arXiv cs.AI· Zefan Wang, Yufan Dang, Xuanwei Shi, Mingqi Gao, Cheng Yin, SikYuen Tam, Hanyu Liu, Yasheng Wang, Yaxi Lu, Songyang Zhang, Xin Cong, Yuan Yao·· 1 天前AI 评分55
JustRUA：GPT-6 机器人使用智能体的受控研究
JustRUA: A Controlled Study of GPT-6 Robot-Use Agents
AI 导读
JustRUA 在统一框架下拆解 GPT-6 机器人使用智能体的设计，分别检验模型外围 harness、动作接口、经验复用和场景几何的影响。研究发现保留完整回合历史并提前终止每次动作的精简 harness 效果显著，场景几何帮助最大，即使深度仅由 RGB 预测。该 harness 不含学习策略、自我改进循环或任务特定调优，仍将 RoboDojo 仿真基准得分从 29.7 提升到 50.3。
正文 · 原文
Authors:Zefan Wang, Yufan Dang, Xuanwei Shi, Mingqi Gao, Cheng Yin, SikYuen Tam, Hanyu Liu, Yasheng Wang, Yaxi Lu, Songyang Zhang, Xin Cong, Yuan Yao
Abstract:GPT-6 Astra can drive a robot arm directly through a fixed harness, and scores of such robot-use agents on RoboDojo have risen quickly since. Reported systems differ in harness and test subset and often change several designs at once, so it is hard to tell which choices produce the gains. We take these designs apart and test them in one framework, along the harness around the model, the action interface, the reuse of experience and the scene geometry given to the model. A lean harness that keeps the full episode history and stops each motion early goes a long way. Reusing experience changes how the model acts but is not a free lunch, and scene geometry helps most, even when depth is predicted from RGB alone. JustRUA starts from the official L3 harness and keeps only what matters most. Even with no learned policy, self-improvement loop or task-specific tuning, this simple harness raises the score on the RoboDojo simulation benchmark substantially, from 29.7 to 50.3.
| Comments: | 30 pages, 14 figures. Project page: this https URL. Code: this https URL |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.16009 [cs.RO] |
| (or arXiv:2611.16009v1 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2611.16009 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zefan Wang [view email]
[v1]
Wed, 11 Nov 2026 21:08:31 UTC (7,868 KB)
来源：arXiv cs.AI · arxiv.org