$R^3$: "This is My SQL, Are You With Me?" A Consensus-Based Multi-Agent System for Text-to-SQL Tasks

Hanchen Xia; Feng Jiang; Naihao Deng; Cunxiang Wang; Guojiang Zhao; Rada Mihalcea; Yue Zhang

$R^3$: "This is My SQL, Are You With Me?" A Consensus-Based Multi-Agent System for Text-to-SQL Tasks

Computation and Language 2024-07-12 v2 Artificial Intelligence Databases

Authors: Hanchen Xia , Feng Jiang , Naihao Deng , Cunxiang Wang , Guojiang Zhao , Rada Mihalcea , Yue Zhang

Abstract

Large Language Models (LLMs) have demonstrated strong performance on various tasks. To unleash their power on the Text-to-SQL task, we propose $R^3$ (Review-Rebuttal-Revision), a consensus-based multi-agent system for Text-to-SQL tasks. $R^3$ outperforms the existing single LLM Text-to-SQL systems as well as the multi-agent Text-to-SQL systems by $1.3\%$ to $8.1\%$ on Spider and Bird. Surprisingly, we find that for Llama-3-8B, $R^3$ outperforms chain-of-thought prompting by over 20\%, even outperforming GPT-3.5 on the development set of Spider.

Keywords

text-to-sql large language model multi-agent systems

Cite

@article{arxiv.2402.14851,
  title  = {$R^3$: "This is My SQL, Are You With Me?" A Consensus-Based Multi-Agent System for Text-to-SQL Tasks},
  author = {Hanchen Xia and Feng Jiang and Naihao Deng and Cunxiang Wang and Guojiang Zhao and Rada Mihalcea and Yue Zhang},
  journal= {arXiv preprint arXiv:2402.14851},
  year   = {2024}
}

Comments

12 pages, 2 figures, 8 tables

$R^3$: "This is My SQL, Are You With Me?" A Consensus-Based Multi-Agent System for Text-to-SQL Tasks

Abstract

Keywords

Cite

Comments

Related papers