Jailbreak Detection on DrAttack
99AccuracyGradSafe
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GradSafeTarget LLM=Llama2-7B2025.12 | 99 | 99 | |
| IMAGTarget LLM=Vicuna-13B2025.12 | 98 | 99 | |
| PPLTarget LLM=Llama3-8B2025.12 | 97 | 98 | |
| IMAGTarget LLM=Llama3-8B2025.12 | 96 | 98 | |
| IMAGTarget LLM=Mistral-7B2025.12 | 86 | 92 | |
| LlamaGTarget LLM=Vicuna-13B2025.12 | 85 | 92 | |
| LlamaGTarget LLM=Mistral-7B2025.12 | 84 | 91 | |
| LlamaGTarget LLM=Vicuna-7B2025.12 | 84 | 91 | |
| IMAGTarget LLM=Vicuna-7B2025.12 | 84 | 91 | |
| IMAGTarget LLM=Llama2-7B2025.12 | 65 | 79 | |
| LlamaGTarget LLM=Llama2-7B2025.12 | 57 | 72 | |
| LlamaGTarget LLM=Llama3-8B2025.12 | 55 | 71 | |
| Self-ExTarget LLM=Mistral-7B2025.12 | 51 | 67 | |
| Self-ExTarget LLM=Llama2-7B2025.12 | 24 | 27 | |
| Self-ExTarget LLM=Llama3-8B2025.12 | 18 | 30 | |
| OAPITarget LLM=Mistral-7B2025.12 | 4 | 7 | |
| OAPITarget LLM=Vicuna-7B2025.12 | 4 | 7 | |
| OAPITarget LLM=Vicuna-13B2025.12 | 4 | 8 | |
| Self-ExTarget LLM=Vicuna-7B2025.12 | 3 | 6 | |
| OAPITarget LLM=Llama2-7B2025.12 | 3 | 7 | |
| OAPITarget LLM=Llama3-8B2025.12 | 3 | 6 | |
| PPLTarget LLM=Mistral-7B2025.12 | 0 | 0 | |
| GradSafeTarget LLM=Mistral-7B2025.12 | 0 | 0 | |
| PPLTarget LLM=Vicuna-7B2025.12 | 0 | 0 | |
| GradSafeTarget LLM=Vicuna-7B2025.12 | 0 | 0 | |
| PPLTarget LLM=Vicuna-13B2025.12 | 0 | 0 | |
| Self-ExTarget LLM=Vicuna-13B2025.12 | 0 | 0 | |
| GradSafeTarget LLM=Vicuna-13B2025.12 | 0 | 0 | |
| PPLTarget LLM=Llama2-7B2025.12 | 0 | 0 | |
| GradSafeTarget LLM=Llama3-8B2025.12 | 0 | 0 |