The field of Text-to-SQL is moving towards more advanced and interactive systems, with a focus on handling complex queries, spatial and temporal reasoning, and real-world database exploration. Researchers are exploring the use of multi-agent frameworks, data-centric pipelines, and reinforcement learning to improve the accuracy and robustness of Text-to-SQL models. Noteworthy papers include: From Questions to Queries: An AI-powered Multi-Agent Framework for Spatial Text-to-SQL, which proposes a multi-agent framework for spatial Text-to-SQL tasks. MTIR-SQL: Multi-turn Tool-Integrated Reasoning Reinforcement Learning for Text-to-SQL, which introduces a multi-turn tool-integrated reasoning reinforcement learning framework for Text-to-SQL. Rethinking Text-to-SQL: Dynamic Multi-turn SQL Interaction for Real-world Database Exploration, which introduces a benchmark for evaluating model performance under evolving user interactions.
Text-to-SQL Research Trends
Sources
From Questions to Queries: An AI-powered Multi-Agent Framework for Spatial Text-to-SQL
From Factoid Questions to Data Product Requests: Benchmarking Data Product Discovery over Tables and Text
DCMM-SQL: Automated Data-Centric Pipeline and Multi-Model Collaboration Training for Text-to-SQL Model
Falcon: A Comprehensive Chinese Text-to-SQL Benchmark for Enterprise-Grade Evaluation
MTIR-SQL: Multi-turn Tool-Integrated Reasoning Reinforcement Learning for Text-to-SQL
From Queries to Insights: Agentic LLM Pipelines for Spatio-Temporal Text-to-SQL
Rethinking Text-to-SQL: Dynamic Multi-turn SQL Interaction for Real-world Database Exploration