AGENTIQL: An Agent-Inspired Multi-Expert Framework for Text-to-SQL Generation
This work provides a robust, scalable, and interpretable approach for semantic parsing tasks, narrowing the gap to GPT-4-based state-of-the-art while using smaller open-source models.
The paper tackles the problem of text-to-SQL generation by addressing limitations in monolithic architectures, such as complex reasoning and schema diversity, with AGENTIQL, an agent-inspired multi-expert framework that improves execution accuracy to 86.07% on the Spider benchmark using 14B models.
LLMs have advanced text-to-SQL generation, yet monolithic architectures struggle with complex reasoning and schema diversity. We propose AGENTIQL, an agent-inspired multi-expert framework that combines a reasoning agent for question decomposition, a coding agent for sub-query generation, and a refinement step for column selection. An adaptive router further balances efficiency and accuracy by selecting between our modular pipeline and a baseline parser. Several steps in the pipeline can be executed in parallel, making the framework scalable to larger workloads. Evaluated on the Spider benchmark, AGENTIQL improves execution accuracy and interpretability and achieves up to 86.07% EX with 14B models using the Planner&Executor merging strategy. The attained performance is contingent upon the efficacy of the routing mechanism, thereby narrowing the gap to GPT-4-based SOTA (89.65% EX) while using much smaller open-source LLMs. Beyond accuracy, AGENTIQL enhances transparency by exposing intermediate reasoning steps, offering a robust, scalable, and interpretable approach to semantic parsing.