Reproduce “Code Transformation Rule Synthesis using LLMs: Potential and Limits” 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
Axel Allain, Aymeric Blot, Djamel Eddine Khelladi, Mathieu Acher
The authors test whether LLMs can turn code diffs into reusable rewrite rules that match code patterns and reuse captured fragments in replacements. They assess rule validity and output matches across tasks and DSLs. On a 43-example Python API-misuse task, GPT-5.4’s Comby rules with retrieval had 95.3% applicability and 63.4% AST match, versus anti-unification’s 97.6% and 0%. This comparison shows a task-specific trade-off, not semantic guarantees.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.