Long-Context Generation on PG-19 (50K context length)
5.79Throughput Speedup (micro)SpecPV-2K
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SpecPV-2KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=2K2025.12 | 5.79 | 3.56 | |
| SpecPV-4KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 5.66 | 3.52 | |
| SpecPV-8KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=8K2025.12 | 5.31 | 3.46 | |
| TriForceBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 3.42 | 2.82 | |
| EAGLE3-YARNBackbone=LLaMA3.1-8B-Instruct2025.12 | 3.06 | 3.46 | |
| TokenSwiftBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 2.06 | 1.77 |