List your company

MLLM Jailbreaking on MM-SafetyBench Physical Harm scenario

6ASRVanilla
2.4826.245073.76Mar 26, 2025

Evaluation results

MethodLinks
VanillaTarget Model=o12025.036
FigStepTarget Model=o12025.036
FigStep-ProTarget Model=o12025.036
HADESTarget Model=o12025.036
FigStep-ProTarget Model=GPT-4o2025.0310
VanillaTarget Model=GPT-4o2025.0316
FigStepTarget Model=GPT-4o2025.0316
HADESTarget Model=GPT-4o2025.0319
FigStepTarget Model=Qwen2-VL-7B2025.0352
JOODTarget Model=o12025.0352
VanillaTarget Model=Qwen2-VL-7B2025.0355
FigStep-ProTarget Model=Qwen2-VL-7B2025.0368
JOODTarget Model=GPT-4o2025.0374
HADESTarget Model=Qwen2-VL-7B2025.0379
JOODTarget Model=Qwen2-VL-7B2025.0394