On the Role of Fault Localization Context for LLM-Based Program Repair

1. 1Conduct a large-scale empirical study to evaluate the impact of fault localization context on LLM-based program repair
2. 2Use a dataset of 500 SWE-bench Verified instances to train and test the model
3. 3Evaluate 61 configurations of fault localization context to determine the optimal amount and type of context needed
4. 4Analyze the results to identify the benefits and limitations of using additional context beyond the predicted buggy location

Software engineers and AI researchers on a team can benefit from understanding how fault localization context affects LLM-based program repair, as it can improve the efficiency and accuracy of their repair tools

💡 The amount and type of fault localization context used can significantly impact the effectiveness of LLM-based program repair

💡 LLM-based program repair: how much fault localization context is needed? 🤔