Math on MATH500 (Speedup, τ)
8.5SpeedupBASTION
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BASTIONModel=Qwen3-4B, Temperature=1, Hardware=H1002026.05 | 8.5 | 10.15 | |
| BASTIONModel=Qwen3-4B, Temperature=0, Hardware=H1002026.05 | 8.05 | 10.2 | |
| BASTIONModel=Qwen3-8B, Temperature=0, Hardware=H1002026.05 | 8.02 | 10.21 | |
| BASTIONModel=Qwen3-8B, Temperature=1, Hardware=H1002026.05 | 7.59 | 10.2 | |
| DFlashModel=Qwen3-4B, Temperature=0, Hardware=H100, Block size=162026.05 | 6.96 | 7.79 | |
| DFlashModel=Qwen3-8B, Temperature=0, Hardware=H100, Block size=162026.05 | 6.86 | 7.93 | |
| DFlashModel=Qwen3-4B, Temperature=1, Hardware=H100, Block size=162026.05 | 5.59 | 6.6 | |
| DFlashModel=Qwen3-8B, Temperature=1, Hardware=H100, Block size=162026.05 | 5.43 | 6.45 | |
| DFlash-TfMTemperature=0.0, Reasoning Mode=Reasoning, Path=Tree, Budget=642026.07 | 5.3 | 11.05 | |
| DFlash-TfMTemperature=0.0, Reasoning Mode=No Reasoning, Path=Tree, Budget=642026.07 | 5.14 | 11.06 | |
| EAGLE-3Model=Qwen3-4B, Temperature=0, Hardware=H100, Tree budget size=602026.05 | 3.04 | 2.54 | |
| EAGLE-3Model=Qwen3-4B, Temperature=1, Hardware=H100, Tree budget size=602026.05 | 3.03 | 2.41 | |
| EAGLE-3Model=Qwen3-8B, Temperature=0, Hardware=H100, Tree budget size=602026.05 | 2.83 | 2.48 | |
| EAGLE-3Model=Qwen3-8B, Temperature=1, Hardware=H100, Tree budget size=602026.05 | 2.75 | 2.35 |