Replication Package for "Rethinking Automated Program Repair: The Impact of Bug Complexity, Fault Localization, and LLMs Cost-efficiency "

The package includes: - The ARC-APR benchmark dataset- Buggy and correct program pairs- Fault localization settings- Experimental scripts- Repair results and evaluation scripts- Statistical analysis data The benchmark is constructed from AtCoder Regular Contest (ARC) submissions and is designed to evaluate the effectiveness and generalizability of LLM-based APR techniques under varying bug complexity and fault localization settings. The APR techniques and analysis scripts are in an anonymous GitHub repo.