Reimagining Text-to-SQL: A Survey of Large Language Model Architectures and Strategies

Text-to-SQL systems translate natural language questions into executable SQL queries. Large language models like ReFoRCE, Agentar-Scale-SQL, Snowflake's Arctic-Text2SQL-R1, and XiYan-SQL have advanced the field. Modular frameworks decompose the task into sub-problems while agentic systems coordinate specialized agents.