Search papers, labs, and topics across Lattice.
The paper introduces EviLink, a novel schema linking approach for Text-to-SQL that reframes the task as uncertainty-aware schema-need inference across multiple plausible SQL paths. EviLink distinguishes between required and path-dependent uncertain schema items, strategically acquiring evidence only when necessary to resolve ambiguities. Experiments on BIRD-Dev and Spider2-Snow demonstrate that EviLink improves schema completeness and relevance while reducing token costs, achieving a 90.15% field-level strict recall rate on Spider2-Snow.
Text-to-SQL models can now achieve higher accuracy with fewer tokens by reasoning about multiple possible query paths and selectively gathering evidence only when uncertain about which schema elements are needed.
Schema linking is a difficult and important step in large-scale Text-to-SQL, where systems must identify a compact yet sufficient schema context from large and ambiguous databases. Existing methods often treat schema linking as deterministic selection around a single SQL path, but complex questions may admit multiple valid realizations with different schema needs. We reframe schema linking as uncertainty-aware schema-need inference over multiple plausible SQL paths, where the system distinguishes required schema items from path-dependent uncertain ones and acquires evidence only where needed. We instantiate this reframing with EviLink, which combines multi-hypothesis schema grounding with uncertainty-guided evidence acquisition. Experiments on BIRD-Dev and Spider2-Snow show that this perspective improves the balance among schema completeness, schema relevance, and token cost. On Spider2-Snow, EviLink achieves 90.15% field-level strict recall rate, uses 123.30K average tokens, and improves downstream SQL generation under a fixed generator.
