Reproduce “Harvesting SWE Benchmarks from Wild Vibe Coding Sessions” 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
Ao Lu, Yuanhao Zeng, Shaocheng Lan, Sijie Xu, Jiacheng Bao, Kuang Yan, Heng Lin, Kan Ren
SessionFarmer turns repository coding histories into autonomous benchmark tasks by grouping sessions and code changes around engineering goals, then packaging the full episode as a task with tests and a reference patch. In CropChat, full-episode statements covered 75% of requirements in sampled multi-session tasks, while the strongest tested model–harness setup reached 39.3% Pass@1, a result limited to selected benchmark tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.