Reproduce “RAG in a Trenchcoat: When Minimal Memory Is Enough for Agentic Systems, and When It Isn't” 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

Jingyu Liu, Zongze Li, Zach Xu, Zhanhui Zhou, Tahseen Rabbani, Dawn Song, Ce Zhang

The authors evaluate whether memory systems help agents apply remembered rules and patterns. BRIDGE stores notes alongside raw conversation blocks, retrieves with keyword and embedding search, and presents blocks chronologically. Across 20-scenario probes, LANGMEM scores 80% versus BRIDGE’s 35% on stated-rule application, while BRIDGE scores 88% versus LANGMEM’s 55% on procedure induction; these results do not establish general rankings.

Summary by alphaXiv.

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