Question Answering on NQ (Efficiency Metrics)
3.51SpeedupSpecBlock+adapt
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SpecBlock+adaptModel=Llama-3.1-8B, Temperature=02026.05 | 3.51 | 5.41 | 15 | |
| SpecBlock+OSDModel=Llama-3.1-8B, Temperature=02026.05 | 3.25 | 5.41 | 18 | |
| SpecBlockModel=Llama-3.1-8B, Temperature=02026.05 | 3 | 4.4 | 16 | |
| SpecBlock+adaptModel=Qwen3-8B, Temperature=02026.05 | 2.69 | 3.75 | 17 | |
| SpecBlock+adaptModel=Qwen3-32B, Temperature=02026.05 | 2.57 | 3.61 | 13 | |
| EAGLE-3Model=Qwen3-8B, Temperature=02026.05 | 2.5 | 4.2 | 29 | |
| SpecBlock+OSDModel=Qwen3-8B, Temperature=02026.05 | 2.49 | 3.75 | 19 | |
| EAGLE-3Model=Llama-3.1-8B, Temperature=02026.05 | 2.42 | 5.39 | 31 | |
| SpecBlock+adaptModel=Llama-3.1-8B, Temperature=1.02026.05 | 2.39 | 4.57 | 15 | |
| EAGLE-3Model=Qwen3-32B, Temperature=02026.05 | 2.38 | 3.96 | 24 | |
| SpecBlock+OSDModel=Qwen3-32B, Temperature=02026.05 | 2.36 | 3.61 | 15 | |
| SpecBlock+adaptModel=Qwen3-8B, Temperature=1.02026.05 | 2.27 | 3.55 | 18 | |
| SpecBlockModel=Qwen3-8B, Temperature=02026.05 | 2.26 | 3.21 | 14 | |
| SpecBlock+adaptModel=Qwen3-32B, Temperature=1.02026.05 | 2.19 | 3.4 | 13 | |
| SpecBlockModel=Qwen3-32B, Temperature=02026.05 | 2.17 | 3.07 | 11 | |
| SpecBlock+OSDModel=Llama-3.1-8B, Temperature=1.02026.05 | 2.17 | 4.59 | 17 | |
| FalconModel=Llama-3.1-8B, Temperature=02026.05 | 2.12 | 3.38 | 26 | |
| SpecBlock+OSDModel=Qwen3-8B, Temperature=1.02026.05 | 2.06 | 3.55 | 19 | |
| SpecBlock+OSDModel=Qwen3-32B, Temperature=1.02026.05 | 1.99 | 3.41 | 14 | |
| EAGLE-3Model=Qwen3-8B, Temperature=1.02026.05 | 1.94 | 4.09 | 29 | |
| SpecBlockModel=Llama-3.1-8B, Temperature=1.02026.05 | 1.92 | 3.6 | 14 | |
| EAGLE-3Model=Qwen3-32B, Temperature=1.02026.05 | 1.87 | 3.95 | 25 | |
| SpecBlockModel=Qwen3-8B, Temperature=1.02026.05 | 1.81 | 3.06 | 14 | |
| ParallelSpecModel=Llama-3.1-8B, Temperature=02026.05 | 1.78 | 2.63 | 8 | |
| SpecBlockModel=Qwen3-32B, Temperature=1.02026.05 | 1.75 | 2.92 | 11 | |
| EAGLE-3Model=Llama-3.1-8B, Temperature=1.02026.05 | 1.66 | 4.27 | 30 | |
| MedusaModel=Llama-3.1-8B, Temperature=02026.05 | 1.44 | 2 | 6 | |
| FalconModel=Llama-3.1-8B, Temperature=1.02026.05 | 1.44 | 2.77 | 27 | |
| ParallelSpecModel=Llama-3.1-8B, Temperature=1.02026.05 | 1.2 | 2.2 | 9 | |
| SpSModel=Llama-3.1-8B, Temperature=02026.05 | 1.06 | 2.09 | 38 | |
| MedusaModel=Llama-3.1-8B, Temperature=1.02026.05 | 1.01 | 1.67 | 7 | |
| SpSModel=Llama-3.1-8B, Temperature=1.02026.05 | 0.7 | 1.76 | 40 |