Reproduce “ECoMEM: Explicit Concept Memory for Memory-Dependent Robot Control” 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

Yize Liu, Ke Wang, Mac Schwager, Yiqing Xu, Jiajun Wu

ECoMEM separates an evidence-grounded record of task progress from the policy that chooses actions. Its Writer updates reusable task-relevant facts from evidence, and its Reader supplies those records as memory tokens to the vision-language-action policy. On RoboMME, it reaches 82.42% task-equal mean success, leading on 15 of 16 tasks; real-robot trials are unpaired and unequal, and control skill remains a limit.

Summary by alphaXiv.

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