Chat on Alpaca (Speedup, τ)
3.59SpeedupBASTION
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BASTIONModel=Qwen3-4B, Temperature=1, Hardware=H1002026.05 | 3.59 | 4.71 | |
| BASTIONModel=Qwen3-4B, Temperature=0, Hardware=H1002026.05 | 3.58 | 4.73 | |
| BASTIONModel=Qwen3-8B, Temperature=1, Hardware=H1002026.05 | 3.39 | 4.73 | |
| BASTIONModel=Qwen3-8B, Temperature=0, Hardware=H1002026.05 | 3.35 | 4.82 | |
| EAGLE-3Model=Qwen3-4B, Temperature=0, Hardware=H100, Tree budget size=602026.05 | 2.55 | 2.17 | |
| DFlashModel=Qwen3-4B, Temperature=0, Hardware=H100, Block size=162026.05 | 2.49 | 3.09 | |
| DFlashModel=Qwen3-8B, Temperature=0, Hardware=H100, Block size=162026.05 | 2.47 | 3.09 | |
| DFlashModel=Qwen3-4B, Temperature=1, Hardware=H100, Block size=162026.05 | 2.4 | 2.95 | |
| DFlashModel=Qwen3-8B, Temperature=1, Hardware=H100, Block size=162026.05 | 2.38 | 2.94 | |
| EAGLE-3Model=Qwen3-4B, Temperature=1, Hardware=H100, Tree budget size=602026.05 | 2.23 | 2.08 | |
| EAGLE-3Model=Qwen3-8B, Temperature=0, Hardware=H100, Tree budget size=602026.05 | 2.05 | 1.73 | |
| EAGLE-3Model=Qwen3-8B, Temperature=1, Hardware=H100, Tree budget size=602026.05 | 1.79 | 1.62 |