Summarization on XSum (Speedup vs. AR)
1.84Speedup vs ARHiSpec
Evaluation Results
| Method | Links | |
|---|---|---|
| HiSpecModel=Llama3-8B2025.10 | 1.84 | |
| HiSpecModel=Llama2-70B2025.10 | 1.63 | |
| HiSpecModel=Llama2-13B2025.10 | 1.55 | |
| HiSpecModel=Llama2-7B2025.10 | 1.5 | |
| LookAheadModel=CodeLlama-34B2025.10 | 1.49 | |
| LookAheadModel=Llama2-13B2025.10 | 1.45 | |
| HiSpecModel=CodeLlama-34B2025.10 | 1.45 | |
| LayerSkipModel=Llama2-7B2025.10 | 1.43 | |
| LookAheadModel=Llama2-7B2025.10 | 1.43 | |
| LayerSkipModel=CodeLlama-34B2025.10 | 1.42 | |
| LookAheadModel=Llama3-8B2025.10 | 1.37 | |
| AdaDecodeModel=CodeLlama-34B2025.10 | 1.37 | |
| LayerSkipModel=Llama3-8B2025.10 | 1.31 | |
| LayerSkipModel=Llama2-13B2025.10 | 1.29 | |
| LayerSkipModel=Llama2-70B2025.10 | 1.29 | |
| SWIFTModel=Llama2-70B2025.10 | 1.1 | |
| AdaDecodeModel=Llama3-8B2025.10 | 1.06 | |
| SWIFTModel=CodeLlama-34B2025.10 | 1.05 | |
| SWIFTModel=Llama2-13B2025.10 | 0.97 |