Text-to-SQL on Spider2-Lite SQLite v1.0 (test)
64.44Pass@1 AccuracyREFORCE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| REFORCEBackbone=GPT-5.4, K (number of candidates)=dynamic2026.05 | 64.44 | — | 58.52 | |
| DIVSKILL-SQLBackbone=GPT-5.4, K (number of candidates)=82026.05 | 62.87 | 84.44 | 71.85 | |
| CHASE-SQLBackbone=GPT-5.4, K (number of candidates)=82026.05 | 61.57 | 83.7 | 66.67 | |
| DIN-SQLBackbone=GPT-5.42026.05 | 37.04 | — | — |