Safety Evaluation on Attack Benchmarks Prompt
100StrongRYuFeng-XGuard
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| YuFeng-XGuardModel Scale=8B2026.01 | 100 | 98.4 | 94.3 | 97.6 | |
| NemotronReasoningModel Scale=4B2026.01 | 99.8 | 93.6 | 82.5 | 92 | |
| Qwen3Guard-GenModel Scale=4B, Evaluation Policy=strict2026.01 | 99.8 | 94.3 | 84.5 | 92.9 | |
| Qwen3Guard-GenModel Scale=8B, Evaluation Policy=strict2026.01 | 99.8 | 95.2 | 85.9 | 93.7 | |
| Qwen3Guard-GenModel Scale=0.6B, Evaluation Policy=strict2026.01 | 99.7 | 94.9 | 81.3 | 92 | |
| YuFeng-XGuardModel Scale=0.6B2026.01 | 99.7 | 97.6 | 94.4 | 97.2 | |
| PolyGuardModel Scale=7B2026.01 | 99.7 | 97.5 | 85.5 | 94.2 | |
| WildGuardModel Scale=7B2026.01 | 99.5 | 80 | 77.8 | 85.8 | |
| NemotronGuardV2Model Scale=8B2026.01 | 99.5 | 83.5 | 66.8 | 83.3 | |
| GPT-OSS-SafeGuardModel Scale=20B2026.01 | 99.2 | 91.4 | 86.7 | 92.4 | |
| Qwen3Guard-GenModel Scale=8B, Evaluation Policy=loose2026.01 | 99 | 68.2 | 68.5 | 78.6 | |
| Qwen3Guard-GenModel Scale=4B, Evaluation Policy=loose2026.01 | 98.9 | 70.3 | 70 | 79.7 | |
| Qwen3Guard-GenModel Scale=0.6B, Evaluation Policy=loose2026.01 | 98.5 | 77.9 | 70.2 | 82.2 | |
| Llama3GuardModel Scale=8B2026.01 | 98.5 | 71.6 | 61.7 | 77.3 | |
| Llama4GuardModel Scale=12B2026.01 | 95.3 | 61.4 | 60.6 | 72.4 | |
| ShieldGemmaModel Scale=9B2026.01 | 90 | 46.9 | 44.4 | 60.4 |