Reproduce “Input Reduction Enhanced LLM-based Program Repair” 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
Boyang Yang, Luyao Ren, Xin Yin, Jiadong Ren, Haoye Tian, Shunfu Jin
ReduceFix uses an LLM-generated, task-specific reducer to shrink failure-inducing tests while preserving disagreement between buggy and reference programs, then supplies the reduced test to an unchanged repair model. On LFTBench’s 200 C++ bugs, reduction succeeds in 95.0% of cases. For GLM-4-9B-chat, pass@10 reaches 10.0%, versus 6.5% with original tests; this overall result does not hold for every model or condition.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.