SQLConductor: Search-to-Policy Learning for Step-wise Text-to-SQL Orchestration

📄 arXiv: 2606.23537v1 📥 PDF

作者: Yizhang Zhu, Zhangyang Peng, Boyan Li, Yuyu Luo

分类: cs.DB, cs.AI, cs.LG

发布日期: 2026-06-22


💡 一句话要点

提出SQLConductor以解决复杂数据库环境下的Text-to-SQL问题

🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture)

关键词: Text-to-SQL 策略学习 蒙特卡洛树搜索 数据库查询 自然语言处理 机器学习 强化学习

📋 核心要点

  1. 现有的Text-to-SQL系统在复杂数据库环境中面临协调推理的挑战,固定的多阶段管道限制了适应性。
  2. SQLConductor通过将Text-to-SQL子任务视为专用动作,并利用搜索到策略学习来动态选择动作,从而实现灵活的工作流编排。
  3. 在BIRD-Dev和其他数据集上的实验表明,SQLConductor达到了73.2%的执行准确率,显著优于之前的方法。

📝 摘要(中文)

Text-to-SQL使用户能够通过自然语言访问关系数据库,但在复杂的数据库环境中,协调推理仍然具有挑战性。现有系统通常使用多阶段管道或针对单个阶段的推理模型,固定的管道依赖于预定义的阶段顺序,限制了其适应性。本文提出SQLConductor,一个逐步编排学习框架,通过将Text-to-SQL子任务形式化为工作流组合的专用动作,并训练一个策略模型,根据中间结果和反馈选择下一个动作。SQLConductor引入了搜索到策略学习,利用蒙特卡洛树搜索探索候选工作流,并通过稳定性估计识别稳健的监督。实验表明,SQLConductor在执行准确性和泛化能力上优于现有方法。

🔬 方法详解

问题定义:本文旨在解决在复杂数据库环境中,现有Text-to-SQL方法因固定管道而导致的适应性不足和推理能力有限的问题。

核心思路:SQLConductor的核心思路是将Text-to-SQL的子任务视为专用动作,通过训练策略模型动态选择下一个动作,以适应中间结果和反馈。

技术框架:SQLConductor的整体架构包括三个主要模块:子任务定义、搜索到策略学习和策略模型训练。首先,定义Text-to-SQL的子任务;其次,使用蒙特卡洛树搜索探索候选工作流;最后,通过稳定性加权的监督微调和课程强化学习训练策略模型。

关键创新:SQLConductor的关键创新在于引入了搜索到策略学习,结合蒙特卡洛树搜索和稳定性估计,能够动态适应不同查询需求,显著提升了执行准确性。

关键设计:在训练过程中,采用稳定性加权的监督微调来优先考虑高质量的编排模式,并通过课程强化学习进一步增强策略模型的性能。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在BIRD-Dev数据集上的实验结果显示,SQLConductor达到了73.2%的执行准确率,显著优于以往方法,展示了其在复杂查询需求下的强大适应能力和泛化性能。

🎯 应用场景

SQLConductor的研究成果在多个领域具有潜在应用价值,包括智能问答系统、数据分析工具和自然语言处理应用。通过提高Text-to-SQL的执行准确性和适应性,能够更好地满足用户在复杂数据库环境中的查询需求,推动相关技术的实际应用和发展。

📄 摘要(原文)

Text-to-SQL enables users to access relational databases via natural language, but real-world settings remain challenging due to coordinated reasoning over complex database environments. Existing systems often use multi-stage pipelines or reasoning models specialized for individual stages. However, fixed pipelines rely on predefined stage orders, limiting their adaptivity to query demands and intermediate evidence. Recent orchestration-based methods provide flexibility by composing specialized modules for each query, but typical plan-then-execute approaches still commit to a complete workflow before execution and cannot adapt to intermediate artifacts and feedback. In this paper, we propose SQLConductor, a step-wise orchestration learning framework for Text-to-SQL. SQLConductor formulates Text-to-SQL subtasks as specialized actions for workflow composition and trains a policy model to select the next action based on intermediate artifacts and feedback. To learn this policy, SQLConductor introduces Search-to-Policy Learning, which uses Monte Carlo Tree Search to explore candidate workflows and stability estimation to identify robust supervision. The policy model is trained with Stability-weighted Supervised Fine-tuning to prioritize high-quality orchestration patterns and further enhanced through Curriculum Reinforcement Learning. This transforms offline workflow search into a deployable policy for step-wise orchestration at inference time. Experiments on BIRD-Dev and out-of-distribution datasets show that SQLConductor achieves superior execution accuracy and strong generalization, reaching 73.2% EX on BIRD-Dev with a compact orchestration policy coordinating frozen larger action models, outperforming prior methods that directly train comparable or larger Text-to-SQL backbones. Further analyses show that the learned policy adapts orchestration to diverse query demands.