Jailbreak Defense on JailbreakBench (ASR GCG, AutoDAN, PAIR)
0ASR (GCG)Circuit Breakers
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Circuit BreakersBase Model=Llama2-7B-chat2026.05 | 0 | 0 | 0 | |
| EVA (text)Base Model=Llama2-7B-chat2026.05 | 0 | 0 | 0 | |
| SafeDecodingBase Model=Llama3.1-8B-Instruct2026.05 | 0 | 1 | 1 | |
| Circuit BreakersBase Model=Llama3.1-8B-Instruct2026.05 | 0 | 0 | 0 | |
| LEDBase Model=Qwen2.5-7B-Instruct2026.05 | 0 | 0 | 1 | |
| EVA (text)Base Model=Qwen2.5-7B-Instruct2026.05 | 0 | 1 | 5 | |
| SafeDecodingBase Model=Llama2-7B-chat2026.05 | 1 | 0 | 3 | |
| EVA (text)Base Model=Mistral-7B-Instruct-v0.22026.05 | 2 | 11 | 16 | |
| EVA (text)Base Model=Llama3.1-8B-Instruct2026.05 | 2 | 0 | 0 | |
| SafeDecodingBase Model=Vicuna-7B-v1.52026.05 | 3 | 18 | 15 | |
| Circuit BreakersBase Model=Mistral-7B-Instruct-v0.22026.05 | 3 | 5 | 12 | |
| LEDBase Model=Llama3.1-8B-Instruct2026.05 | 3 | 8 | 0 | |
| Circuit BreakersBase Model=Qwen2.5-7B-Instruct2026.05 | 3 | 0 | 0 | |
| LEDBase Model=Llama2-7B-chat2026.05 | 8 | 2 | 4 | |
| LEDBase Model=Mistral-7B-Instruct-v0.22026.05 | 10 | 12 | 10 | |
| Circuit BreakersBase Model=Vicuna-7B-v1.52026.05 | 15 | 12 | 3 | |
| EVA (text)Base Model=Vicuna-7B-v1.52026.05 | 17 | 8 | 11 | |
| SafeDecodingBase Model=Mistral-7B-Instruct-v0.22026.05 | 18 | 14 | 22 | |
| LoRABase Model=Vicuna-7B-v1.52026.05 | 32 | 22 | 20 | |
| LEDBase Model=Vicuna-7B-v1.52026.05 | 34 | 8 | 6 | |
| LoRABase Model=Qwen2.5-7B-Instruct2026.05 | 36 | 42 | 50 | |
| LoRABase Model=Llama3.1-8B-Instruct2026.05 | 37 | 35 | 9 | |
| LoRABase Model=Mistral-7B-Instruct-v0.22026.05 | 38 | 62 | 34 | |
| VanillaBase Model=Qwen2.5-7B-Instruct2026.05 | 42 | 80 | 55 | |
| SafeDecodingBase Model=Qwen2.5-7B-Instruct2026.05 | 42 | 73 | 43 | |
| VanillaBase Model=Llama2-7B-chat2026.05 | 46 | 27 | 4 | |
| VanillaBase Model=Llama3.1-8B-Instruct2026.05 | 46 | 31 | 11 | |
| LoRABase Model=Llama2-7B-chat2026.05 | 50 | 1 | 2 | |
| VanillaBase Model=Mistral-7B-Instruct-v0.22026.05 | 79 | 95 | 89 | |
| VanillaBase Model=Vicuna-7B-v1.52026.05 | 89 | 73 | 77 |