List your company

Jailbreak Attack Robustness on JBB-Behaviors

10ASR (PAIR)RA
6.4830.245477.76Oct 1, 2025

Evaluation results

MethodLinks
RABase Model=LLaDA2025.101072.345
RABase Model=LLaDA1.52025.101671.747
MOSABase Model=LLaDA1.52025.102578.370
RA w/o interBase Model=LLaDA, Ablation=without intervention2025.1026.375.771.3
MOSABase Model=LLaDA2025.1027.377.766.3
DPOBase Model=LLaDA2025.1031.388.374
DPOBase Model=LLaDA1.52025.103690.774.3
SFTBase Model=LLaDA2025.1036.794.371
RA w/o interBase Model=LLaDA1.5, Ablation=without intervention2025.1038.779.368.5
SFTBase Model=LLaDA1.52025.103991.370.7
OriginalBase Model=LLaDA2025.1044.392.781.3
OriginalBase Model=LLaDA1.52025.1045.396.781.7
RABase Model=MMaDA2025.1046.381.755.3
RA w/o interBase Model=MMaDA, Ablation=without intervention2025.1054.377.672
MOSABase Model=MMaDA2025.105975.771
DPOBase Model=MMaDA2025.1067.582.385.7
SFTBase Model=MMaDA2025.10929593.5
OriginalBase Model=MMaDA2025.109879.393