ReToolSQL: Agentic Reinforcement Learning for Robust Text-to-SQL
Artificial Intelligence
Summary
The authors present ReToolSQL, a new method to improve converting natural language questions into SQL queries by training in two steps. First, they teach the model using examples of good reasoning, and then they let it learn from its own attempts by checking and fixing its SQL based on execution results. This two-stage training helps the model handle difficult questions better and improve accuracy. Their approach achieved top performance on a benchmark without needing extra human-labeled data beyond the test set. Overall, the authors show that combining supervised learning and reinforcement feedback can make a single model more reliable at writing correct SQL from text.
Authors
Pratik Kakkar, Chandra Dhir, Ravi Shankar, Pareekshit Reddy Gaddam, Anup Shirgaonkar
Abstract
Recent work has shown that reinforcement learning from execution feedback can substantially improve text-to-SQL performance, often enabling smaller models to match or exceed much larger systems. However, most existing approaches treat SQL generation as a single-turn task, limiting the model's ability to recover from errors through iterative refinement. We present ReToolSQL, a two-stage training framework for text-to-SQL that combines (i) a supervised warm-start on rejection-sampled reasoning traces with (ii) agentic reinforcement fine-tuning (RFT) over multi-turn tool-use trajectories. The key insight is that the two stages act on complementary axes, the supervised fine-tuning (SFT) on verified privileged-teacher traces expands the set of solvable questions (raising pass@k coverage on the hardest cases), while RFT converts that expanded capability into higher single-pass accuracy by teaching the model when to verify, what evidence to retrieve, and how to repair faulty SQL from execution feedback. Applied to Gemma 4 instruction-tuned (31B), RFT alone achieves 73.66% execution accuracy (EX) on the BIRD-SQL development benchmark (74.12% EX with self-consistency). Initializing RFT from the SFT checkpoint (SFT$\to$RFT) yields our strongest model at 74.32% EX single-pass and 74.77% EX with self-consistency. At the time of writing, this ranked first on the BIRD single-model development-set leaderboard. The approach uses composite rewards anchored on execution correctness, requires no human annotation beyond the benchmark itself, and operates within a single dense 31B model, showing that a properly designed SFT$\to$RFT pipeline over tool-use trajectories is a practical path toward robust enterprise-grade text-to-SQL.