Reproduce “Towards Practical and Useful Automated Program Repair for Debugging” 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
Qi Xin, Haojun Wu, Steven P. Reiss, Jifeng Xuan
This paper outlines PracAPR, a vision for an interactive, IDE-integrated automated program repair system designed to assist developers in debugging without relying on comprehensive test suites or frequent program re-execution. It aims to provide rapid, effective repair suggestions for both local and complex multi-location bugs.
Key insights
- Test-free fault localization and re-execution-free patch validation are feasible and effective, demonstrated by the prior ROSE framework, allowing for rapid repair suggestions.
- Effective LLM-based local repair requires augmented prompts with rich dynamic execution and failure information, addressing the 'shallow problem diagnosis' of current methods.
- Multi-location bugs for a single fault can be categorized into 8 'partial patch relationships,' providing a structured basis for designing targeted global repair strategies.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.