Long-Context Generation on PG-19 20K context length
4Throughput Speedup (micro)SpecPV-4K
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SpecPV-4KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 4 | 3.57 | |
| SpecPV-2KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=2K2025.12 | 3.83 | 3.42 | |
| SpecPV-8KBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=8K2025.12 | 3.54 | 3.51 | |
| EAGLE3-YARNBackbone=LLaMA3.1-8B-Instruct2025.12 | 2.74 | 3.49 | |
| TriForceBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 2.44 | 2.82 | |
| TokenSwiftBackbone=LLaMA3.1-8B-Instruct, Partial KV Budget=4K2025.12 | 1.7 | 1.73 |