List your company

Jailbreaking on HarmBench

84Attack Success Rate (ASR)MASKFORGE
9.6428.94548.2567.555May 5, 2026May 9, 2026May 14, 2026May 18, 2026May 23, 2026May 27, 2026Jun 1, 2026

Evaluation results

MethodLinks
MASKFORGEVictim model=LLaDA-Instruct2026.0684———4.74
UJEM-KLTarget Model=LLaVA-1.52026.0582.3————
UJEM-KLTarget Model=LLaVA-1.52026.0582.3————
MASKFORGEVictim model=Dream-Instruct2026.0680———4.49
UJEMTarget Model=LLaVA-1.52026.0579.4————
UJEMTarget Model=LLaVA-1.52026.0579.4————
UJEM-KLTarget Model=InternVL3.52026.0579.1————
UJEM-KLTarget Model=InternVL3.52026.0579.1————
MASKFORGEVictim model=LLaDA 1.52026.0679———4.74
MASKFORGEVictim model=AVG.2026.0679———4.64
SEATarget Model=LLaVA-1.52026.0578.5————
SEATarget Model=LLaVA-1.52026.0578.5————
UJEM-KLTarget Model=Qwen2.5-VL2026.0578.4————
UJEM-KLTarget Model=Qwen2.5-VL2026.0578.4————
MASKFORGEVictim model=LLaDA2.0-mini2026.0678———4.7
ForceTarget Model=LLaVA-1.52026.0577.9————
ForceTarget Model=LLaVA-1.52026.0577.9————
ForceTarget Model=InternVL3.52026.0577.3————
ForceTarget Model=InternVL3.52026.0577.3————
DIJAVictim model=MMaDA-CoT2026.0677.1———4.03
SEATarget Model=InternVL3.52026.0576.8————
SEATarget Model=InternVL3.52026.0576.8————
FigStepTarget Model=LLaVA-1.52026.0576.4————
FigStepTarget Model=LLaVA-1.52026.0576.4————
SEATarget Model=Qwen2.5-VL2026.0576.1————
SEATarget Model=Qwen2.5-VL2026.0576.1————
DIJAVictim model=LLaDA-Instruct2026.0675.8———4.03
DIJAVictim model=LLaDA 1.52026.0675.8———4.07
ForceTarget Model=Qwen2.5-VL2026.0574.8————
ForceTarget Model=Qwen2.5-VL2026.0574.8————
DIJAVictim model=Dream-Instruct2026.0674.6———4.01
UJEMTarget Model=InternVL3.52026.0574.2————
UJEMTarget Model=InternVL3.52026.0574.2————
MASKFORGEVictim model=MMaDA-CoT2026.0674———4.53
PAIRVictim model=MMaDA-CoT2026.0673———3.19
FigStepTarget Model=Qwen2.5-VL2026.0572.8————
FigStepTarget Model=Qwen2.5-VL2026.0572.8————
AutoDANVictim model=MMaDA-CoT2026.0672.5———3.71
UJEMTarget Model=Qwen2.5-VL2026.0571.6————
UJEMTarget Model=Qwen2.5-VL2026.0571.6————
AutoDANVictim model=LLaDA-Instruct2026.0671.5———3.52
FigStepTarget Model=InternVL3.52026.0570.1————
FigStepTarget Model=InternVL3.52026.0570.1————
PADVictim model=MMaDA-CoT2026.0669.7———3.98
UJATarget Model=LLaVA-1.52026.0568.9————
UJATarget Model=LLaVA-1.52026.0568.9————
AutoDANVictim model=LLaDA 1.52026.0668———3.45
Set TheoryTarget Model=Gemini-3-Flash, Encoding Category=LLM-based2026.0567.3————
Set TheoryTarget Model=Gemini-2.0-Flash, Encoding Category=LLM-based2026.0565.4————
UJATarget Model=Qwen2.5-VL2026.0565.3————
UJATarget Model=Qwen2.5-VL2026.0565.3————
DIJAVictim model=AVG.2026.0664.8———3.57
UJATarget Model=InternVL3.52026.0563.7————
UJATarget Model=InternVL3.52026.0563.7————
DIRECTVictim model=MMaDA-CoT2026.0662.3———3.48
Formal LogicTarget Model=Gemini-2.0-Flash, Encoding Category=LLM-based2026.0561.6————
Set TheoryTarget Model=GPT-4o-Mini, Encoding Category=LLM-based2026.0560.4————
Formal LogicTarget Model=Llama-3-8B, Encoding Category=LLM-based2026.0560.4————
AutoDANVictim model=AVG.2026.0660.3———3.3
Set TheoryTarget Model=Llama-3-8B, Encoding Category=LLM-based2026.0559.1————
Quantum MechanicsTarget Model=Gemini-3-Flash, Encoding Category=LLM-based2026.0559.1————
Formal LogicTarget Model=GPT-4o-Mini, Encoding Category=LLM-based2026.0558.5————
Formal LogicTarget Model=Gemini-3-Flash, Encoding Category=LLM-based2026.0558.5————
Formal LogicTarget Model=Claude-3.7-Sonnet, Encoding Category=LLM-based2026.0557.2————
Set TheoryTarget Model=Claude-3.7-Sonnet, Encoding Category=LLM-based2026.0556.6————
PADVictim model=Dream-Instruct2026.0654.7———3.94
AutoDANVictim model=Dream-Instruct2026.0653———3.2
PADVictim model=LLaDA-Instruct2026.0652.7———3.76
PADVictim model=AVG.2026.0652.2———3.72
PADVictim model=LLaDA 1.52026.0651.7———3.65
Quantum MechanicsTarget Model=Claude-3.7-Sonnet, Encoding Category=LLM-based2026.0550.9————
Quantum MechanicsTarget Model=GPT-4o-Mini, Encoding Category=LLM-based2026.0550.3————
Quantum MechanicsTarget Model=Gemini-2.0-Flash, Encoding Category=LLM-based2026.0549.7————
Set TheoryTarget Model=Vicuna-13B, Encoding Category=LLM-based2026.0544.7————
Quantum MechanicsTarget Model=Llama-3-8B, Encoding Category=LLM-based2026.0544.7————
Formal LogicTarget Model=Vicuna-13B, Encoding Category=LLM-based2026.0540.9————
AutoDANVictim model=LLaDA2.0-mini2026.0636.5———2.62
PAIRVictim model=LLaDA 1.52026.0635———2
PADVictim model=LLaDA2.0-mini2026.0632.3———3.28
Formal LogicTarget Model=GPT-5, Encoding Category=LLM-based2026.0532.1————
PAIRVictim model=AVG.2026.0631.8———1.94
Formal LogicTarget Model=GPT-5-Mini, Encoding Category=LLM-based2026.0531.4————
PAIRVictim model=LLaDA-Instruct2026.0630.5———1.85
DIRECTVictim model=AVG.2026.0628.7———1.98
Set TheoryTarget Model=GPT-5, Encoding Category=LLM-based2026.0527.7————
DIRECTVictim model=LLaDA-Instruct2026.0627.2———1.79
DIRECTVictim model=LLaDA 1.52026.0626.2———1.8
Set TheoryTarget Model=GPT-5-Mini, Encoding Category=LLM-based2026.0524.5————
AIMVictim model=MMaDA-CoT2026.0624.5———2.55
BaselineTarget Model=Vicuna-13B, Encoding Category=Unencoded2026.0522————
Quantum MechanicsTarget Model=Vicuna-13B, Encoding Category=LLM-based2026.0521.4————
DIJAVictim model=LLaDA2.0-mini2026.0620.7———1.73
DIRECTVictim model=LLaDA2.0-mini2026.0619.1———1.51
BaselineTarget Model=Gemini-3-Flash, Encoding Category=Unencoded2026.0516.4————
Quantum MechanicsTarget Model=GPT-5-Mini, Encoding Category=LLM-based2026.0514.5————
AIMVictim model=LLaDA-Instruct2026.0614———1.9
PAIRVictim model=LLaDA2.0-mini2026.0614———1.3
Quantum MechanicsTarget Model=GPT-5, Encoding Category=LLM-based2026.0513.2————
BaselineTarget Model=GPT-4o-Mini, Encoding Category=Unencoded2026.0512.6————
AIMVictim model=LLaDA 1.52026.0612.5———1.85