TRUST-SQL: Tool-Integrated Multi-Turn Reinforcement Learning for Text-to-SQL over Unknown Schemas

TL;DR

TRUST-SQL introduces a tool-integrated, multi-turn reinforcement learning approach for Text-to-SQL tasks in unknown schemas, effectively identifying relevant database subsets without pre-loaded metadata, and significantly improving performance over existing methods.

Contribution

It proposes a novel framework combining structured multi-phase reasoning and a dual-track strategy to handle unknown schemas in Text-to-SQL, outperforming baseline models without schema preloading.

Findings

Achieves 30.6% and 16.6% absolute improvements on two model variants.

Outperforms baselines relying on schema prefilling.

Demonstrates effectiveness across five benchmarks.

Abstract

Text-to-SQL parsing has achieved remarkable progress under the Full Schema Assumption. However, this premise fails in real-world enterprise environments where databases contain hundreds of tables with massive noisy metadata. Rather than injecting the full schema upfront, an agent must actively identify and verify only the relevant subset, giving rise to the Unknown Schema scenario we study in this work. To address this, we propose TRUST-SQL (Truthful Reasoning with Unknown Schema via Tools). We formulate the task as a Partially Observable Markov Decision Process where our autonomous agent employs a structured four-phase protocol to ground reasoning in verified metadata. Crucially, this protocol provides a structural boundary for our novel Dual-Track GRPO strategy. By applying token-level masked advantages, this strategy isolates exploration rewards from execution outcomes to resolve credit assignment, yielding a 9.9% relative improvement over standard GRPO. Extensive experiments across five benchmarks demonstrate that TRUST-SQL achieves an average absolute improvement of 30.6% and 16.6% for the 4B and 8B variants respectively over their base models. Remarkably, despite operating entirely without pre-loaded metadata, our framework consistently matches or surpasses strong baselines that rely on schema prefilling.