Long-Context Generation on PG-19 60K context length
6.29Throughput Speedup (micro)SpecPV-2K
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SpecPV-2KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=2K2025.12 | 6.29 | 3.57 | |
| SpecPV-4KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 6.15 | 3.56 | |
| SpecPV-8KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=8K2025.12 | 5.8 | 3.57 | |
| TriForceBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 3.17 | 2.61 | |
| EAGLE3-YARNBackbone=LLaMA3.1-8B-Instruct2025.12 | 3.11 | 3.48 | |
| TokenSwiftBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 2.11 | 1.75 |