Mathematical Reasoning on MATH 500 (Accuracy, Avg. Tokens, E3)
93.16AccuracyVanilla
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VanillaBackbone=o4-mini, Mode=Direct2025.05 | 93.16 | 711.2 | 12.2 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-LLaMA-70B, Scheduling Strategy=Exponential2025.05 | 93.04 | 1,469.29 | 5.89 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-LLaMA-70B, Scheduling Strategy=Cosine2025.05 | 92.88 | 1,487.83 | 5.8 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-LLaMA-70B, Scheduling Strategy=Weighted2025.05 | 92.64 | 1,557.64 | 5.51 | |
| Global BudgetBackbone=DeepSeek-R1-Distill-LLaMA-70B, Mode=Planned2025.05 | 92.56 | 1,661.24 | 5.16 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-LLaMA-70B, Scheduling Strategy=Linear2025.05 | 92.32 | 1,529.98 | 5.57 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-LLaMA-70B, Scheduling Strategy=Uniform2025.05 | 92.28 | 1,575.04 | 5.41 | |
| VanillaBackbone=DeepSeek-R1-Distill-LLaMA-70B, Mode=Planned2025.05 | 92.12 | 2,022.38 | 4.2 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-LLaMA-70B, Scheduling Strategy=Polynomial2025.05 | 91.92 | 1,514.43 | 5.58 | |
| VanillaBackbone=o4-mini, Mode=Planned2025.05 | 91.88 | 539.36 | 15.65 | |
| Global BudgetBackbone=o4-mini, Mode=Direct2025.05 | 91.84 | 636.41 | 13.25 | |
| Global BudgetBackbone=o4-mini, Mode=Planned2025.05 | 91.84 | 586.18 | 14.39 | |
| PLAN-AND-BUDGETBackbone=o4-mini, Scheduling Strategy=Weighted2025.05 | 91.64 | 538.22 | 15.6 | |
| PLAN-AND-BUDGETBackbone=o4-mini, Scheduling Strategy=Uniform2025.05 | 91.36 | 525.53 | 15.88 | |
| PLAN-AND-BUDGETBackbone=o4-mini, Scheduling Strategy=Cosine2025.05 | 91.32 | 522.89 | 15.95 | |
| Global BudgetBackbone=DeepSeek-R1-Distill-Qwen-32B, Mode=Planned2025.05 | 91.24 | 1,552.62 | 5.36 | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-32B, Mode=Planned2025.05 | 91.04 | 1,883.73 | 4.4 | |
| PLAN-AND-BUDGETBackbone=o4-mini, Scheduling Strategy=Exponential2025.05 | 90.88 | 525.51 | 15.72 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-Qwen-32B, Scheduling Strategy=Exponential2025.05 | 90.8 | 1,389.75 | 5.93 | |
| Global BudgetBackbone=DeepSeek-R1-Distill-LLaMA-70B, Mode=Direct2025.05 | 90.8 | 1,810.83 | 4.55 | |
| Global BudgetBackbone=QwQ-32B, Mode=Direct2025.05 | 90.56 | 2,565.18 | 3.2 | |
| PLAN-AND-BUDGETBackbone=o4-mini, Scheduling Strategy=Linear2025.05 | 90.56 | 534.45 | 15.34 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-Qwen-32B, Scheduling Strategy=Weighted2025.05 | 90.48 | 1,485.99 | 5.51 | |
| VanillaBackbone=DeepSeek-R1-Distill-LLaMA-70B, Mode=Direct2025.05 | 90.44 | 2,286.63 | 3.58 | |
| PLAN-AND-BUDGETBackbone=o4-mini, Scheduling Strategy=Polynomial2025.05 | 90.36 | 525 | 15.55 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-Qwen-32B, Scheduling Strategy=Uniform2025.05 | 90.16 | 1,440.7 | 5.64 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-Qwen-32B, Scheduling Strategy=Linear2025.05 | 90.04 | 1,336.27 | 6.07 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-Qwen-32B, Scheduling Strategy=Polynomial2025.05 | 90.04 | 1,371.59 | 5.91 | |
| PLAN-AND-BUDGETBackbone=DeepSeek-R1-Distill-Qwen-32B, Scheduling Strategy=Cosine2025.05 | 89.88 | 1,365.51 | 5.92 | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-32B, Mode=Direct2025.05 | 89.76 | 2,105.12 | 3.83 | |
| Global BudgetBackbone=DeepSeek-R1-Distill-Qwen-32B, Mode=Direct2025.05 | 89.6 | 1,526.15 | 5.26 | |
| PLAN-AND-BUDGETBackbone=QwQ-32B, Scheduling Strategy=Uniform2025.05 | 88.68 | 2,397.16 | 3.28 | |
| PLAN-AND-BUDGETBackbone=QwQ-32B, Scheduling Strategy=Cosine2025.05 | 88.6 | 2,306.83 | 3.4 | |
| PLAN-AND-BUDGETBackbone=QwQ-32B, Scheduling Strategy=Polynomial2025.05 | 88.27 | 2,346.94 | 3.32 | |
| Global BudgetBackbone=QwQ-32B, Mode=Planned2025.05 | 88.2 | 2,671.6 | 2.91 | |
| PLAN-AND-BUDGETBackbone=QwQ-32B, Scheduling Strategy=Linear2025.05 | 88.13 | 2,346.35 | 3.31 | |
| PLAN-AND-BUDGETBackbone=QwQ-32B, Scheduling Strategy=Exponential2025.05 | 87.9 | 2,320.04 | 3.33 | |
| PLAN-AND-BUDGETBackbone=QwQ-32B, Scheduling Strategy=Weighted2025.05 | 87.45 | 2,479.46 | 3.08 | |
| VanillaBackbone=QwQ-32B, Mode=Planned2025.05 | 85.3 | 3,309.69 | 2.2 | |
| VanillaBackbone=QwQ-32B, Mode=Direct2025.05 | 84.88 | 3,523.72 | 2.04 |