Reproduce “Just-in-Time Memory: Learning to Curate Task-Adaptive Memory for LLM 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

Yefan Zhou, Yang Li, Zeyu Leo Liu, Semih Yavuz, Shafiq Joty

JITMEM stores raw successful trajectories and, when a new task arrives, retrieves relevant traces and uses a curator to produce a temporary task-specific briefing for a frozen executor. On ALFWorld, Qwen3-8B JITMEM reached 77.4% success versus 61.2% for SkillOS. It also outperformed reported baselines on WebShop and τ²-bench, though τ²-bench gains were domain-dependent and training-free.

Summary by alphaXiv.

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