JustRUA:GPT-6 機械人操作智能體的受控研究
JustRUA: A Controlled Study of GPT-6 Robot-Use Agents
JustRUA 在統一評測框架下研究 GPT-6 機械人操作智能體,發現保留完整回合歷史並提早停止動作的精簡評測框架可提升 RoboDojo 表現。由官方 L3 評測框架精簡而成的方案,在沒有學習策略、自我改進循環或任務專項微調下,將 RoboDojo 模擬基準分數由 29.7 提升至 50.3。場景幾何資訊帶來最大的額外收益,即使深度只由 RGB 預測亦然。
Published on Nov 11
Authors:
,
,
,
,
,
,
,
,
,
,
Abstract
A simplified evaluation harness that keeps full episode history and stops motions early substantially improves robot-use agent scores on RoboDojo, with scene geometry providing the largest additional benefit.
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.
View arXiv page View PDF Project page GitHub Add to collection
Get this paper in your agent:
hf papers read 2611.16009
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash
Models citing this paper 0
No model linking this paper
Cite arxiv.org/abs/2611.16009 in a model README.md to link it from this page.
Datasets citing this paper 1
ZefanW/robodojo-devset-replays
Viewer • Updated 30 minutes ago • 6.22k • 208
Spaces citing this paper 1
Collections including this paper 0
No Collection including this paper
Add this paper to a collection to link it from this page.
來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co