LLM Routing on OOD datasets (test)
89AccuracyOracle
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OracleScenario=Performance First, Alpha (α)=0.22025.10 | 89 | 0.33 | 82 | |
| DiSRouterScenario=Performance First, Alpha (α)=0.2, Protocol=+ RL2025.10 | 75 | 0.67 | 62 | |
| Largest LLMScenario=Performance First, Alpha (α)=0.22025.10 | 74 | 0.9 | 56 | |
| AutomixScenario=Performance First, Alpha (α)=0.22025.10 | 74 | 1.1 | 52 | |
| DiSRouterScenario=Performance First, Alpha (α)=0.2, Protocol=SFT2025.10 | 73 | 0.62 | 61 | |
| FORCScenario=Performance First, Alpha (α)=0.22025.10 | 70 | 0.75 | 56 | |
| GraphRouterScenario=Performance First, Alpha (α)=0.22025.10 | 69 | 0.67 | 56 | |
| FrugalGPTScenario=Performance First, Alpha (α)=0.22025.10 | 63 | 0.41 | 54 | |
| RandomScenario=Performance First, Alpha (α)=0.22025.10 | 59 | 0.46 | 50 | |
| RouteLLMScenario=Performance First, Alpha (α)=0.22025.10 | 42 | 0.22 | 38 | |
| Smallest LLMScenario=Performance First, Alpha (α)=0.22025.10 | 36 | 0.1 | 34 |