List your company

Jailbreak Attack Transferability on Llama-2-7b-chat finetuned variants v1 (test)

60.4Transfer Success Rate (TSR)PGP
-2.41613.89230.246.508Dec 14, 2025

Evaluation results

MethodLinks
PGPAttacker Knowledge=fully black-box2025.1260.4
GCGAttacker Knowledge=white-box2025.1247.4
GCG EnsembleAttacker Knowledge=fully black-box2025.1240.2
Guiding-GCGAttacker Knowledge=pretrain–to–finetune setting2025.1228.8
SCAVAttacker Knowledge=pretrain–to–finetune setting2025.1228
TUJAAttacker Knowledge=pretrain–to–finetune setting2025.1224.8
GCGAttacker Knowledge=pretrain–to–finetune setting2025.1224.4
SEAAttacker Knowledge=pretrain–to–finetune setting2025.1220.4
LSGM_LILAAttacker Knowledge=pretrain–to–finetune setting2025.1216
AutoDanAttacker Knowledge=white-box2025.1215.8
AutoDanAttacker Knowledge=pretrain–to–finetune setting2025.126.2
DI-GCGAttacker Knowledge=pretrain–to–finetune setting2025.123.2
L4AAttacker Knowledge=pretrain–to–finetune setting2025.121.8
PIFAttacker Knowledge=pretrain–to–finetune setting2025.120
DANAttacker Knowledge=fully black-box2025.120
PAIRAttacker Knowledge=fully black-box2025.120