List your company

Jailbreak Attack on MM-SafetyBench (tiny)

99.16ASRMIDAS
-3.966422.806849.5876.3532Feb 28, 2026

Evaluation results

MethodLinks
MIDASTarget Model=Gemini-2.5-FT2026.0299.164.35
MIDASTarget Model=QVQ-Max2026.0298.654.21
MIDASTarget Model=Gemini-2.5-Pro2026.0292.173.94
VisCRATarget Model=QVQ-Max2026.0282.23.53
MIDASTarget Model=GPT-5-Chat2026.0281.543.49
MIDASTarget Model=GPT-4o2026.0261.072.53
VisCRATarget Model=Gemini-2.5-FT2026.0249.72.29
HIMRDTarget Model=Gemini-2.5-Pro2026.0238.31.38
VisCRATarget Model=GPT-4o2026.0237.121.68
VisCRATarget Model=Gemini-2.5-Pro2026.0235.921.69
HIMRDTarget Model=GPT-4o2026.0226.40.95
HIMRDTarget Model=GPT-5-Chat2026.0226.40.97
FigStepTarget Model=QVQ-Max2026.0222.780.94
FigStepTarget Model=Gemini-2.5-FT2026.0220.560.71
VisCRATarget Model=GPT-5-Chat2026.0220.240.97
HIMRDTarget Model=QVQ-Max2026.0216.81.1
SI-AttackTarget Model=QVQ-Max2026.0213.941.29
FigStepTarget Model=GPT-4o2026.0211.930.51
FigStepTarget Model=Gemini-2.5-Pro2026.0211.820.46
FigStepTarget Model=GPT-5-Chat2026.0211.820.49
HIMRDTarget Model=Gemini-2.5-FT2026.0210.20.67
SI-AttackTarget Model=Gemini-2.5-FT2026.025.520.89
SI-AttackTarget Model=GPT-5-Chat2026.025.520.63
SI-AttackTarget Model=Gemini-2.5-Pro2026.020.60.86
SI-AttackTarget Model=GPT-4o2026.0200.38