
A VLA with grounded chain of thought and look-ahead spatial reasoning, trained on a dataset built from BridgeV2 with 60,000 auto-annotated trajectories.具备落地链式思考和前瞻空间推理的 VLA，基于 BridgeV2 构建并自动标注的 6 万条轨迹数据集训练。
@misc{sun2024emma,
title = {{Emma-X: An Embodied Multimodal Action Model with Grounded Chain of Thought and Look-ahead Spatial Reasoning}},
author = {Qi Sun and Pengfei Hong and Tej Deep Pala and Vernon Toh and U-Xuan Tan and Deepanway Ghosal and Soujanya Poria},
year = {2024},
eprint = {2412.11974},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.11974},
}
Systematic study of what actually drives VLA manipulation performance, from model design to data; introduces the RoboVLMs family.系统研究哪些因素真正决定 VLA 操作性能（模型设计到数据），并提出 RoboVLMs 系列。
@misc{li2024what,
title = {{What Matters in Building Vision-Language-Action Models for Generalist Robots}},
author = {Xinghang Li and Peiyan Li and Long Qian and Minghuan Liu and Dong Wang and Jirong Liu and Bingyi Kang and Xiao Ma and Xinlong Wang and Di Guo and Tao Kong and Hanbo Zhang and Huaping Liu},
year = {2024},
eprint = {2412.14058},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.14058},
}
RoboUniView learns a unified view representation so a VLM-based policy depends less on camera specs and mounting positions.RoboUniView 学习统一视角表征，降低基于 VLM 的策略对相机参数和安装位置的依赖。
@misc{liu2024robouniview,
title = {{RoboUniView: Visual-Language Model with Unified View Representation for Robotic Manipulation}},
author = {Fanfan Liu and Feng Yan and Liming Zheng and Chengjian Feng and Yiyang Huang and Lin Ma},
year = {2024},
eprint = {2406.18977},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2406.18977},
}
ReKep — a VLM writes manipulation tasks as Python cost functions over 3D keypoints (relational keypoint constraints), solved by a hierarchical optimizer in real time; zero-shot, no task-specific training.ReKep：由视觉语言模型把操作任务写成定义在三维关键点上的 Python 代价函数（关系关键点约束），再由分层优化器实时求解；零样本，无需针对任务训练。
@misc{huang2024rekep,
title = {{ReKep: Spatio-Temporal Reasoning of Relational Keypoint Constraints for Robotic Manipulation}},
author = {Wenlong Huang and Chen Wang and Yunzhu Li and Ruohan Zhang and Li Fei-Fei},
year = {2024},
eprint = {2409.01652},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2409.01652},
}
Combines tactile perception with language so an embodied system can reason about physical properties learned by touch; contributes the PhysiCLeAR dataset.把触觉感知与语言结合，让具身系统对触摸获得的物理属性做常识推理；贡献 PhysiCLeAR 数据集。
@misc{yu2024octopi,
title = {{Octopi: Object Property Reasoning with Large Tactile-Language Models}},
author = {Samson Yu and Kelvin Lin and Anxing Xiao and Jiafei Duan and Harold Soh},
year = {2024},
eprint = {2405.02794},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2405.02794},
}
Generates humanoid whole-body motion from language, starting from human motion priors and refining with LLM commonsense reasoning. Oral at CoRL 2024.从语言描述生成人形全身动作：以人体动作先验初始化，再用 LLM 的常识推理加以修正。CoRL 2024 Oral。
@misc{jiang2024harmon,
title = {{Harmon: Whole-Body Motion Generation of Humanoid Robots from Language Descriptions}},
author = {Zhenyu Jiang and Yuqi Xie and Jinhan Li and Ye Yuan and Yifeng Zhu and Yuke Zhu},
year = {2024},
eprint = {2410.12773},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2410.12773},
}
Extends NaVid into one video VLA model for several navigation tasks (instruction following, object search, question answering, human following).把 NaVid 扩展为单一视频 VLA 模型，统一处理多种导航任务（指令导航、目标搜索、具身问答、跟随人）。
@misc{zhang2024uni,
title = {{Uni-NaVid: A Video-based Vision-Language-Action Model for Unifying Embodied Navigation Tasks}},
author = {Jiazhao Zhang and Kunyu Wang and Shaoan Wang and Minghan Li and Haoran Liu and Songlin Wei and Zhongyuan Wang and Zhizheng Zhang and He Wang},
year = {2024},
eprint = {2412.06224},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.06224},
}
A VLA outputs mid-level navigation commands in language that an RL locomotion policy executes, so legged robots follow natural-language routes; adds Isaac Lab VLN benchmarks.VLA 先用语言输出中层导航指令，再由强化学习运动策略执行，让腿式机器人按自然语言指令导航；并提供 Isaac Lab 上的 VLN 基准。
@misc{cheng2024navila,
title = {{NaVILA: Legged Robot Vision-Language-Action Model for Navigation}},
author = {An-Chieh Cheng and Yandong Ji and Zhaojing Yang and Zaitian Gongye and Xueyan Zou and Jan Kautz and Erdem Bıyık and Hongxu Yin and Sifei Liu and Xiaolong Wang},
year = {2024},
eprint = {2412.04453},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.04453},
}
100 task categories over 2,000+ objects that need common sense, world knowledge and multi-step reasoning, not just instruction following; evaluates VLAs, embodied agents and VLMs in one suite.100 类任务、2000 多个物体，任务需要常识、世界知识与多步推理，而不只是照指令执行；在同一套件中评测 VLA、具身智能体与 VLM。
@misc{zhang2024vlabench,
title = {{VLABench: A Large-Scale Benchmark for Language-Conditioned Robotics Manipulation with Long-Horizon Reasoning Tasks}},
author = {Shiduo Zhang and Zhe Xu and Peiju Liu and Xiaopeng Yu and Yuan Li and Qinghui Gao and Zhaoye Fei and Zhangyue Yin and Zuxuan Wu and Yu-Gang Jiang and Xipeng Qiu},
year = {2024},
eprint = {2412.18194},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.18194},
}
@misc{driess2023palm,
title = {{PaLM-E: An Embodied Multimodal Language Model}},
author = {Danny Driess and Fei Xia and Mehdi S. M. Sajjadi and Corey Lynch and Aakanksha Chowdhery and Brian Ichter and Ayzaan Wahid and Jonathan Tompson and Quan Vuong and Tianhe Yu and Wenlong Huang and Yevgen Chebotar and Pierre Sermanet and Daniel Duckworth and Sergey Levine and Vincent Vanhoucke and Karol Hausman and Marc Toussaint and Klaus Greff and Andy Zeng and Igor Mordatch and Pete Florence},
year = {2023},
eprint = {2303.03378},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2303.03378},
}
Survey of how pretrained foundation models are used in robot perception, decision-making and control, and what holds them back — scarce robot data, safety and uncertainty, real-time execution.综述预训练基础模型如何用于机器人的感知、决策与控制，以及主要瓶颈：机器人数据稀缺、安全性与不确定性量化、实时执行。
@misc{firoozi2023foundation,
title = {{Foundation Models in Robotics: Applications, Challenges, and the Future}},
author = {Roya Firoozi and Johnathan Tucker and Stephen Tian and Anirudha Majumdar and Jiankai Sun and Weiyu Liu and Yuke Zhu and Shuran Song and Ashish Kapoor and Karol Hausman and Brian Ichter and Danny Driess and Jiajun Wu and Cewu Lu and Mac Schwager},
year = {2023},
eprint = {2312.07843},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2312.07843},
}
@misc{xiao2023robot,
title = {{Robot Learning in the Era of Foundation Models: A Survey}},
author = {Xuan Xiao and Jiahang Liu and Zhipeng Wang and Yanmin Zhou and Yong Qi and Qian Cheng and Bin He and Shuo Jiang},
year = {2023},
eprint = {2311.14379},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2311.14379},
}
Survey and meta-analysis arguing robots are limited by task-specific design, and how foundation models could enable general-purpose ones.综述与元分析：指出机器人受限于任务专用设计，并讨论基础模型如何走向通用机器人。
@misc{hu2023toward,
title = {{Toward General-Purpose Robots via Foundation Models: A Survey and Meta-Analysis}},
author = {Yafei Hu and Quanting Xie and Vidhi Jain and Jonathan Francis and Jay Patrikar and Nikhil Keetha and Seungchan Kim and Yaqi Xie and Tianyi Zhang and Hao-Shu Fang and Shibo Zhao and Shayegan Omidshafiei and Dong-Ki Kim and Ali-akbar Agha-mohammadi and Katia Sycara and Matthew Johnson-Roberson and Dhruv Batra and Xiaolong Wang and Sebastian Scherer and Chen Wang and Zsolt Kira and Fei Xia and Yonatan Bisk},
year = {2023},
eprint = {2312.08782},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2312.08782},
}
Gato — one Transformer with one set of weights plays games, captions images and stacks blocks with a real arm.Gato：同一套权重的 Transformer 既能玩游戏、写图像描述，也能用真实机械臂堆积木。
@misc{reed2022generalist,
title = {{A Generalist Agent}},
author = {Scott Reed and Konrad Zolna and Emilio Parisotto and Sergio Gomez Colmenarejo and Alexander Novikov and Gabriel Barth-Maron and Mai Gimenez and Yury Sulsky and Jackie Kay and Jost Tobias Springenberg and Tom Eccles and Jake Bruce and Ali Razavi and Ashley Edwards and Nicolas Heess and Yutian Chen and Raia Hadsell and Oriol Vinyals and Mahyar Bordbar and Nando de Freitas},
year = {2022},
eprint = {2205.06175},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2205.06175},
}
Casts many tabletop tasks as interleaved text-and-image prompts to one Transformer policy, with the VIMA-Bench suite for systematic generalization tests.把多种桌面任务统一成图文交错的提示，交给一个 Transformer 策略执行，并配套 VIMA-Bench 用于系统性泛化评测。
@misc{jiang2022vima,
title = {{VIMA: General Robot Manipulation with Multimodal Prompts}},
author = {Yunfan Jiang and Agrim Gupta and Zichen Zhang and Guanzhi Wang and Yongqiang Dou and Yanjun Chen and Li Fei-Fei and Anima Anandkumar and Yuke Zhu and Linxi Fan},
year = {2022},
eprint = {2210.03094},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2210.03094},
}
PerAct — a Perceiver Transformer over voxelized RGB-D observations that predicts the next keyframe gripper pose as voxel classification; one language-conditioned policy for many RLBench and real tasks.PerAct：在体素化的 RGB-D 观测上运行 Perceiver Transformer，把下一关键帧夹爪位姿当作体素分类来预测；一个语言条件策略即可覆盖众多 RLBench 与真机任务。
Chains an LLM (parses landmarks from the instruction), CLIP (finds them in images) and a visual navigation model to follow route instructions on a real robot without language-annotated robot data.将大语言模型（从指令中解析地标）、CLIP（在图像中找到地标）与视觉导航模型串联，无需带语言标注的机器人数据即可在真实机器人上按路线指令导航。
@misc{shah2022lm,
title = {{LM-Nav: Robotic Navigation with Large Pre-Trained Models of Language, Vision, and Action}},
author = {Dhruv Shah and Blazej Osinski and Brian Ichter and Sergey Levine},
year = {2022},
eprint = {2207.04429},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2207.04429},
}
CLIPort — combines CLIP's semantic "what" stream with Transporter's spatial "where" stream for language-conditioned tabletop manipulation; an early bridge from web vision-language models to robot actions.CLIPort：把 CLIP 提供的语义"是什么"通路与 Transporter 的空间"在哪里"通路结合，实现语言条件的桌面操作；是较早把网络视觉语言模型接入机器人动作的工作。
@misc{shridhar2021cliport,
title = {{CLIPort: What and Where Pathways for Robotic Manipulation}},
author = {Mohit Shridhar and Lucas Manuelli and Dieter Fox},
year = {2021},
eprint = {2109.12098},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2109.12098},
}
Vision-language-action (VLA) models are trained as general-purpose foundation models for robot learning, and require fine-tuning for each downstream task and robot. This raises practical questions such as how much data and training updates are needed, how checkpoints should be selected, and whether simulation can be leveraged to alleviate the cost of real data collection. We study these questions on a task that requires both precise manipulation and spatial understanding of the scene. In this task, an SO-101 arm moves a chess king between two squares that are given only by their names, such …
@misc{kripner2026vla,
title = {{VLA Adaptation with Real and Simulated Data: A Case Study on Robotic Chess}},
author = {Matěj Kripner and Milan Straka},
year = {2026},
eprint = {2611.12676},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.12676},
}
Yuguang Yang, Zhewen Tan, Canyu Chen, Linlin Yang et al.
Abstract摘要
Autoregressive VLA planning usually predicts continuous trajectories as the final output, but driving instructions often describe actions rather than ego states, which weakens direct conditioning on the final continuous trajectory. We address this gap with TrajCond, which introduces a discrete trajectory-geometric condition for VLA planning. Specifically, we represent TrajCond as a 1D-Brohan differential action trajectory and learn it with a two-stage training framework before final continuous trajectory prediction. With the simultaneous autoregressive generation of discrete and numeric …
@misc{yang2026autovla,
title = {{AutoVLA-D: Learn a Trajectory-Geometric Condition from Reasoning for Autonomous Driving VLA}},
author = {Yuguang Yang and Zhewen Tan and Canyu Chen and Linlin Yang and Cheng Chi and Chunyang Liu and Kehua Sheng and Jinyu Yang and others},
year = {2026},
eprint = {2611.12382},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.12382},
}
Physical intelligence requires embodied agents to ground language in observations and execute actions while retaining useful skills. Vision-language-action (VLA) policies offer a shared interface, but task-specific post-training often produces separate specialists. Consolidation is difficult: mixed-task reinforcement learning provides sparse trajectory-level credit, while offline distillation may miss student-visited observations and endpoint distillation does not supervise intermediate action-generation states. We introduce VLA Multi-teacher On-Policy Distillation (VLA-MOPD), a protocol for …
@misc{lv2026vla,
title = {{VLA-MOPD: On-Policy Distillation for Multi-Task VLA Policy Consolidation}},
author = {Yongxuan Lv and Zhen Fang and Lichen Man and Hao Li and Pu Li and Hengyi Wang and Langfeng Cao and Lin Chen and others},
year = {2026},
eprint = {2611.12286},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.12286},
}
Jingxuan Chen, Junshi Luo, Chencheng Dong, Xiaojian Nie et al.
Abstract摘要
Recent Vision-Language-Action (VLA) models have begun incorporating thermal sensing, but existing approaches typically treat thermal images as ordinary visual inputs, providing limited support for temperature-aware relational reasoning in manipulation. To address this limitation, we propose Thermo-VLA with a ThermoRes Encoder, which decomposes thermal images into cold and hot attribute representations and injects them into the RGB token through language-conditioned residual modulation, enabling temperature-driven target selection and continuous manipulation. We further construct six …
@misc{chen2026thermo,
title = {{Thermo-VLA: A Thermal-Aware Vision-Language-Action Model with Gated Residual Modulation for Robotic Manipulation}},
author = {Jingxuan Chen and Junshi Luo and Chencheng Dong and Xiaojian Nie and Tongge Huang and Qiang Huang},
year = {2026},
eprint = {2611.11981},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11981},
}
Generalist robot policies often fail at one step of a task they can otherwise perform. We ask whether such a failure can be repaired by editing the policy's weights directly, without gradients, training data, or a change to the instruction. Told to put the cube then the banana in the bowl, a zero-shot $π_{0.5}$ vision-language-action policy in RoboLab releases the cube before it reaches the bowl. Replaying 21 of these decisions under rewordings shows that it binds "in the bowl" to the wrong object. Closed-form instruction distillation copies the per-layer effect of an explicit rewording into …
Fleets of driving vision-language-action (VLA) models are developed one domain at a time: each operator finetunes a shared base model on private data, producing specialized experts that are costly to consolidate into a single model. Weight merging loses accuracy as heterogeneous domains accumulate, and dense mixture-of-experts routing is prohibitively expensive to combine them. We present DeltaForge, to our knowledge the first post-hoc mixture-of-experts VLA for driving, which consolidates N independently finetuned per-domain models into a single multi-domain model from their weights and a …
@misc{wang2026deltaforge,
title = {{DeltaForge: A Post-hoc Mixture-of-Experts VLA for Multi-Domain Autonomous Driving}},
author = {Bingzhang Wang and Mo Poorsartep and Yasaman Esfandiari},
year = {2026},
eprint = {2611.11856},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2611.11856},
}
Bingzhang Wang, Nathan Zhao, Jialeng Ni, Jiajin Zheng et al.
Abstract摘要
Vision-language-action (VLA) models pair a vision-language backbone with an action head, but in closed-loop control the backbone sits on the control-critical path: while it computes, the robot keeps moving, so each executed action is conditioned on an increasingly stale observation. A larger backbone therefore directly blocks reactivity, especially in humanoid dexterous manipulation. We present DASH, a Decoupled Asynchronous System for Humanoids that removes this coupling. DASH splits the policy into a slow vision-language model (System 2) and a fast flow-matching action head (System 1) that …
@misc{wang2026decoupled,
title = {{Decoupled Asynchronous System for Humanoids (DASH): Latency-Agnostic Scaling of Vision-Language-Action Models for Humanoid Manipulation}},
author = {Bingzhang Wang and Nathan Zhao and Jialeng Ni and Jiajin Zheng and Yin Zhang and Kehua Chen and Jian Yao},
year = {2026},
eprint = {2611.11854},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11854},
}
Jiaxin Guo, Wenzhen Dong, Taoyu Wu, Haomin Kuang et al.
Abstract摘要
Understanding 3D scene dynamics in minimally invasive surgery (MIS) is a fundamental step toward physically intelligent and autonomous surgical systems. However, existing public datasets offer limited scalability and diversity in real-world clinical scenarios. Moreover, annotations typically focus on either tissue or instruments alone, overlooking tool-tissue interactions. This gap limits applications such as augmented reality guidance and autonomous surgery driven by vision-language-action (VLA) models or world models, which demand large-scale, high-quality 3D supervision. In this work, we …
@misc{guo2026surgtap,
title = {{SurgTAP-3D: A Large-Scale Cross-Surgery Video Dataset with Contact-Aware 3D Motion Tracking}},
author = {Jiaxin Guo and Wenzhen Dong and Taoyu Wu and Haomin Kuang and Mengya Xu and Kejian Shi and Ziyi Wang and Tongfan Guan and others},
year = {2026},
eprint = {2611.11697},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.11697},
}
Amit Jadhav, Shaurya Beriwala, David Q. Liu, Yanfei Liu
Abstract摘要
Failure monitors for vision-language-action (VLA) policies are trained offline and then frozen. After deployment, lighting, objects and cameras drift, and the policy itself is updated or moved to new tasks. On recorded real-robot runs, after a switch of task or policy version, a frozen monitor fit on the old task ranks failures little better than chance. We present Streaming Failure Forecasting (SFF), a monitor that keeps learning from the one label every episode provides: whether it succeeded. SFF keeps the offline monitor's network fixed and refits only its last linear layer after each …
@misc{jadhav2026streaming,
title = {{Streaming Failure Forecasting: VLA Failure Monitors That Keep Learning After Deployment}},
author = {Amit Jadhav and Shaurya Beriwala and David Q. Liu and Yanfei Liu},
year = {2026},
eprint = {2611.11688},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11688},
}
Zhang Xi-Jia, Roman Yakunin, Shivang Chopra, Chengyue Huang et al.
Abstract摘要
Large-scale pretraining is reducing the amount of adaptation required to deploy embodied foundation models, but it is not guaranteed that a pretrained model already provide the level of performance required in every deployment. Practitioners therefore need to understand how well a pretrained model matches a target deployment and how much adaptation data is left to be collected. We study whether a small set of successful target-domain demonstrations can reveal whether and how much target-domain data will be required before any adaptation. We introduce FlowEval, a set of training-free metrics …
@misc{xijia2026floweval,
title = {{FlowEval: How Much Adaptation Does an Embodied Foundation Model Need at Deployment?}},
author = {Zhang Xi-Jia and Roman Yakunin and Shivang Chopra and Chengyue Huang and Che Chen and Siddharth Karamcheti and Zsolt Kira},
year = {2026},
eprint = {2611.11650},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11650},
}
Advances in spatial understanding have enabled frontier models to power coding agents that control a real robot arm zero-shot and end-to-end, without task-specific demonstrations or retraining. This brings two new safety challenges. First, the extensive knowledge of frontier models lets these agents perform tasks that earlier VLA and WAM paradigms could not, opening the door to new forms of capability misuse. Second, coding agents can access digital resources such as web pages and emails. Threats in these resources, such as indirect prompt injection, can carry over into the physical world …
@misc{sun2026when,
title = {{When Coding Agents Get Arms: Safety Risks Across the Digital-Physical Boundary}},
author = {Yiming Sun and Mi Zhang},
year = {2026},
eprint = {2611.11438},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11438},
}
Training vision-language-action (VLA) models to follow gameplay instructions across diverse games is limited by the cost of collecting demonstrations. We introduce a framework that automatically generates game environments and expert demonstrations. An LLM-powered agent uses game-building tools to construct executable environments and tasks targeting specified skills. Programmatic experts then use privileged simulator state to generate aligned instructions, rendered observations, and keyboard and mouse actions. We train our game policy, Script2Play, by post-training a 4B vision-language model …
@misc{liu2026from,
title = {{From Game Generation to Game Playing: Fully Automated Ability Training for Vision-Language-Action Models}},
author = {Changran Liu and Qi Wu and Wenchao Ma and Sharon X. Huang and Haomiao Jiang},
year = {2026},
eprint = {2611.11136},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.11136},
}
Zhengyang Cheng, He Zhu, Zeyu Guo, Zihan Liu et al.
Abstract摘要
Vision-language-action (VLA) models enable robots to understand and execute complex instructions, but their autoregressive (AR) action generation introduces significant latency on resource-constrained robotic platforms. Speculative decoding can accelerate AR-based inference by parallelizing action generation and verification; however, existing approaches rely on costly draft models, which are inefficient on edge platforms. In this paper, we propose Zora, a hardware-software co-design framework for accelerating AR-based VLA inference. Specifically, Zora introduces a lightweight …
@misc{cheng2026efficient,
title = {{Efficient VLA Inference via Retrieval-Based Speculation and Hardware Acceleration}},
author = {Zhengyang Cheng and He Zhu and Zeyu Guo and Zihan Liu and Anbang Wu and Yiming Gan and Jingwen Leng and Yu Feng and others},
year = {2026},
eprint = {2611.11098},
archivePrefix = {arXiv},
primaryClass = {cs.AR},
url = {https://arxiv.org/abs/2611.11098},
}
Embodied intelligence systems rely on fast and reliable inference to enable real-time perception, reasoning, and control. Recent advances in large-scale artificial intelligence (AI) models, including Large Language Models (LLMs) and Vision-Language-Action (VLA) models, have significantly enhanced the capabilities of embodied agents in complex environments. However, deploying these models remains challenging due to the limited computational resources and energy budgets of robotic platforms. A fundamental question is where inference should be executed. Collaborative inference, which partitions …
@misc{sun2026where,
title = {{Where Should Models Run? A Collaborative Inference Framework for Embodied Intelligence in 6G}},
author = {Wanlu Sun and Zhenhui Yuan and Xiang Ji and Yi Han and Qing Li and Gan Zheng},
year = {2026},
eprint = {2611.11023},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2611.11023},
}
Pre-training large parametric "Vision-Language-Action" (VLA) models on large demonstration datasets has recently yielded promising results for general-purpose manipulation. However, these models still struggle under even small domain shifts relative to the training data, such as changes in camera viewpoint, background, object identity, or robot morphology. Motivated by the success of Retrieval-Augmented Generation (RAG) in language models, we pursue retrieval to boost VLA performance. Specifically, we seek to leverage retrieval from domain-mismatched, "in-the-wild" demonstration datasets like …
Kevin Park, Izat Rasyad, Dvij Kalaria, Shankar Sastry
Abstract摘要
Training vision-language-action (VLA) policies typically depends on substantial robot demonstration data, making human teleoperation a costly and cumbersome bottleneck, particularly for whole-body humanoid interaction. Reinforcement learning in simulation offers an alternative for scaling this data, but designing task-specific rewards remains challenging. To address this, we propose DreamVLA, which uses generated motion trajectories to guide reinforcement learning (RL) in simulation, converting them into reliable task execution trajectories. These trajectories train VLA models in simulation …
@misc{park2026dreamvla,
title = {{DreamVLA: From Generated Human Motions to Humanoid Vision-Language-Action Policies}},
author = {Kevin Park and Izat Rasyad and Dvij Kalaria and Shankar Sastry},
year = {2026},
eprint = {2611.10872},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10872},
}
Ahmed Zeer, Osama Shbib, Muzaffer Kaan Yuce, Himmet Toprak Kesgin et al.
Abstract摘要
Search allows a robotic policy to evaluate candidate futures before committing to an action. Motivated by test-time planning for vision-language action (VLA) models, we study selective search over action chunks in PushT using a frozen Action Chunking with Transformers (ACT) policy. Beam search generates perturbed alternatives around the policy prediction and scores them through rollouts from cloned simulator states. Across 50 evaluation episodes, beam search raises task success from 0% to 18% and increases average cumulative reward from 39.44 to 103.2. We introduce One Step Ahead (OSA), a …
@misc{zeer2026look,
title = {{Look Ahead: Search-Driven Reasoning in Embodied AI}},
author = {Ahmed Zeer and Osama Shbib and Muzaffer Kaan Yuce and Himmet Toprak Kesgin and Mehmet Fatih Amasyali},
year = {2026},
eprint = {2611.10769},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2611.10769},
}
Jungin Park, Chaoran Zhu, Changjae Oh, Kwanghoon Sohn
Abstract摘要
Understanding three-dimensional physical structure has become essential for vision-language-action models (VLAs). However, existing approaches either rely on explicit geometric inputs at inference time or adopt a single visual pathway, potentially compromising the visual-semantic knowledge inherited from pre-trained models. We present SPARK-VLA, Spatial Representation distillation and Knowledge routing for VLAs, a preserve-specialize-route framework for spatially adapting VLAs without requiring external 3D inputs or experts at inference. SPARK-VLA consists of two key components: (1) spatial …
Human demonstrations provide a scalable source of diverse manipulation data, but leveraging them for robot learning remains challenging due to differences in embodiment, viewpoint, action spaces, and temporal structure. We introduce BATCH, a cross-embodiment learning framework that explicitly aligns human and robot behaviors. BATCH learns self-supervised temporal representations with a cycle-consistency objective to establish cross-embodiment correspondences without manual annotation. We further introduce two complementary training objectives that leverage these correspondences within an …
Robot datasets are costly to collect and are usually tied to a specific embodiment, making it difficult to reuse demonstrations across robots with different geometry, appearance, and kinematics. We introduce UniLatent, a cross-embodiment transfer framework that reduces the embodiment-induced observation gap in a shared latent space. UniLatent renders paired source-target robot observations in simulation and aligns a source visual encoder to a frozen target encoder. A policy trained on encoded source demonstrations can then run directly on target observations, with no pixel-level image …
@misc{ding2026unilatent,
title = {{UniLatent: Cross-Embodiment Transfer via Latent Observation Alignment}},
author = {Yufei Ding and Yi Du and Weiduo Yuan and Hao Jiang and Dylan Goetting and Lawrence Yunliang Chen and Daniel Seita and Yue Wang},
year = {2026},
eprint = {2611.10061},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10061},
}
Pretrained vision language action (VLA) models show strong zero-shot performance, but are rarely deployed directly and are often first post-trained on demonstrations from the target setting. Yet VLAs are typically compared by zero-shot or full-data fine-tuned success, revealing little about the data and compute required for adaptation. We study adaptation efficiency for six VLAs ($\pi_{0}$, $\pi_{0.5}$, and GR00T N1-N1.7) on four LIBERO suites, varying both demonstrations and training steps. This reveals differences hidden by full-data fine-tuning: models within two points on the full …
@misc{chand2026how,
title = {{How Many Demonstrations Does a VLA Need? On the Sample Efficiency of VLA Finetuning}},
author = {Rahul Chand and Miguel Villa Floran},
year = {2026},
eprint = {2611.09964},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09964},
}
Chen Chen, Omid Bagherieh, Naveen Nikamanth Appiah Balaji, Jesse T. Gonzalez et al.
Abstract摘要
Recent work on robot manipulation has focused on specialist policies trained on demonstrations. We show that generalist AI models can control manipulators more reliably than specialist policies such as vision-language-action models (VLAs) and world-action models (WAMs). We systematically analyze agentic manipulation policies (AMPs), in which a generalist AI model commands the manipulator directly, adapting its actions as the task unfolds. We present AMP1, a reference harness that runs unchanged across single-arm, bimanual, and mobile manipulation. AMP1 achieves state-of-the-art results on the …
@misc{chen2026agentic,
title = {{On Agentic Manipulation Policies}},
author = {Chen Chen and Omid Bagherieh and Naveen Nikamanth Appiah Balaji and Jesse T. Gonzalez and Negin Heravi and Fahad Islam and Xueting Li and Jean-Baptiste Regli and others},
year = {2026},
eprint = {2611.09892},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09892},
}
Jingye Zhang, Choo Sin Wai, Dake Zhong, Zhifang Liu et al.
Abstract摘要
Vision-language-action (VLA) policies have become a promising paradigm for end-to-end robot control, but their effectiveness still depends heavily on large volumes of demonstration data. In manipulation tasks, this dependence is compounded by structured spatiotemporal uncertainty in the action space: multiple nearby actions or trajectories may successfully solve the same task, while semantically equivalent demonstrations often unfold at different speeds. Standard pointwise supervision treats each demonstrated action as an exact target, imposing an unnecessary learning burden under such …
@misc{zhang2026sure,
title = {{SURE:Spatiotemporal Uncertainty Representation for Efficient Vision-Language-Action Learning}},
author = {Jingye Zhang and Choo Sin Wai and Dake Zhong and Zhifang Liu and Yuhong Zhang and Siyao Xiao and Guorui Song and Guocun Wang and others},
year = {2026},
eprint = {2611.09684},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09684},
}
Qishen Wang, Pengxiang Ding, Shifeng Bao, Mingyang Sun et al.
Abstract摘要
Vision-language-action (VLA) models typically generate actions in chunks, organizing continuous robot execution into separately predicted segments and thus introducing potential discontinuities at action-chunk transitions. We propose ODEAction, a chunk-free action generation paradigm that replaces action-chunk prediction with Continuous Action Evolution. Given visual observations, language instructions, and robot proprioceptive information, ODEAction learns an action transition field defined over action space, where a transition vector characterizes the evolution rate of the current action …
@misc{wang2026odeaction,
title = {{ODEAction: From Predicting Chunks to Evolving Actions}},
author = {Qishen Wang and Pengxiang Ding and Shifeng Bao and Mingyang Sun and Hengtao Li and Yi Gao and Baohua Yin and Yang Liu and others},
year = {2026},
eprint = {2611.09672},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09672},
}
Memory plays a crucial role in long-horizon task execution and cross-task generalization for Vision-Language-Action (VLA) models operating in complex 3D scenes. However, existing VLA memory mechanisms suffer from memory noise, particularly in complex 3D scenes, as stored episodic memories contain extensive redundant details about scene layouts, object configurations, and robot-specific states, which obscure generalizable manipulation principles and hinder cross-task transfer. To address this challenge, we first introduce a hippocampus-inspired memory consolidation mechanism that synthesizes …
@misc{zhao2026himem,
title = {{HiMEM: Hippocampus-Inspired Memory Consolidation, Pruning, and Composition for Vision-Language-Action Models in 3D Scenes}},
author = {Boran Zhao and Yanbin Hu and Jun Ye and Zhenxian Hu and Yuxin Zhai and Jin Cui and Pengju Ren},
year = {2026},
eprint = {2611.09355},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.09355},
}
Shengjun Deng, Yinghan Zhang, Chengrui Wei, Liu Liang et al.
Abstract摘要
Language-guided embodied person tracking requires a mobile agent to identify a target from a referring description and consistently follow the target under egocentric observations. We present DobotTrack, a target-grounded vision-language-action (VLA) framework that integrates auxiliary grounding supervision with efficient temporal modeling. We systematically investigate five auxiliary tasks---person segmentation, bounding-box grounding, visual-spatial classification (VSC), future DINO feature prediction, and future-frame prediction---and adopt complementary mask and bounding-box supervision …
@misc{deng2026dobottrack,
title = {{DobotTrack: Target-Grounded Vision-Language-Action Learning for Embodied Person Tracking}},
author = {Shengjun Deng and Yinghan Zhang and Chengrui Wei and Liu Liang and Songyuan Liu and Pengkun Zheng and Yajun Zheng and Ying Zhang and others},
year = {2026},
eprint = {2611.09041},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09041},
}
Large vision-language-action (VLA) models generalize well but are difficult to run in real time on a robot, and small VLAs that run there fail more often. Remote inference over beyond-5G and 6G networks lets a robot use a large VLA on a server, but querying the server for every action chunk limits the number of robots per server. We propose VLA handover: the robot runs a small VLA (S-VLA) and switches to a large VLA (L-VLA) on a server only when the S-VLA is likely to fail. We formulate the switching as minimizing the request cost under a required success probability and, since this problem …
Jiachen Tao, Gengyu Zhang, Haoxuan Wang, Bing Liu et al.
Abstract摘要
Vision-Language-Action (VLA) models promise a single language-conditioned policy adaptable to many manipulation behaviors, yet deployment additionally requires skill retention and long-horizon composition: a policy must retain previously acquired skills and reuse them inside composite tasks whose behavior changes across phases. The standard recipe absorbs every new task into one shared update, which provides neither a parameter-level locus for reusable skills nor a mechanism for invoking each skill along the trajectory. We observe that this missing structure is latent rather than absent: …
@misc{tao2026prism,
title = {{PRISM: Primitive Routing via In-context Skill Mixing for Retentive VLA Adaptation}},
author = {Jiachen Tao and Gengyu Zhang and Haoxuan Wang and Bing Liu and Yan Yan},
year = {2026},
eprint = {2611.08917},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08917},
}
Operators can state safety constraints for robots as prohibitions such as "do not put the bowl on the stove". We study whether OpenVLA follows such prohibitions and where it represents the goal it executes. The 10 tasks of LIBEROGoal share one scene, so we can label the goal that each rollout executes. We combine closed-loop rollouts with activation patching, linear probes, activation steering and runtime monitors. On LIBERO-Goal, negation had no detectable effect on which task OpenVLA executed. The model executed the forbidden task in 79% of negated rollouts and completed 78% of plain tasks. …
Sungwon Moon, Minji Kim, Hyemin Boo, Daeun Song et al.
Abstract摘要
Vision-language-action (VLA) models perform versatile manipulation tasks from language instructions, but their inference cost limits how fast a robot can respond in closed-loop control. Visual token pruning reduces this cost; however, prior methods estimate token redundancy from encoded visual embeddings or attention, only after the visual encoder has processed the entire observation. We propose \ours, a training-free token pruning method that moves the pruning decision ahead of the vision encoder by grounding it in a geometric prior derived from the robot's own action. The action chunk …
@misc{moon2026actprune,
title = {{ActPrune: Action-Grounded Visual Token Pruning for Efficient Vision-Language-Action Models}},
author = {Sungwon Moon and Minji Kim and Hyemin Boo and Daeun Song and Jiyoung Lee},
year = {2026},
eprint = {2611.08827},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08827},
}
Luca Macesanu, Tianluo Zhang, Aditya Patil, Archit Sharma et al.
Abstract摘要
Recent advances in robot foundation models have highlighted the potential of in-context learning (ICL) for adapting robot behavior from demonstrations without further fine-tuning. Previous work has used retrieval-based post-training to equip pretrained vision-language-action (VLA) models with the capability for ICL by teaching them to condition action predictions on retrieved visually-relevant demonstration chunks. However, the effectiveness of retrieval hinges on selecting context relevant to the current execution stage, which visual similarity alone may not capture. We introduce VICTR, a …
@misc{macesanu2026victr,
title = {{VICTR: Value-Informed In-Context Learning for Vision-Language-Action Models}},
author = {Luca Macesanu and Tianluo Zhang and Aditya Patil and Archit Sharma and Kevin Angers and Christopher M. Clark and Chen Feng and Juexiao Zhang},
year = {2026},
eprint = {2611.08698},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08698},
}
Meituan Robotics Team, Caicheng Wang, Chen Zhao, Chengpeng Jin et al.
Abstract摘要
We present Meituan-Robotics-0 (MR0), a vision-language-action foundation model and embodied-agent system for stationary-base desktop manipulation. MR0 combines 147 million vision-language samples with 15,060 hours of curated robot trajectories spanning 12 embodiments. A three-stage training recipe aligns multimodal representations with robot behavior through discrete action-token pre-training, transfers these priors to continuous cross-embodiment control through flow-matching mid-training, and adapts the policy through real-world post-training. We introduce Rollback, which pairs replaced …
@misc{team2026meituan,
title = {{Meituan-Robotics-0: A Vision-Language-Action Foundation Model for Desktop Manipulation}},
author = {Meituan Robotics Team and Caicheng Wang and Chen Zhao and Chengpeng Jin and Duo Xu and Fengjiao Chen and Guanglu Wan and Heqing Yang and others},
year = {2026},
eprint = {2611.08577},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08577},
}
Arpit Bahety, Skand Peri, Sriniket Ambatipudi, Chen Xu et al.
Abstract摘要
Vision-Language-Action (VLA) models can perform a wide range of robotic manipulation tasks, but often fail during deployment. Teaching them to recover from these failures requires substantial corrective data: existing approaches learn to recover from failures by directly associating corrective actions with the observations in which they were demonstrated. For instance, DAgger-like fine-tuning can require tens to hundreds of interactions, yet the resulting policy may still fail to generalize the learned corrections beyond the demonstrated scenarios. We hypothesize that semantics can capture …
@misc{bahety2026oopsierecover,
title = {{OopsieRecover: Learning Transferable VLA Corrections from Semantic Interventions}},
author = {Arpit Bahety and Skand Peri and Sriniket Ambatipudi and Chen Xu and Haruki Nishimura and Masha Itkina and Roberto Mart\'in-Mart\'in},
year = {2026},
eprint = {2611.08462},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08462},
}
Michal P. Podolinsky, Yunhao Yang, Daniel Little, Hong Lu et al.
Abstract摘要
Vision-Language-Action models (VLAs) enable general-purpose embodied intelligence, but their performance is brittle under distribution shift, where unanticipated failures can damage equipment or endanger people. Deploying VLAs in real-world systems therefore requires a runtime monitor that is timely, calibrated, and informative while being efficient enough to run on the edge. Existing methods fail to fulfill these requirements. We present VLA-Tug, a lightweight probe over the latent layers of a VLA backbone that improves early failure detection and bounds the missed-detection and false-alarm …
@misc{podolinsky2026failure,
title = {{Failure Prediction and Classification in Vision-Language-Action Models Via Calibrated Latent Probing}},
author = {Michal P. Podolinsky and Yunhao Yang and Daniel Little and Hong Lu and Rohan Siva and Olin Goudey and Neel P. Bhatt and Matthias Scheutz and others},
year = {2026},
eprint = {2611.08425},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08425},
}
Mohammed Hassan (Celine), Zhenyang Chen (Celine), Zheng Wang (Celine), Zhixin Zhu (Celine) et al.
Abstract摘要
Vision-Language-Action (VLA) models enable end-to-end robot control, but inference is increasingly expensive. Each decision processes hundreds of visual tokens, often across multiple cameras, leading to substantial computation while allowing irrelevant background content to distract action prediction under distribution shift. We argue that these issues stem from two complementary forms of redundancy: many visual tokens are spatially irrelevant, while attention over the remaining tokens is often misallocated. Based on this observation, we propose T3R, a Training-Free Two-Stage Token Refinement …
@misc{celine2026t,
title = {{T3R: Training-Free Two-Stage Token Refinement Towards Efficient and Robust VLA Models}},
author = {Mohammed Hassan (Celine) and Zhenyang Chen (Celine) and Zheng Wang (Celine) and Zhixin Zhu (Celine) and Tianlong Chen (Celine) and Yingyan (Celine) and Lin and Chaojian Li},
year = {2026},
eprint = {2611.08413},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08413},
}
Fuhao Li, Wenxuan Song, Qizhou Huang, Ziyang Zhou et al.
Abstract摘要
Reliable robotic manipulation requires both semantic understanding and precise control in three-dimensional space. However, vision-language-action (VLA) foundations largely inherit visual representations from 2D vision-language pre-training, while spatial supervision is often introduced only during task-specific adaptation. Developing spatial understanding as a transferable foundation-model capability therefore calls for learning from broader robot experience. We present SLA, a Spatial-Language-Action Model that scales implicit spatial alignment to continual pre-training and maintains it …
@misc{li2026sla,
title = {{SLA: Scaling Spatial-Language-Action Model via Continual Pre-training}},
author = {Fuhao Li and Wenxuan Song and Qizhou Huang and Ziyang Zhou and Jingbo Wang and Junwu E and Pengwei Wang and Yan Wang and others},
year = {2026},
eprint = {2611.08373},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08373},
}
Nikita Kuzmin, Georgii Demianchuk, Mariya Lezina, Yara Mahmoud et al.
Abstract摘要
Social touch interactions, such as handshakes, fist bumps, and hugs, require humanoid robots to infer human intent, generate contact-rich motions, and maintain safe physical contact. We present AffectiveVLA, an intent-gated Vision-Language-Action (VLA) pipeline deployed on a Unitree G1 humanoid with dexterous hands. A lightweight RGB-based model recognizes interaction intent from two onboard cameras and converts it into a language prompt for the VLA policy. Low-confidence predictions suppress motion execution, reducing unintended activations. During physical contact, interaction-specific …
@misc{kuzmin2026affectivevla,
title = {{AffectiveVLA: Safe and Intent-Aware VLA Control for Human-Humanoid Social Interaction}},
author = {Nikita Kuzmin and Georgii Demianchuk and Mariya Lezina and Yara Mahmoud and Iana Zhura and Miguel Altamirano Cabrera and Dzmitry Tsetserukou},
year = {2026},
eprint = {2611.07712},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07712},
}
Chengcheng Wang, Wei He, Wenhao Li, Hongguang Li et al.
Abstract摘要
Most vision-language-action (VLA) policies decouple manipulation from cognition by bolting an external action mechanism onto a vision-language model (VLM)--either quantizing actions into discrete tokens whose errors bear no stable correspondence to physical displacements, or delegating continuous trajectory generation to a separate action expert whose gradients compete with or are blocked from updating the VLM backbone. We argue that manipulation and cognition should instead be embodied in the same model, shifting from a decoupled VLA to an Embodied Vision-Language Model (eVLM). We introduce …
@misc{wang2026concertina,
title = {{Concertina: Embodying Manipulation and Cognition in One Diffusion Vision-Language Model}},
author = {Chengcheng Wang and Wei He and Wenhao Li and Hongguang Li and Kai Han and Jianyuan Guo and Chang Xu},
year = {2026},
eprint = {2611.07513},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07513},
}
Weichen Fan, Haiwen Diao, Penghao Wu, Siyuan Hu et al.
Abstract摘要
Vision-language-action models are typically built on a simple premise: perception is first compressed by visual encoders, and action reasoning begins only afterwards. Yet manipulation often hinges on precisely the visual signals most vulnerable to this abstraction, e.g., local geometry, boundaries, depth, and subtle state changes. This raises a fundamental question: should action policies inherit a fixed visual endpoint, or directly access evolving visual computation? We introduce NEO-VLA, a structurally encoder-free vision-language-action framework. Images are tokenized by lightweight patch …
@misc{fan2026from,
title = {{From Pixels to Actions - Towards Native Vision-Language-Action Policies}},
author = {Weichen Fan and Haiwen Diao and Penghao Wu and Siyuan Hu and Ziqi Huang and Tianqi Liu and Mingze Sun and Ziyi Wang and others},
year = {2026},
eprint = {2611.06918},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.06918},
}
Michal Korniak, Marco Bagatella, Ralf R\"omer, Michal Nauman et al.
Abstract摘要
The introduction of generalist robot policies has enabled a single, large model to learn to perform a variety of tasks from a vast pre-training corpus. While these models often exhibit powerful zero-shot generalization capabilities, they typically require an adaptation phase to achieve sufficient performance in deployment. Iterated self-improvement via reinforcement learning (RL) has demonstrated promise in enabling this adaptation, yet such approaches have been largely explored in single-task settings - where the goal is to simply improve on a narrow deployment task - which lies in stark …
@misc{korniak2026reinforcement,
title = {{Reinforcement Learning Enables Scaling Generalist Robot Policy Improvement}},
author = {Michal Korniak and Marco Bagatella and Ralf R\"omer and Michal Nauman and Yarden As and Pranav Vignesh and Manan Agarwal and Leander Diaz-Bone and others},
year = {2026},
eprint = {2611.06905},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.06905},
}