By taking the union of all shortest paths between LLM-identified source and destination tables, SchemaGraphSQL reaches 95.71% table recall and 62.91% execution accuracy on BIRD dev, but the SOTA framing depends on recall-weighted metrics and a test-set-chosen configuration.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SchemaGraphSQL: Efficient Schema Linking with Pathfinding Graph Algorithms for Text-to-SQL on Large-Scale Databases
By taking the union of all shortest paths between LLM-identified source and destination tables, SchemaGraphSQL reaches 95.71% table recall and 62.91% execution accuracy on BIRD dev, but the SOTA framing depends on recall-weighted metrics and a test-set-chosen configuration.