List your company
ResearchFormerly WizWand

Jailbreak Attack on Llama2-7b five finetuned variants

0Average ASRPIF
-2.50414.39831.348.202Dec 14, 2025
Updated 8mo ago

Evaluation results

MethodLinks
PIFAttacker knowledge (Threat Model)=Pretrain-to-finetune2025.120
DANAttacker knowledge (Threat Model)=Fully black-box2025.120
PAIRAttacker knowledge (Threat Model)=Fully black-box2025.120
L4AAttacker knowledge (Threat Model)=Pretrain-to-finetune2025.121.8
DI-GCGAttacker knowledge (Threat Model)=Pretrain-to-finetune2025.123.2
AutoDan (adaptation)Attacker knowledge (Threat Model)=Pretrain-to-finetune2025.126.2
AutoDan (white)Attacker knowledge (Threat Model)=White-box2025.1215.8
LSGM_LILA (adaptation)Attacker knowledge (Threat Model)=Pretrain-to-finetune2025.1216
SEAAttacker knowledge (Threat Model)=Pretrain-to-finetune2025.1220.4
TUJA (adaptation)Attacker knowledge (Threat Model)=Pretrain-to-finetune2025.1224.8
GCG (adaptation)Attacker knowledge (Threat Model)=Pretrain-to-finetune2025.1225.6
SCAV (adaptation)Attacker knowledge (Threat Model)=Pretrain-to-finetune2025.1228
Guiding-GCGAttacker knowledge (Threat Model)=Pretrain-to-finetune2025.1233.2
GCG EnsembleAttacker knowledge (Threat Model)=Fully black-box2025.1240.2
GCG (white)Attacker knowledge (Threat Model)=White-box2025.1247.4
PGPAttacker knowledge (Threat Model)=Fully black-box2025.1262.6