Reproduce “Better Automatic Program Repair by Using Bug Reports and Tests Together” 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
Manish Motwani, Yuriy Brun
The study develops Blues, an unsupervised information retrieval-based fault localization technique utilizing bug reports, and RAFL, a novel unsupervised method for combining arbitrary fault localization techniques. Integrating SBIR (Blues combined with Spectrum-Based Fault Localization via RAFL) into automated program repair tools enabled the correct repair of 7 defects in Defects4J previously unaddressed by 14 other repair tools, while also improving overall patch quality for FL-sensitive tools.
Key insights
- Combining fault localization information from natural-language bug reports and test execution data leads to superior localization accuracy compared to using either source independently.
- Unsupervised rank aggregation methods for fault localization can achieve better performance than supervised learning-based approaches, removing the dependency on extensive training data.
- Enhanced fault localization directly improves the quality of patches generated by automated program repair tools, leading to more correctly fixed defects.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.