Reproduce “Learning Project-wise Subsequent Code Edits via Interleaving Neural-based Induction and Tool-based Deduction” 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
Chenyan Liu, Yun Lin, Yuhuan Huang, Jiaxin Chang, Binhang Qi, Bo Jiang, Zhiyong Huang, Jin Song Dong
TRACE combines IDE services and neural models to recommend related code edits. A learned invoker routes predictable syntactic changes to tools such as rename or reference lookup; neural models locate and generate context-dependent edits. In a simulation of 500 held-out commits, TRACE's top-1 localization match rate was 35.18%, versus 11.47% for CoEdPilot; ground-truth edits advanced later steps, not autonomous editing.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.