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

TL;DR

Rose-SQL is a training-free, in-context learning framework that uses Role-State evolution to improve multi-turn Text-to-SQL performance with small-scale large reasoning models, surpassing fine-tuned models.

Contribution

It introduces Role-State as a structural blueprint and a method to trace its evolution, enabling context-dependent SQL parsing without fine-tuning.

Findings

Outperforms in-context learning baselines at 4B scale.

Surpasses state-of-the-art fine-tuned models at 8B and 14B scales.

Shows consistent gains across different reasoning backbones.

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.