Safety Evaluation on X (test)
91.57F1 ScoreSingGuard-2B
Evaluation Results
| Method | Links | |
|---|---|---|
| SingGuard-2BAccess=Open-Source2026.06 | 91.57 | |
| SingGuard-4BAccess=Open-Source2026.06 | 91.46 | |
| SingGuard-8BAccess=Open-Source2026.06 | 91.46 | |
| Qwen3Guard-8B-looseAccess=Open-Source, Configuration=loose2026.06 | 91.03 | |
| Qwen3-VL-235BAccess=Open-Source2026.06 | 91.02 | |
| GuardReasoner-VL-7BAccess=Open-Source2026.06 | 90.67 | |
| Llama Guard 3Access=Open-Source2026.06 | 90.41 | |
| LlamaGuard32025.12 | 90.4 | |
| Qwen3Guard-8B-strictAccess=Open-Source, Configuration=strict2026.06 | 89.82 | |
| ShieldGemma-9BAccess=Open-Source2026.06 | 86.52 | |
| GGuard2025.12 | 86.4 | |
| YuFeng-XGuard-Reason-8BAccess=Open-Source2026.06 | 83.87 | |
| GPT-5.1Access=Closed-Source2026.06 | 66.67 | |
| Gemini3-ProAccess=Closed-Source2026.06 | 59.84 | |
| WildGuardAccess=Open-Source2026.06 | 59.29 | |
| OpenAI2025.12 | 55.8 | |
| GraniteGuardianAccess=Open-Source2026.06 | 29.77 |