LLM Serving on Evol-CodeAlpaca
234.09p90 Latency (ms/token)Oracle SJF
Evaluation Results
| Method | Links | |
|---|---|---|
| Oracle SJFModel=Llama, Workload Mode=Burst Mode2025.09 | 234.09 | |
| Oracle SJFModel=DeepSeek-R1, Workload Mode=Burst Mode2025.09 | 269.06 | |
| PARSModel=Llama, Workload Mode=Burst Mode2025.09 | 279.26 | |
| Cross-Model PARSModel=Llama, Workload Mode=Burst Mode2025.09 | 290.88 | |
| Pointwise SJFModel=Llama, Workload Mode=Burst Mode2025.09 | 292.19 | |
| Listwise SJFModel=Llama, Workload Mode=Burst Mode2025.09 | 310.87 | |
| PARSModel=DeepSeek-R1, Workload Mode=Burst Mode2025.09 | 327.73 | |
| Listwise SJFModel=DeepSeek-R1, Workload Mode=Burst Mode2025.09 | 412.53 | |
| Cross-Model PARSModel=DeepSeek-R1, Workload Mode=Burst Mode2025.09 | 435.29 | |
| Pointwise SJFModel=DeepSeek-R1, Workload Mode=Burst Mode2025.09 | 446.98 | |
| FCFSModel=Llama, Workload Mode=Burst Mode2025.09 | 699.22 | |
| FCFSModel=DeepSeek-R1, Workload Mode=Burst Mode2025.09 | 860.7 |