Safety Classification on WildGuardMix-p (test)
93.2F1 ScoreOracle
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OracleCondition=Perfect Routing2025.02 | 93.2 | 92.4 | 93.9 | 30.81 | |
| SafeRouteRouting=SafeRoute (Ours), Small Model=Llama-Guard-3-1B, Large Model=Granite-Guardian-3-8B2025.02 | 84.8 | 87.9 | 81.9 | 12.19 | |
| TSRouting=Thresholding, Small Model=Llama-Guard-3-1B, Large Model=Granite-Guardian-3-8B2025.02 | 84.3 | 78 | 91.6 | 28.64 | |
| BCRouting=Bayesian-based, Small Model=Llama-Guard-3-1B, Large Model=Granite-Guardian-3-8B2025.02 | 83.3 | 79.8 | 87.1 | 16.77 | |
| Granite-Guardian-3-8BModel=Large (Granite-Guardian-3-8B)2025.02 | 83.2 | 75.7 | 92.3 | 25.65 | |
| CCRouting=Calibration-based, Small Model=Llama-Guard-3-1B, Large Model=Granite-Guardian-3-8B2025.02 | 83 | 81.7 | 84.2 | 16.57 | |
| EntRouting=Entropy-based, Small Model=Llama-Guard-3-1B, Large Model=Granite-Guardian-3-8B2025.02 | 82.4 | 83.7 | 81.2 | 19.61 | |
| RandomRouting=Random, Small Model=Llama-Guard-3-1B, Large Model=Granite-Guardian-3-8B2025.02 | 79.5 | 79.5 | 79.6 | 21.41 | |
| Llama-Guard-3-1BModel=Small (Llama-Guard-3-1B)2025.02 | 75.2 | 82.7 | 69 | 6.46 |