Jailbreak Defense on JailbreakBench
70Rate of Response SafetyDefaultSystem
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DefaultSystemLLM=Llama-2-7b-chat2025.06 | 70 | 100 | 30 | |
| SystemEmbedderLLM=Llama-2-7b-chat2025.06 | 56.67 | 100 | 43.33 | |
| DefaultSystemLLM=Llama-3.1-8b2025.06 | 30 | 100 | 70 | |
| SystemEmbedderLLM=Llama-3.1-8b2025.06 | 30 | 100 | 70 | |
| LoRA*LLM=Mistral-7B-v0.22025.06 | 23.33 | 100 | 76.67 | |
| SysformerLLM=Phi-3.5-mini2025.06 | 20 | 90 | 70 | |
| SysformerLLM=zephyr-7b-beta2025.06 | 16.67 | 93.33 | 76.66 | |
| LoRA*LLM=Phi-3.5-mini2025.06 | 16.67 | 60 | 43.33 | |
| DefaultSystemLLM=Mistral-7B-v0.22025.06 | 13.33 | 83.33 | 70 | |
| SystemEmbedderLLM=Mistral-7B-v0.22025.06 | 13.33 | 86.67 | 73.34 | |
| LoRA*LLM=Llama-2-7b-chat2025.06 | 10 | 96.67 | 86.67 | |
| LoRA*LLM=Llama-3.1-8b2025.06 | 10 | 96.67 | 86.67 | |
| SysformerLLM=Mistral-7B-v0.22025.06 | 10 | 100 | 90 | |
| DefaultSystemLLM=zephyr-7b-beta2025.06 | 6.67 | 33.33 | 26.66 | |
| SystemEmbedderLLM=zephyr-7b-beta2025.06 | 6.67 | 40 | 33.33 | |
| SysformerLLM=Llama-2-7b-chat2025.06 | 6.67 | 90 | 83.33 | |
| LoRA*LLM=zephyr-7b-beta2025.06 | 3.33 | 86.67 | 73.33 | |
| SysformerLLM=Llama-3.1-8b2025.06 | 3.33 | 96.67 | 93.34 | |
| DefaultSystemLLM=Phi-3.5-mini2025.06 | 3.33 | 10 | 6.67 | |
| SystemEmbedderLLM=Phi-3.5-mini2025.06 | 3.33 | 16.67 | 13.34 |