Self-Bootstrapping Automated Program Repair: Using LLMs to Generate and Evaluate Synthetic Training Data for Bug Repair

1. 1Generate synthetic training data using LLMs to supplement limited real-world data
2. 2Evaluate the generated data to ensure its quality and diversity
3. 3Use the synthetic data to train APR models, improving their ability to repair bugs across multiple programming languages
4. 4Fine-tune the APR models using the synthetic data to adapt to new bug types and programming languages

Software engineers and AI researchers on a team can benefit from this approach as it enhances automated program repair capabilities, improving the overall quality and efficiency of the software development process

💡 LLMs can be used to generate high-quality synthetic training data, enhancing the capabilities of automated program repair systems

💡 LLMs can generate synthetic training data for automated program repair!