Safety Classification on XSTest, OpenAI, ToxicC., BeaverT., RTP-LX, XSafety French
84.4XSTest AccuracyLlamaGuard 3
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| LlamaGuard 3Size=8B2025.02 | 84.4 | 78.1 | 50.1 | 69.5 | 48.8 | 40.3 | 61.9 | |
| LlamaGuard 2Size=8B2025.02 | 81.6 | 74.5 | 39.7 | 68.6 | 40 | 35.4 | 56.6 | |
| DuoGuardSize=0.5B2025.02 | 79.2 | 67.1 | 62.8 | 81.3 | 91 | 54.7 | 72.7 | |
| ShieldGemmaSize=2B2025.02 | 63.3 | 36.8 | 28.7 | 50.1 | 21.5 | 23.9 | 37.4 | |
| LlamaGuard 3Size=1B2025.02 | 43 | 37.8 | 19.5 | 50.9 | 54.9 | 61.3 | 44.6 |