Mathematical Dialogue Evaluation on MathDial (test)
21AccuracyDPO
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DPOMode=Standard Mode2025.06 | 21 | 0 | 1.85 | 1.9 | 2.41 | 2 | 1.77 | 1.58 | |
| StepDPOMode=Standard Mode2025.06 | 20.67 | 0 | 1.86 | 1.87 | 2.43 | 2.03 | 1.78 | 1.55 | |
| STaR-GATE-DMode=Standard Mode2025.06 | 20 | 0 | 1.55 | 1.63 | 1.95 | 1.79 | 1.51 | 1.31 | |
| STaR-GATEMode=Standard Mode2025.06 | 10.67 | 0 | 2.29 | 2.25 | 2.94 | 2.58 | 2.3 | 1.56 | |
| SwiftMode=Standard Mode2025.06 | 9.67 | 0 | 2.43 | 2.41 | 3.09 | 2.68 | 2.45 | 1.66 | |
| RefitMode=Standard Mode2025.06 | 6 | 0 | 2.43 | 2.5 | 3.15 | 2.68 | 2.41 | 1.63 | |
| BaseMode=Standard Mode2025.06 | 0 | 0 | 1.9 | 2.2 | 2.54 | 1.81 | 2.01 | 1.2 |