SQLConductor: Search-to-Policy Learning for Step-wise Text-to-SQL Orchestration
2026-06-22 • Databases
DatabasesArtificial IntelligenceMachine Learning
AI summaryⓘ
The authors present SQLConductor, a new method that helps computers translate natural language questions into SQL database queries more flexibly and accurately. Unlike previous systems that follow a fixed series of steps, SQLConductor chooses actions one at a time based on results seen so far, allowing it to adapt as needed. They teach the system using a search process combined with training techniques that focus on stable and good workflows. Their method performs better on benchmark tests and generalizes well to new situations, showing it can handle complex database questions more effectively.
Text-to-SQLRelational databasesOrchestration learningMonte Carlo Tree SearchPolicy modelCurriculum Reinforcement LearningExecution accuracyWorkflow compositionStability-weighted training
Authors
Yizhang Zhu, Zhangyang Peng, Boyan Li, Yuyu Luo
Abstract
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.