Safety Classification on XSTestResponse
0.96F1 ScoreAprielGuard
Evaluation Results
| Method | Links | |
|---|---|---|
| AprielGuardSize=8B, Reasoning=true2025.12 | 0.96 | |
| AprielGuardSize=8B, Reasoning=false2025.12 | 0.95 | |
| Qwen3GuardSize=8B, Mode=loose, Reasoning=false2025.12 | 0.94 | |
| Qwen3GuardSize=8B, Mode=strict, Reasoning=false2025.12 | 0.92 | |
| Llama Guard 2Reasoning=false2025.12 | 0.91 | |
| Llama Guard 3Reasoning=false2025.12 | 0.9 | |
| Llama Guard 4Reasoning=false2025.12 | 0.9 | |
| IBM Granite Guardian 3.3Size=8B, Reasoning=false2025.12 | 0.9 | |
| IBM Granite Guardian 3.2Size=5B, Reasoning=false2025.12 | 0.89 | |
| gpt-oss-safeguardSize=20B, Reasoning=true2025.12 | 0.89 | |
| IBM Granite Guardian 3.1Size=2B, Reasoning=false2025.12 | 0.88 | |
| IBM Granite Guardian 3.3Size=8B, Reasoning=true2025.12 | 0.86 | |
| IBM Granite Guardian 3.2Size=3B, Reasoning=false2025.12 | 0.85 | |
| ShieldGemmaSize=9B, Reasoning=false2025.12 | 0.83 |