Mathematical Reasoning on MPMath
79.77AccuracyQwen2.5(14B)T
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5(14B)TGeneration Length=80962026.07 | 79.77 | — | |
| Qwen2.5(14B)TGeneration Length=10242026.07 | 78.86 | — | |
| RG-OPD (ours)Generation Length=10242026.07 | 42.91 | 5.58 | |
| RG-OPD (ours)Generation Length=80962026.07 | 42.88 | 5.5 | |
| TSD-KD [10]Generation Length=10242026.07 | 38.91 | 1.58 | |
| TSD-KD [10]Generation Length=80962026.07 | 38.91 | 1.53 | |
| Qwen2.5(1.5B)SGeneration Length=80962026.07 | 37.38 | 0 | |
| Qwen2.5(1.5B)SGeneration Length=10242026.07 | 37.33 | 0 | |
| Reverse KL [1]Generation Length=80962026.07 | 37.28 | -0.1 | |
| Reverse KL [1]Generation Length=10242026.07 | 37.26 | -0.07 | |
| TSD-KD [10]*Generation Length=10242026.07 | 36.9 | -0.43 |