List your company

MLLM Safety Evaluation on OOD-MMSafe Malicious Mode 1.0 (test)

1.97RA Scoreo4-Mini
0.58680.94591.3051.6641Mar 10, 2026

Evaluation results

MethodLinks
o4-MiniModel Type=Closed-source2026.031.970.21.980.91.990
GPT-5.1Model Type=Closed-source2026.031.961.51.961.320
Qwen3VL-8BModel Type=Open-source2026.031.913.71.923.31.980.4
Gemini-3-ProModel Type=Closed-source2026.031.883.51.933.11.732.9
Gemma-3-12BModel Type=Open-source2026.031.873.71.789.21.950.4
Gemini-3-Pro-ThinkModel Type=Closed-source2026.031.855.71.875.41.831.8
Qwen3VL-32BModel Type=Open-source2026.031.854.81.941.93.5
Qwen3VL-4BModel Type=Open-source2026.031.828.11.88.61.913.1
Doubao-1.6Model Type=Closed-source2026.031.86.81.827.31.91.1
QwenVL2.5-7BModel Type=Open-source2026.031.2311.237.81.617.7
InternVL-3.5-38BModel Type=Open-source2026.031.09291.6217.81.2418.5
InternVL-3.5-8BModel Type=Open-source2026.030.9936.31.31331.3112.3
LLaVA-1.5-7BModel Type=Open-source2026.030.6454.50.5770.11.0222.9