Reproduce “JustRUA: A Controlled Study of GPT-6 Robot-Use Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zefan Wang, Yufan Dang, Xuanwei Shi, Mingqi Gao, Cheng Yin, SikYuen Tam, Hanyu Liu, Yasheng Wang and 4 more
The authors evaluate JustRUA, a robot-agent configuration using GPT-6 Astra without a learned policy. It keeps image history, limits motions to 20 steps, adds gripper readings and predicted depth, and uses code to calculate target coordinates. On 111 RoboDojo episodes, its score rises from 29.7 to 50.3; this is not an official leaderboard result, and the final step also removes the decision cap.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.