Embodied foundation models are expected to benefit from data scaling like large language models, but face a much tighter data bottleneck. Teleoperated real-robot trajectories remain the dominant pretraining source due to their precise action supervision and embodiment alignment, yet their scalability is limited by high collection cost, acquisition difficulty, and low behavioral and environmental diversity.
Ego4WAM investigates how various properties of egocentric human data—such as human‑robot alignment, data duration, task diversity, and supervision type—affect robot learning. The study, conducted under a unified world‑action model framework, shows that aligned demonstrations improve out‑of‑distribution generalization and lower the amount of target‑task robot data needed. It also finds that video‑only supervision remains effective, and that data duration and task diversity influence downstream capabilities in distinct ways, as validated on real robots and RoboDojo.
By Zhihao Sun, Liu Liu, Xinjiang Wang, Haoyi Jiang, Wei Feng, Huiqiang Zhang, Xiaosong Jia, Zhizhong Su, Zuxuan Wu
arXiv:2610.33685v1 Announce Type: cross
Abstract: Vision-Language-Action (VLA) models are bottlenecked by the scarcity of robot data. Egocentric human manipulation videos offer an abundant, diverse,...
By Qingyuan Cai, Bo Zhang, Yongzhen Huang, Jianyu Chen
Egocentric human data offer a path to scaling robot learning beyond costly robot demonstrations, yet the embodiment gap makes raw human trajectories a poor supervisory target for control. Our key insi...
EgoLAP is a VLA pre‑training framework that learns from both human and robot trajectories by expressing motion intent as temporally abstracted language actions linked to motion‑level reasoning grounded in scene geometry, physics, and object affordances. It jointly trains on a shared language‑based action chain‑of‑thought, enabling the transfer of human experience to robot control. In extensive real‑world and simulated experiments, EgoLAP achieves 80.1% mean real‑world task progress, a 2.3× improvement over alternative action representations, and outperforms composite reasoning formats.
By Lihan Zha, Shresth Grover, Tenny Yin, Samuel M. Bateman, Hengkai Pan, Mengchao Zhang, Aykut Onol, Allen Z. Ren, Dhruv Shah, Anirudha Majumdar
Recent progress in large-scale imitation learning for robot manipulation has been driven by leveraging datasets across a wide range of robot embodiments. However, achieving significant cross-embodiment transfer is often still challenging.
Open‑HumanWAM presents the first fully open recipe for scaling egocentric human video pretraining to 50,000 hours, providing the Open‑HumanWAM‑50K corpus densely annotated with hand motion and natural language. Using this corpus, the authors demonstrate a log‑linear scaling law between pretraining data and robot success, and release HumanWAM‑Vanilla, which achieves high success rates on RoboTwin 2.0 and RoboDojo after post‑training. They further extend the approach with HumanWAM‑Goal, a visual‑goal‑conditioned model that achieves state‑of‑the‑art zero‑shot performance on simulation and physical benchmarks and exhibits embodied in‑context learning.
By Juncheng Ma, Jiaxin Li, Jiankai Tu, Bo Liang, Guodong Lv, Yufan Deng, Junzhe Tang, Liangzhu Yang, Chen Bian, Shukai Gong, Yintianrun Zhang, Zicong Cao, Baohao Zhu, Jianxin Bi, Tengfei Yan, Yinyuan Zhang, Kewei Zhang, Xinglong Chen, Kaiqi Chen, Duomin Wang, Xiaotian Tang, Liang Tan, Li Li, Yuan Xu, Haosen Yang, Yuqi Wang, Jingfei Xia, Songyan Huang, Zhiyang Dou, Zhen Dong, Enze Xie, Tat-Seng Chua, Daquan Zhou
Open‑HumanScale presents the first fully open recipe for scaling egocentric human video pretraining to 50,000 hours, supported by the curated Open‑HumanScale‑50K corpus annotated with hand motion and natural language. Using this corpus, the authors develop HumanWAM‑Vanilla, a world–action model that achieves high success rates on RoboTwin 2.0 and RoboDojo after robot post‑training, and HumanWAM‑Goal, a visual‑goal‑conditioned model that further improves zero‑shot performance and demonstrates embodied in‑context learning on physical robots. The paper also releases the dataset, curation pipelines, pretrained checkpoints, and evaluation tools to enable reproducible scaling research in embodied AI.
By Juncheng Ma, Jiaxin Li, Jiankai Tu, Bo Liang, Guodong Lv, Yufan Deng, Junzhe Tang, Liangzhu Yang, Chen Bian, Zicong Cao, Baohao Zhu, Jianxin Bi, Tengfei Yan, Yinyuan Zhang, Kewei Zhang, Xinglong Chen, Kaiqi Chen, Duomin Wang, Xiaotian Tang, Liang Tan, Li Li, Yuan Xu, Haosen Yang, Yuqi Wang, Jingfei Xia, Songyan Huang, Zhiyang Dou, Zhen Dong, Enze Xie, Tat-Seng Chua, Daquan Zhou
arXiv:2607. 27549v1 Announce Type: cross Abstract: Recent progress in large-scale imitation learning for robot manipulation has been driven by leveraging datasets across a wide range of robot embodiments.
By Ajay Sridhar, Jensen Gao, Jonathan Yang, Jean Mercat, Suneel Belkhale, Dorsa Sadigh
arXiv:2604.19734v2 Announce Type: replace-cross
Abstract: Scaling humanoid foundation models is bottlenecked by the scarcity of robotic data. While massive egocentric human data offers a scalable alt...
By Boyu Chen, Yi Chen, Lu Qiu, Jerry Bai, Yuying Ge, Yixiao Ge
The paper introduces VLAct, a Vision‑Language‑Action model that focuses on representation‑centric continued pre‑training rather than merely scaling robot data. VLAct is trained on diverse, multi‑embodiment robot data and preserves a broad VLM prior while encouraging shared action semantics across embodiments. Experiments across simulation, real‑world, and unseen‑embodiment settings show that VLAct consistently outperforms existing industrial VLA systems, achieving high success rates with only a modest compute budget and open‑source data.
By Senqiao Yang, Chengyao Wang, Yuxin Chen, Zixuan Wang, Longxiang Tang, Haokun Gui, Jinhui Ye, Changsheng Lu, Xiaoyang Wu, Mingkang Zhu, Pengguang Chen, Shu Liu, Zhuotao Tian, Hengshuang Zhao, Bei Yu, Jiaya Jia
EgoActionBench is a new benchmark for dense action captioning in egocentric videos, featuring 1,407 episodes and 41,132 fine‑grained action annotations that specify which hand (left, right, or both) performs each temporally localized atomic action. The benchmark introduces an evaluation protocol that assesses semantic fidelity, temporal overlap, action precision, and recall, thereby measuring what happens, when it happens, and which hand is involved. Twenty baseline models were evaluated, and two backbones (Qwen3‑VL‑4B and LLaVA‑OneVision‑2‑8B) were fine‑tuned and reinforced, achieving significant score improvements and outperforming all open‑source baselines.