Reproduce “Sentry: Learning to Recover from LLM Agent Failures at Test Time” 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
Changxiu Ji, Amy Lu, Qizheng Zhang, Kunle Olukotun
Sentry helps long-running tool-using agents recover from failures by keeping verified repair lessons outside persistent context and retrieving up to five lessons when a failure occurs. It adds lessons only after a verifier judges the behavior locally resolved. Across four benchmarks with Qwen3.5-9B, it reports a 37% average relative gain over runtime baselines, not a guarantee of task success.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.