Text-to-SQL Execution on SQL Execution Benchmarks Spider, BIRD
98.36Execution Accuracy (Spider)DSIR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DSIRMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 98.36 | 92.05 | 95.2 | |
| MIRA-GroupMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 98.26 | 89.9 | 94.08 | |
| Raw MixtureMid-training Protocol=Full Corpus, Token Budget=50B2026.05 | 98.07 | 90.29 | 94.18 | |
| DataManMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 97.58 | 90.09 | 93.84 | |
| MIRA-SourceMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 97.49 | 91.26 | 94.38 | |
| MIRA-GlobalMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 97 | 91.53 | 94.26 | |
| RandomMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 96.52 | 91.07 | 93.79 | |
| PPLMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 90.99 | 90.33 | 90.66 | |
| +SFTMid-training Protocol=No Mid-training, Token Budget=02026.05 | 79.98 | 48.5 | 64.24 | |
| Base ModelMid-training Protocol=No Mid-training, Token Budget=02026.05 | 11.22 | 1.37 | 6.29 |