Long-Context Generation on PG-19 40K context length
5.46Throughput Speedup (micro)SpecPV-4K
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SpecPV-4KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 5.46 | 3.67 | |
| SpecPV-2KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=2K2025.12 | 5.21 | 3.55 | |
| SpecPV-8KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=8K2025.12 | 4.88 | 3.56 | |
| TriForceBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 3.1 | 2.8 | |
| EAGLE3-YARNBackbone=LLaMA3.1-8B-Instruct2025.12 | 3.02 | 3.57 | |
| TokenSwiftBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 1.98 | 1.77 |