SQL-Trail: Multi-Turn Reinforcement Learning with Interleaved Feedback for Text-to-SQL
It addresses the problem of robust Text-to-SQL generation for database querying, offering a novel agentic workflow that is incremental over single-pass methods.
The paper tackles the performance gap between AI systems and human experts in Text-to-SQL generation by introducing SQL-Trail, a multi-turn reinforcement learning framework that iteratively refines queries using execution feedback, achieving up to 18x higher data efficiency and outperforming larger proprietary systems by 5% on average.
While large language models (LLMs) have substantially improved Text-to-SQL generation, a pronounced gap remains between AI systems and human experts on challenging benchmarks such as BIRD-SQL. We argue this gap stems largely from the prevailing single-pass paradigm, which lacks the iterative reasoning, schema exploration, and error-correction behaviors that humans naturally employ. To address this limitation, we introduce SQL-Trail, a multi-turn reinforcement learning (RL) agentic framework for Text-to-SQL. Rather than producing a query in one shot, SQL-Trail interacts with the database environment and uses execution feedback to iteratively refine its predictions. Our approach centers on two key ideas: (i) an adaptive turn-budget allocation mechanism that scales the agent's interaction depth to match question difficulty, and (ii) a composite reward panel that jointly incentivizes SQL correctness and efficient exploration. Across benchmarks, SQL-Trail sets a new state of the art and delivers strong data efficiency--up to 18x higher than prior single-pass RL state-of-the-art methods. Notably, our 7B and 14B models outperform substantially larger proprietary systems by 5% on average, underscoring the effectiveness of interactive, agentic workflows for robust Text-to-SQL generation.