List your company

Jailbreak Defense on MaliciousInstruct

0ASR (GCG)SafeDecoding
-3.7621.624772.38May 14, 2026

Evaluation results

MethodLinks
SafeDecodingBase Model=Llama3.1-8B-Instruct2026.05010
LEDBase Model=Llama3.1-8B-Instruct2026.05021
EVA (text)Base Model=Llama3.1-8B-Instruct2026.05000
LEDBase Model=Qwen2.5-7B-Instruct2026.05002
EVA (text)Base Model=Qwen2.5-7B-Instruct2026.05004
SafeDecodingBase Model=Vicuna-7B-v1.52026.051811
EVA (text)Base Model=Vicuna-7B-v1.52026.05155
SafeDecodingBase Model=Llama2-7B-chat2026.05100
Circuit BreakersBase Model=Llama2-7B-chat2026.05110
EVA (text)Base Model=Llama2-7B-chat2026.05100
Circuit BreakersBase Model=Llama3.1-8B-Instruct2026.05110
Circuit BreakersBase Model=Qwen2.5-7B-Instruct2026.05110
Circuit BreakersBase Model=Vicuna-7B-v1.52026.052712
LEDBase Model=Vicuna-7B-v1.52026.055105
LoRABase Model=Vicuna-7B-v1.52026.0583216
LEDBase Model=Llama2-7B-chat2026.05821
Circuit BreakersBase Model=Mistral-7B-Instruct-v0.22026.05101910
LEDBase Model=Mistral-7B-Instruct-v0.22026.05122512
SafeDecodingBase Model=Mistral-7B-Instruct-v0.22026.05142416
EVA (text)Base Model=Mistral-7B-Instruct-v0.22026.05141210
LoRABase Model=Llama2-7B-chat2026.053200
LoRABase Model=Mistral-7B-Instruct-v0.22026.05405544
LoRABase Model=Llama3.1-8B-Instruct2026.0542476
VanillaBase Model=Llama2-7B-chat2026.0545300
SafeDecodingBase Model=Qwen2.5-7B-Instruct2026.05489745
VanillaBase Model=Llama3.1-8B-Instruct2026.0569464
LoRABase Model=Qwen2.5-7B-Instruct2026.05729044
VanillaBase Model=Qwen2.5-7B-Instruct2026.05809654
VanillaBase Model=Vicuna-7B-v1.52026.05948386
VanillaBase Model=Mistral-7B-Instruct-v0.22026.05949692