ZAS-SQL: Distilling Rules from Failures for Zero-Shot Text-to-SQL
Researchers have developed new methods to improve the accuracy of Text-to-SQL systems, which translate natural language questions into database queries. TAHOE uses an automated hint optimization system to learn from errors and guide large language models, significantly boosting performance on benchmarks like Spider 2.0-Snow. SOMA-SQL addresses ambiguity by generating synthetic query logs and using execution probing to resolve underspecified questions, outperforming state-of-the-art baselines. ZAS-SQL distills rules from failure cases to improve zero-shot Text-to-SQL performance, establishing a new state-of-the-art that surpasses some few-shot and fine-tuning methods. AI
IMPACT These advancements in Text-to-SQL systems could significantly improve the accessibility and efficiency of database querying for a wider range of users.