SWE-Fixer leverages open-source LLMs to effectively resolve GitHub issues, offering an accessible alternative to proprietary models for software engineering tasks.

• Trains specialized open-source LLMs to fix code based on user-reported issues
• Combines file retrieval, code editing, and patch generation capabilities
• Delivers comparable performance to proprietary models while ensuring reproducibility and transparency
• Provides a cost-effective solution for organizations looking to implement AI-assisted software maintenance

This research matters because it democratizes AI tools for software engineering, enabling more organizations to benefit from automated issue resolution without proprietary model limitations.

SWE-Fixer: Training Open-Source LLMs for Effective and Efficient GitHub Issue Resolution