Our paper "SQL-of-Thought: Multi-agentic Text-to-SQL with Guided Error Correction" has been accepted at the 4th Workshop on Deep Learning for Code (DL4C), co-located with NeurIPS 2025 in San Diego, California!
The Challenge
Converting natural language queries into SQL queries is a hard problem with direct practical implications for making databases more accessible.
Our Approach: SQL-of-Thought
We propose SQL-of-Thought: a multi-agent framework that decomposes the Text2SQL task into:
- Schema linking
- Subproblem identification
- Query plan generation
- SQL generation
- Guided correction loop
Key Innovation
Unlike prior systems that rely only on execution-based static correction, we introduce taxonomy-guided dynamic error modification informed by in-context learning.
Results
SQL-of-Thought achieves state-of-the-art results on the Spider dataset and its variants, combining guided error taxonomy with reasoning-based query planning.
This work was led by Saumya Chaturvedi (MPI-SWS), in collaboration with Aman Chadha (Apple).