Prompt-only Safety Routing on ToxicChat
56.82Routing F1SafeRoute
Evaluation Results
| Method | Links | |
|---|---|---|
| SafeRouteProtocol=Prompt-only, Small Model=Llama-Guard-3-1B, Large Model=Llama-Guard-3-8B2025.02 | 56.82 | |
| SafeRoutesmall_model=Llama-Guard-3-1B, large_model=Granite-Guardian-3-8B2025.02 | 48.87 | |
| EntropyProtocol=Prompt-only, Small Model=Llama-Guard-3-1B, Large Model=Llama-Guard-3-8B2025.02 | 40.02 | |
| Entropysmall_model=Llama-Guard-3-1B, large_model=Granite-Guardian-3-8B2025.02 | 38.99 | |
| CCProtocol=Prompt-only, Small Model=Llama-Guard-3-1B, Large Model=Llama-Guard-3-8B2025.02 | 31.35 | |
| +CCsmall_model=Llama-Guard-3-1B, large_model=Granite-Guardian-3-8B2025.02 | 31.25 | |
| +TSsmall_model=Llama-Guard-3-1B, large_model=Granite-Guardian-3-8B2025.02 | 22.77 | |
| +BCsmall_model=Llama-Guard-3-1B, large_model=Granite-Guardian-3-8B2025.02 | 20.64 | |
| TSProtocol=Prompt-only, Small Model=Llama-Guard-3-1B, Large Model=Llama-Guard-3-8B2025.02 | 20.04 | |
| BCProtocol=Prompt-only, Small Model=Llama-Guard-3-1B, Large Model=Llama-Guard-3-8B2025.02 | 18.54 |