Multimodal Safety on MM-Safety
99.32F1 ScoreLLaVAShield
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVAShieldCategory=Open-Source Models2026.06 | 99.32 | |
| ShieldGemma-2Category=Open-Source Models2026.06 | 96.56 | |
| Qwen3-VL-235BCategory=Open-Source Models2026.06 | 88.18 | |
| SingGuard-8BCategory=Our Models2026.06 | 88.12 | |
| Gemini3-ProCategory=Closed-Source Models2026.06 | 87.65 | |
| SingGuard-4BCategory=Our Models2026.06 | 86.05 | |
| GPT-5.1Category=Closed-Source Models2026.06 | 85.68 | |
| SingGuard-2BCategory=Our Models2026.06 | 84.09 | |
| SafeGuard-VLCategory=Open-Source Models2026.06 | 68.66 | |
| GuardReasoner-VL-7BCategory=Open-Source Models2026.06 | 66.98 | |
| Qwen3-VL-8BCategory=Open-Source Models2026.06 | 64.06 | |
| LlamaGuard4Category=Open-Source Models2026.06 | 61.47 | |
| Qwen3-VL-4BCategory=Open-Source Models2026.06 | 61.03 | |
| LlamaGuard3-Vision-11BCategory=Open-Source Models2026.06 | 51.06 | |
| LlavaGuardCategory=Open-Source Models2026.06 | 27.27 |