SWE-SQL: Illuminating LLM Pathways to Solve User SQL Issues in Real-World Applications
Jinyang Li, Xiaolong Li, Ge Qu, Per Jacobsson, Bowen Qin, Binyuan Hui, Shuzheng Si, Nan Huo, Xiaohan Xu, Yue Zhang, Ziwei Tang, Yuanshuai Li, Florensia Widjaja, Xintong Zhu, Feige Zhou, Yongfeng Huang, Yannis Papakonstantinou, Fatma Ozcan, Chenhao Ma, Reynold Cheng
2025-06-25
Summary
This paper talks about SWE-SQL, a new system and benchmark designed to help large language models debug SQL queries more effectively in real-world situations.
What's the problem?
The problem is that while AI models are good at turning text into SQL code, they struggle with fixing SQL queries when there are errors, which is a big challenge for database users and developers.
What's the solution?
The researchers created a new large set of real SQL problems and built an environment called Six-Gym where models are trained to find and fix these SQL issues. They also introduced an open-source AI agent called Bird-Fixer that learns to debug SQL queries better than existing systems.
Why it matters?
This matters because improving SQL debugging with AI helps users manage databases more easily and accurately, reduces errors, and supports privacy by using open-source models instead of relying only on proprietary software.
Abstract
A new benchmark and training environment for debugging SQL issues using advanced open-source models significantly improves their performance compared to proprietary solutions.