Reproduce “Knowledge-Enhanced Program Repair for Data Science Code” 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
Shuyin Ouyang, Jie M. Zhang, Zeyu Sun, Albert Merono Penuela
DSrepair repairs LLM-generated data-science code using API knowledge from a knowledge graph and execution-derived clues that localize failures. It retrieves documentation for seven libraries and analyzes code against supplied tests. On DS-1000, it fixed 104 of 562 GPT-3.5-turbo repair targets (18.51%), versus 72 (12.81%) for Self-Repair. This result is limited to data-science code and the paper’s evaluation setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.