MTSQL-R1: Towards Long-Horizon Multi-Turn Text-to-SQL via Agentic Training
This work addresses the challenge of generating accurate and coherent SQL queries from conversational text for database users, representing an incremental advance by integrating agentic training into a known bottleneck.
The paper tackles the problem of multi-turn Text-to-SQL by addressing issues like non-executable or incoherent outputs in existing systems, and introduces MTSQL-R1, an agentic training framework that uses execution feedback and dialogue memory for iterative refinement, achieving consistent performance improvements on COSQL and SPARC benchmarks.
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.