Reproduce “LoopWAM: Closed-Loop Generalist Policy Improvement from Deployment Rollouts” 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

Wenxuan Guo, Yingping Liang, Zhaoguo Wang, Ziyuan Li, Angyuan Ma, Zhenyu Wu, Chengkun Wang, Bingyao Yu and 6 more

LoopWAM trains a robot policy to learn from deployment rollouts by predicting actions and their visual consequences together. It conditions on rollout quality, reducing action imitation for lower-quality behavior while retaining video supervision, including for failures. On RoboChallenge Table30-V2, aggregate success rose from 23.67% to 43.33% over three rounds; these checkpoint comparisons do not isolate the cause.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.