Reproduce “SUMONE: A Simple and Efficient Summarizer for Plug-and-Play Agent Memory” 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
Jizhe Zhang, Ang Li, Chenheng Zhang, Yisen Wang
SUMONE trains a separate summarizer using masked reconstruction, where a frozen model attempts to recover hidden history from the summary and remaining context. The downstream agent stays unchanged, and reconstruction is removed at deployment. In the default OpenCode setup, DeepSearchQA F1 is 0.48 versus 0.52 with Full History, while estimated cost falls from $0.170 to $0.038 per query.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.