List your company

Jailbreak Attack on AdvBench GPT-4o

6ASRGCG (transfer)
2.3227.165276.84Dec 3, 2025

Evaluation results

MethodLinks
GCG (transfer)Attack setting=Transferred2025.126——
DoublespeakAttack setting=Optimization-free, Zero-shot=true2025.1231——
Adaptive (transfer)Attack setting=Transferred2025.1242——
PAIRAttack setting=Optimization-based2025.1262——
TAPAttack setting=Optimization-based2025.1288——
Adv. ReasoningAttack setting=Optimization-based2025.1294——
FlipAttackAttack setting=Optimization-free2025.1298——
ADVPROMPTERTarget LLM=GPT-4O, Setting=Black-box2025.10—27.868.9
ATOXIATarget LLM=GPT-4O, Setting=Black-box2025.10—48.282.4
RED-BANDIT (UCB)Target LLM=GPT-4O, Setting=Black-box, Policy=UCB2025.10—29.893.3
RED-BANDIT (ϵ-GREEDY)Target LLM=GPT-4O, Setting=Black-box, Policy=ϵ-greedy2025.10—36.685.6