Mathematical Reasoning on MATH500 (Pass@1, #Tk)
62Pass@1 RateDRP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DRPBackbone=Phi-4-Reasoning2026.01 | 62 | 2,122 | |
| DRPBackbone=GPT-oss-20B2026.01 | 61.46 | 4,137 | |
| FROSTBackbone=Phi-4-Reasoning2026.01 | 59.8 | 344.37 | |
| TALEBackbone=Phi-4-Reasoning2026.01 | 58 | 1,874.43 | |
| FROSTBackbone=GPT-oss-20B2026.01 | 58 | 680.89 | |
| BaseBackbone=Phi-4-Reasoning2026.01 | 54.8 | 1,721.95 | |
| TALEBackbone=GPT-oss-20B2026.01 | 54.54 | 3,878.87 | |
| ThinkLessBackbone=Phi-4-Reasoning2026.01 | 54.14 | 1,101.21 | |
| ThinkLessBackbone=GPT-oss-20B2026.01 | 54.1 | 2,206 | |
| BaseBackbone=GPT-oss-20B2026.01 | 54 | 1,575.36 | |
| SelfBudgeterBackbone=Phi-4-Reasoning2026.01 | 53.47 | 1,195.18 | |
| SelfBudgeterBackbone=GPT-oss-20B2026.01 | 53.4 | 2,285 | |
| TALEBackbone=Magistral-Small-1.12026.01 | 30.4 | 723.91 | |
| FROSTBackbone=Magistral-Small-1.12026.01 | 30.4 | 98.2 | |
| ThinkLessBackbone=Magistral-Small-1.12026.01 | 25 | 1,405 | |
| SelfBudgeterBackbone=Magistral-Small-1.12026.01 | 23 | 1,520 | |
| DRPBackbone=Magistral-Small-1.12026.01 | 21 | 1,680.33 | |
| BaseBackbone=Magistral-Small-1.12026.01 | 14.8 | 1,389.89 |