A Preview of XiYan-SQL: A Multi-Generator Ensemble Framework for Text-to-SQL

TL;DR

XiYan-SQL is a multi-generator ensemble framework that combines semi-structured schema representation, in-context learning, and fine-tuning to improve natural language to SQL translation, achieving state-of-the-art accuracy across multiple datasets.

Contribution

The paper introduces XiYan-SQL, a novel ensemble framework that integrates multiple generation strategies and a new schema representation to enhance SQL query generation from natural language.

Findings

Achieves 75.63% accuracy on Bird benchmark.

Outperforms previous methods on multiple datasets.

Effectively balances diversity and quality in SQL candidate generation.

Abstract

To tackle the challenges of large language model performance in natural language to SQL tasks, we introduce XiYan-SQL, an innovative framework that employs a multi-generator ensemble strategy to improve candidate generation. We introduce M-Schema, a semi-structured schema representation method designed to enhance the understanding of database structures. To enhance the quality and diversity of generated candidate SQL queries, XiYan-SQL integrates the significant potential of in-context learning (ICL) with the precise control of supervised fine-tuning. On one hand, we propose a series of training strategies to fine-tune models to generate high-quality candidates with diverse preferences. On the other hand, we implement the ICL approach with an example selection method based on named entity recognition to prevent overemphasis on entities. The refiner optimizes each candidate by correcting logical or syntactical errors. To address the challenge of identifying the best candidate, we fine-tune a selection model to distinguish nuances of candidate SQL queries. The experimental results on multiple dialect datasets demonstrate the robustness of XiYan-SQL in addressing challenges across different scenarios. Overall, our proposed XiYan-SQL achieves the state-of-the-art execution accuracy of 75.63% on Bird benchmark, 89.65% on the Spider test set, 69.86% on SQL-Eval, 41.20% on NL2GQL. The proposed framework not only enhances the quality and diversity of SQL queries but also outperforms previous methods.