Reproduce “DISCERN: Preserve at Write, Resolve at Read for Persistent LLM-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
Chirag Ahuja, Vladislav Tanov, Ramcharan Kakarla, Jonathan Taws
DISCERN preserves enterprise-agent memories in an append-only event log and resolves entities and changing facts at query time, rather than collapsing them during storage. On synthetic ENT-310, it reaches 0.861 macro answer accuracy versus 0.806 for the strongest baseline, and 0.800 entity-resolution accuracy, where baselines reach at most 0.400. Entity resolution has only 15 questions, limiting precision.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.