titleSystematically Exploring the Capabilities of GPT-6 Astra as Embodied Policies
authorsGalbot Team, Xuchuan Chen, Xiaoqian Cheng, Yu Deng, Lihe Ding, Shaocong Dong, Xiangjun Gao, Haozhe Jia, Zekai Li, Zhoujian Li, Yunrui Lian, Sikai Liang, Chenghuai Lin, Dairu Liu, Jiahang Liu, Qingtao Liu, Yuxuan Ma, Zekun Qi, Jiayi Su, He Wang, Ruochen Xu, Tianyu Xu, Xudong Xu, Zhe Xu, Mi Yan, Siming Yan, Li Yi, Ruixi Yu, Jinlu Zhang, Yintianrun Zhang, Zhikai Zhang, Zhizheng Zhang, Yixin Zheng, Weiyi Zhu
date2026-09-29
venuearXiv (preprint, not peer reviewed)
idarXiv:2609.38537
affiliationGalbot
upvotes21 (Hugging Face Daily Papers, 2026-11-01)
linksabs pdf hf code
abstractGPT-6 Astra exhibits a remarkable ability to generate numerical robot actions, extending its role beyond high-level planning. To assess Astra's capabilities as general-purpose embodied policies, we conduct comprehensive evaluations across six domains, examining direct control, cooperation with learned policies, and feedback-driven adaptation. In gripper manipulation, Astra can correct task targets and prepare contact conditions for subsequent policy execution; hybrid control with π0.5 achieves 48% success on the evaluated RoboDojo subset. In dexterous manipulation, hybrid control achieves 50% success in ten experience-guided DexJoCo trials, while direct in-hand control struggles to coordinate finger contacts. In mobile manipulation, hybrid control reaches 38.7% success on the evaluated RoboCasa365. In navigation, Astra leads our local comparisons, reaching 92% success on RxR instruction following and 82% on HM3D object search, although search incurs substantial detours. In locomotion, dense motion-reference generation remains unreliable: none of five sequential attempts on a single obstacle course reaches the goal, despite improvements in stability and forward progress. In humanoid loco-manipulation, Astra exceeds baseline methods on 13 of 30 HumanoidBench tasks with pretrained whole-body controllers. These findings reveal a gap between useful task decisions and reliable physical control. Inference latency further constrains practical control: across 50 RoboDojo instances per condition, policy-assisted and direct control consume 624.8 million and 1.132 billion tokens. A 30-second locomotion run requires 250 model calls averaging 39.86 seconds each, with physics paused during inference.

cite

@misc{team2026systematically,
  title={Systematically Exploring the Capabilities of GPT-6 Astra as Embodied Policies},
  author={Galbot Team and Xuchuan Chen and Xiaoqian Cheng and Yu Deng and Lihe Ding and Shaocong Dong and Xiangjun Gao and Haozhe Jia and Zekai Li and Zhoujian Li and Yunrui Lian and Sikai Liang and Chenghuai Lin and Dairu Liu and Jiahang Liu and Qingtao Liu and Yuxuan Ma and Zekun Qi and Jiayi Su and He Wang and Ruochen Xu and Tianyu Xu and Xudong Xu and Zhe Xu and Mi Yan and Siming Yan and Li Yi and Ruixi Yu and Jinlu Zhang and Yintianrun Zhang and Zhikai Zhang and Zhizheng Zhang and Yixin Zheng and Weiyi Zhu},
  year={2026},
  eprint={2609.38537},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2609.38537}
}