# JustRUA: A Controlled Study of GPT-6 Robot-Use Agents

DevFeed: [JustRUA: A Controlled Study of GPT-6 Robot-Use Agents](<https://devfeed.tech/articles/justrua-a-controlled-study-of-gpt-6-robot-use-agents-132187.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.16009>)

Author: Zefan Wang, Yufan Dang, Xuanwei Shi, Mingqi Gao, Cheng Yin, SikYuen Tam, Hanyu Liu, Yasheng Wang, Yaxi Lu, Songyang Zhang, Xin Cong, Yuan Yao

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [gpt-6-astra](<https://devfeed.tech/topics/gpt-6-astra.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Robot Navigation](<https://devfeed.tech/topics/robot-navigation.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [arm](<https://devfeed.tech/tags/arm.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [experience](<https://devfeed.tech/tags/experience.md>), [framework](<https://devfeed.tech/tags/framework.md>), [full](<https://devfeed.tech/tags/full.md>), [gpt-6](<https://devfeed.tech/tags/gpt-6.md>), [gpt-6-astra](<https://devfeed.tech/tags/gpt-6-astra.md>), [harness](<https://devfeed.tech/tags/harness.md>), [history](<https://devfeed.tech/tags/history.md>), [improvement](<https://devfeed.tech/tags/improvement.md>), [interface](<https://devfeed.tech/tags/interface.md>), [lean](<https://devfeed.tech/tags/lean.md>), [model](<https://devfeed.tech/tags/model.md>), [motion](<https://devfeed.tech/tags/motion.md>), [policy](<https://devfeed.tech/tags/policy.md>), [raises](<https://devfeed.tech/tags/raises.md>), [robot](<https://devfeed.tech/tags/robot.md>), [study](<https://devfeed.tech/tags/study.md>)

## AI overview

The study reports that a lean GPT-6 Astra robot-use harness raises RoboDojo simulation performance from 29.7 to 50.3 without a learned policy, self-improvement loop, or task-specific tuning. In a controlled framework, the authors examine harness design, action interface, experience reuse, and scene geometry. They find that retaining the full episode history and stopping motions early helps substantially; reusing experience changes the model's behavior but is not an unqualified benefit. Scene geometry contributes most, including when depth is predicted from RGB alone. The supplied text is an abstract, so it does not provide the detailed experimental setup or results for individual design choices.

## Source excerpt

arXiv:2611.16009v1 Announce Type: cross 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.