MTSQL-R1: Towards Long-Horizon Multi-Turn Text-to-SQL via Agentic Training

2025-10-12

AI Summary

Existing approaches simplify multi-turn Text-to-SQL into single-turn translation, neglecting execution validation and dialogue state consistency—leading to unexecutable or semantically incoherent SQL queries. This work proposes an agent-based training framework that models multi-turn semantic parsing as a Markov decision process, integrating database execution feedback and persistent dialogue memory to establish a closed-loop reasoning cycle: “generate → execute → verify → revise.” Innovatively combining reinforcement learning with environment-driven iterative optimization, our method significantly outperforms strong baselines on COSQL and SPARC. It is the first to achieve long-horizon, verifiable, and coherent controllable conversational SQL generation. Empirical results validate the critical role of execution-awareness and memory-guided reasoning in multi-turn semantic parsing.

Technology Category

• Multiagent Systems: Multiagent Learning
• Natural Language Processing: Conversational AI/Dialog Systems
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Multi-turn Text-to-SQL aims to translate a user's conversational utterances into executable SQL while preserving dialogue coherence and grounding to the target schema. However, most existing systems only regard this task as a simple text translation task and follow a short-horizon paradigm, generating a query per turn without execution, explicit verification, and refinement, which leads to non-executable or incoherent outputs. We present MTSQL-R1, an agentic training framework for long-horizon multi-turn Text-to-SQL. We cast the task as a Markov Decision Process (MDP) in which an agent interacts with (i) a database for execution feedback and (ii) a persistent dialogue memory for coherence verification, performing an iterative propose to execute -> verify -> refine cycle until all checks pass. Experiments on COSQL and SPARC demonstrate that MTSQL-R1 consistently outperforms strong baselines, highlighting the importance of environment-driven verification and memory-guided refinement for conversational semantic parsing. Full recipes (including code, trained models, logs, reasoning trajectories, etc.) will be released after the internal review to contribute to community research.

Problem

• Addresses multi-turn text-to-SQL translation with dialogue coherence
• Solves non-executable SQL through execution feedback and iterative refinement
• Enhances conversational semantic parsing via environment verification and memory guidance

Innovation

• Agentic training framework for multi-turn Text-to-SQL
• Iterative execute-verify-refine cycle with database feedback
• Persistent dialogue memory for coherence verification