# Text-to-SQL

FLINT, a Text-to-SQL system for production financial databases, grounds natural-language questions in question-specific table constraints instead of relying on column-name similarity. General systems do well on benchmarks like Spider and BIRD, where schemas are shallow and column values are readable, but financial databases store concepts as opaque integer keys, so even simple queries need multiple joins and those methods fall below 50%. FLINT resolves concepts with a lookup agent, retrieves structurally similar query templates, and prunes large schemas along foreign-key chains; on two datasets totaling 359 questions over production financial schemas it outperforms baselines using the same LLM, and it is deployed in a financial data retrieval service.

FLINT, a Text-to-SQL system for production financial databases, grounds natural-language questions in question-specific table constraints instead of relying on column-name similarity. General systems do well on benchmarks like Spider and BIRD, where schemas are shallow and column values are readable, but financial databases store concepts as opaque integer keys, so even simple queries need multiple joins and those methods fall below 50%. FLINT resolves concepts with a lookup agent, retrieves structurally similar query templates, and prunes large schemas along foreign-key chains; on two datasets totaling 359 questions over production financial schemas it outperforms baselines using the same LLM, and it is deployed in a financial data retrieval service.

The work argues that in agentic Text-to-SQL a correct execution result does not necessarily mean the underlying underspecification has been adequately resolved, since the agent may silently make unverified assumptions that happen to match the intended answer; it attributes this partly to premature clarification termination and introduces PlanPool, which externalizes the clarification plan as a mutable question pool where every planned question must be explicitly asked or dropped before submission and newly discovered ambiguities can be added during interaction, consistently improving ambiguity coverage and reducing silent failures across three benchmarks derived from BIRD-Interact and Spider while maintaining competitive execution accuracy.

The work argues that in agentic Text-to-SQL a correct execution result does not necessarily mean the underlying underspecification has been adequately resolved, since the agent may silently make unverified assumptions that happen to match the intended answer; it attributes this partly to premature clarification termination and introduces PlanPool, which externalizes the clarification plan as a mutable question pool where every planned question must be explicitly asked or dropped before submission and newly discovered ambiguities can be added during interaction, consistently improving ambiguity coverage and reducing silent failures across three benchmarks derived from BIRD-Interact and Spider while maintaining competitive execution accuracy.