Reproduce “Self-Evolving Agent Harness” 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
Chuyang Wei, Haoxiang Guan, Minghang Zhu, Kexin Zhou, Jiyan He, Xiren Zhou, Huanhuan Chen, Yifei Shen
The authors define a framework for agents to revise persistent software around fixed model weights. It distinguishes updates to instructions, memory, and workflows from changes to executable components, and specifies conditional controls for admitting revisions. In a reported SpreadsheetBench comparison, GPT-5.4-nano with HarnessOpt plus skill scores 77.58%, versus 76.20% for GPT-5.5 with SkillOpt; this is not a matched-cost comparison.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.