QDA-SQL: Questions Enhanced Dialogue Augmentation for Multi-Turn Text-to-SQL

arXiv cs.AI · 6 hours ago

Researchers introduced QDA-SQL, a data augmentation method that uses LLMs to generate multi-turn question-answer pairs for improving fine-tuned models on Text-to-SQL tasks. The method incorporates validation and correction mechanisms to handle ambiguous and unanswerable questions. Fine-tuned models using QDA-SQL showed improved SQL statement accuracy and better performance on complex multi-turn Text-to-SQL tasks.