Reproduce “GrACE: Generation using Associated Code Edits” 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
Priyanshu Gupta, Avishree Khare, Yasharth Bajpai, Saikat Chakraborty, Sumit Gulwani, Aditya Kanade, Arjun Radhakrishna, Gustavo Soares and 1 more
GrACE conditions code generation on earlier edits associated with a target code change, representing the target and context edits for Codex or CodeT5 to predict a replacement. In the SerializationException example, the earlier exception change helps identify the needed import. On filtered C3PO, Codex-Davinci top-1 Exact Match rises from 37.09% to 67.92%; this benchmark excludes some difficult cases.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.