Speculative Decoding on CNN_DM
1MATLLaMA-3.1-8B-IT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LLaMA-3.1-8B-ITTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.10 | 1 | 18.41 | |
| Qwen-3-8BTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.10 | 1 | 14.7 | |
| Qwen-3-32BTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.10 | 1 | 8.51 | |
| LLaMA-3.1-8B-ITBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.10 | 1 | 17.86 | |
| Qwen-3-8BBackbone=Qwen-3-8B, Inference Setting=Greedy2025.10 | 1 | 14.49 | |
| Qwen-3-32BBackbone=Qwen-3-32B, Inference Setting=Greedy2025.10 | 1 | 8.06 | |
| Vanilla EagleTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.10 | 2.58 | 19.31 | |
| EagleBackbone=Qwen-3-32B, Inference Setting=Greedy2025.10 | 2.59 | 18.67 | |
| EXP3SpecTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.10 | 3.33 | 36.94 | |
| EXP3SpecTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.10 | 3.37 | 24.57 | |
| EXP3SpecBackbone=Qwen-3-8B, Inference Setting=Greedy2025.10 | 3.39 | 37.53 | |
| UCBSpecTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.10 | 3.45 | 24.97 | |
| UCBSpecTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.10 | 3.52 | 39.43 | |
| EXP3SpecBackbone=Qwen-3-32B, Inference Setting=Greedy2025.10 | 3.52 | 25.32 | |
| UCBSpecBackbone=Qwen-3-8B, Inference Setting=Greedy2025.10 | 3.58 | 41.07 | |
| UCBSpecBackbone=Qwen-3-32B, Inference Setting=Greedy2025.10 | 3.64 | 26.93 | |
| UCBSpecTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.10 | 3.72 | 45.4 | |
| EXP3SpecBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.10 | 3.81 | 49.7 | |
| EXP3SpecTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.10 | 3.82 | 46.25 | |
| UCBSpecBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.10 | 3.94 | 50.47 | |
| Vanilla EagleTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.10 | 4.05 | 45.47 | |
| EagleBackbone=Qwen-3-8B, Inference Setting=Greedy2025.10 | 4.07 | 46.2 | |
| HedgeSpecTarget Model=Qwen-3-32B, Inference Mode=Non-greedy2025.10 | 4.78 | 33.48 | |
| HedgeSpecTarget Model=Qwen-3-8B, Inference Mode=Non-greedy2025.10 | 4.96 | 53.5 | |
| Vanilla EagleTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.10 | 5.06 | 62.55 | |
| HedgeSpecBackbone=Qwen-3-8B, Inference Setting=Greedy2025.10 | 5.1 | 54.97 | |
| HedgeSpecBackbone=Qwen-3-32B, Inference Setting=Greedy2025.10 | 5.19 | 33.38 | |
| EagleBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.10 | 5.31 | 67.26 | |
| HedgeSpecTarget Model=LLaMA-3.1-8B-IT, Inference Mode=Non-greedy2025.10 | 5.58 | 65.11 | |
| HedgeSpecBackbone=LLaMA-3.1-8B-IT, Inference Setting=Greedy2025.10 | 5.88 | 70.68 |