Reproduce “Jetriever: Shifting Agent Memory Retrieval from Similarity to Utility” 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

Ruizhe Li, Licheng Zhang, Mingxuan Du, Zhendong Mao

Jetriever ranks stored exchanges by whether each would help answer a query, using Jev to score every candidate rather than first filtering by textual similarity. On 125 InMind tasks, it achieved 84.8% R@1, versus 1.6% for BM25; top-five retrieval produced a 76.0% application score. Results apply to tested pools, while full-pool scoring scales linearly with memory size.

Summary by alphaXiv.

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