Reproduce “REM: What to Store, When to Recall, What to Forget in Experience Memory for Coding 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

Xinyi Lisa Chen, Pedro Herrero-Vidal, Xiaohan Chen, Xiaojing Yu, Cris Liu, Bingqing Song, Bin Xu, Suleiman A. Khan and 1 more

REM organizes past repository bug-fixing work into four phases and exposes detailed edits only after failed validation. On SWE-ContextBench Lite, its retrieval match@1 reached 68.7%, versus 55.6% for LangMem. With Sonnet 4.5, it resolved 27.6% of tasks versus 24.9% without memory, but the pooled difference was not statistically significant. A model sweep found memory could reduce stronger agents’ resolution.

Summary by alphaXiv.

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