Rose-SQL: Role-State Evolution Guided Structured Reasoning for Multi-Turn Text-to-SQL

2026-05-05

AI Summary

This work addresses the challenges of modeling context dependence in multi-turn Text-to-SQL tasks, where existing approaches suffer from unstable API-based inference or high fine-tuning costs. The authors propose Rose-SQL, a training-free framework that introduces Role-State as a structured blueprint to capture contextual semantics and employs structural isomorphism checking to track its evolution across dialogue history, enabling structure-aware contextual reasoning. Leveraging in-context learning with large reasoning models, Rose-SQL significantly outperforms current fine-tuned models on the SParC and CoSQL benchmarks: Qwen3-4B surpasses prior in-context learning baselines, while Qwen3-8B and Qwen3-14B achieve state-of-the-art performance.

Abstract

Recent advances in Large Reasoning Models (LRMs) trained with Long Chain-of-Thought have demonstrated remarkable capabilities in code generation and mathematical reasoning. However, their potential in multi-turn Text-to-SQL tasks remains largely underexplored. Existing approaches typically rely on unstable API-based inference or require expensive fine-tuning on small-scale models. In this work, we present Rose-SQL, a training-free framework that leverages small-scale LRMs through in-context learning to enable accurate context-dependent parsing. We introduce the Role-State, a fine-grained representation that bridges the structural gap between schema linking and SQL generation by serving as a structural blueprint. To handle conversational dependencies, Rose-SQL traces the evolution of Role-State through historical context via structural isomorphism checks, guiding the model to infer the possible SQL composition for the current question through verified interaction trajectories. Experiments on the SParC and CoSQL benchmarks show that, within the Qwen3 series, Rose-SQL outperforms in-context learning baselines at the 4B scale and substantially surpasses state-of-the-art fine-tuned models at the 8B and 14B scales, while showing consistent gains on additional reasoning backbones.

Problem

• Multi-Turn Text-to-SQL
• Context-Dependent Parsing
• Conversational Dependencies
• Schema Linking
• SQL Generation

Innovation

• Role-State
• structured reasoning
• multi-turn Text-to-SQL
• in-context learning
• structural isomorphism