Reproduce “LEDGER: Auditable Agent Memory Under Change” 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
Md Aidul Islam, Md Nasir Uddin Shuvo, Muhammad Waseem, Pekka Abrahamsson
LEDGER addresses multi-session coding-agent memory by storing facts with sources and time intervals. A language model proposes facts and routes questions, while explicit rules determine current values, source precedence, and deletion. On a synthetic coding workload, LEDGER+EP fallback reached 91.1% answer accuracy versus 83.3% for full replay; evidence counts favored fallback, but the workload tests question answering, not downstream coding or real-world source reliability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.