List your company
ResearchFormerly WizWand

Jailbreak Defense on JailbreakBench (ASR GCG, AutoDAN, PAIR)

0ASR (GCG)Circuit Breakers
-3.5620.4744.568.53May 14, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Circuit BreakersBase Model=Llama2-7B-chat2026.05000
EVA (text)Base Model=Llama2-7B-chat2026.05000
SafeDecodingBase Model=Llama3.1-8B-Instruct2026.05011
Circuit BreakersBase Model=Llama3.1-8B-Instruct2026.05000
LEDBase Model=Qwen2.5-7B-Instruct2026.05001
EVA (text)Base Model=Qwen2.5-7B-Instruct2026.05015
SafeDecodingBase Model=Llama2-7B-chat2026.05103
EVA (text)Base Model=Mistral-7B-Instruct-v0.22026.0521116
EVA (text)Base Model=Llama3.1-8B-Instruct2026.05200
SafeDecodingBase Model=Vicuna-7B-v1.52026.0531815
Circuit BreakersBase Model=Mistral-7B-Instruct-v0.22026.053512
LEDBase Model=Llama3.1-8B-Instruct2026.05380
Circuit BreakersBase Model=Qwen2.5-7B-Instruct2026.05300
LEDBase Model=Llama2-7B-chat2026.05824
LEDBase Model=Mistral-7B-Instruct-v0.22026.05101210
Circuit BreakersBase Model=Vicuna-7B-v1.52026.0515123
EVA (text)Base Model=Vicuna-7B-v1.52026.0517811
SafeDecodingBase Model=Mistral-7B-Instruct-v0.22026.05181422
LoRABase Model=Vicuna-7B-v1.52026.05322220
LEDBase Model=Vicuna-7B-v1.52026.053486
LoRABase Model=Qwen2.5-7B-Instruct2026.05364250
LoRABase Model=Llama3.1-8B-Instruct2026.0537359
LoRABase Model=Mistral-7B-Instruct-v0.22026.05386234
VanillaBase Model=Qwen2.5-7B-Instruct2026.05428055
SafeDecodingBase Model=Qwen2.5-7B-Instruct2026.05427343
VanillaBase Model=Llama2-7B-chat2026.0546274
VanillaBase Model=Llama3.1-8B-Instruct2026.05463111
LoRABase Model=Llama2-7B-chat2026.055012
VanillaBase Model=Mistral-7B-Instruct-v0.22026.05799589
VanillaBase Model=Vicuna-7B-v1.52026.05897377