Multilingual text query safety on RTP-LX
84.65AccuracySingGuard-8B
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SingGuard-8BSource=Open-Source Models, Model Scale=8B2026.06 | 84.65 | 93.16 | 85.95 | 89.41 | 88.72 | |
| SingGuard-4BSource=Open-Source Models, Model Scale=4B2026.06 | 81.7 | 93.65 | 81.25 | 87.01 | 87.3 | |
| Qwen3Guard-8B-strictSource=Open-Source Models, Model Scale=8B, Safety Policy=strict2026.06 | 80.24 | 89.85 | 83.2 | 86.4 | 87.12 | |
| SingGuard-2BSource=Open-Source Models, Model Scale=2B2026.06 | 80.1 | 93.99 | 78.65 | 85.64 | 85.49 | |
| GraniteGuardianSource=Open-Source Models2026.06 | 79.95 | 82.52 | 93.16 | 87.51 | 83.71 | |
| YuFeng-XGuard-Reason-8BSource=Open-Source Models, Model Scale=8B2026.06 | 79.15 | 94.6 | 76.73 | 84.73 | 86.48 | |
| Gemini3-ProSource=Closed-Source Models2026.06 | 71.18 | 93.09 | 68.55 | 78.96 | 78.19 | |
| GuardReasoner-VL-7BSource=Open-Source Models, Model Scale=7B2026.06 | 66.08 | 86.27 | 65.45 | 74.43 | 80.98 | |
| GPT-5.1Source=Closed-Source Models2026.06 | 57.41 | 97.15 | 44.85 | 61.36 | 70.07 | |
| WildGuardSource=Open-Source Models2026.06 | 52.85 | 97.34 | 38.54 | 55.22 | 66.07 | |
| Qwen3Guard-8B-looseSource=Open-Source Models, Model Scale=8B, Safety Policy=loose2026.06 | 50.65 | 97.62 | 35.42 | 51.98 | 68.03 | |
| Llama Guard 3Source=Open-Source Models2026.06 | 45.71 | 95.45 | 29.42 | 44.97 | 58.63 | |
| ShieldGemma-9BSource=Open-Source Models, Model Scale=9B2026.06 | 44.54 | 98.15 | 26.98 | 42.32 | 46.07 | |
| Qwen3-VL-235BSource=Open-Source Models, Model Scale=235B2026.06 | 35.7 | 97.13 | 16.86 | 28.73 | 31.38 |