Jailbreak Detection on AutoDAN
99AccuracyIMAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| IMAGTarget LLM=Vicuna-13B2025.12 | 99 | 99 | |
| IMAGTarget LLM=Llama2-7B2025.12 | 98 | 99 | |
| GradSafeTarget LLM=Llama2-7B2025.12 | 96 | 98 | |
| IMAGTarget LLM=Llama3-8B2025.12 | 96 | 98 | |
| IMAGTarget LLM=Mistral-7B2025.12 | 95 | 97 | |
| IMAGTarget LLM=Vicuna-7B2025.12 | 94 | 97 | |
| LlamaGTarget LLM=Mistral-7B2025.12 | 77 | 87 | |
| LlamaGTarget LLM=Vicuna-13B2025.12 | 75 | 76 | |
| LlamaGTarget LLM=Vicuna-7B2025.12 | 72 | 83 | |
| LlamaGTarget LLM=Llama3-8B2025.12 | 60 | 75 | |
| Self-ExTarget LLM=Mistral-7B2025.12 | 56 | 72 | |
| LlamaGTarget LLM=Llama2-7B2025.12 | 38 | 55 | |
| GradSafeTarget LLM=Llama3-8B2025.12 | 37 | 54 | |
| Self-ExTarget LLM=Llama2-7B2025.12 | 27 | 31 | |
| PPLTarget LLM=Llama3-8B2025.12 | 23 | 36 | |
| Self-ExTarget LLM=Llama3-8B2025.12 | 16 | 26 | |
| OAPITarget LLM=Llama2-7B2025.12 | 9 | 16 | |
| OAPITarget LLM=Llama3-8B2025.12 | 8 | 14 | |
| OAPITarget LLM=Mistral-7B2025.12 | 5 | 10 | |
| GradSafeTarget LLM=Mistral-7B2025.12 | 5 | 10 | |
| OAPITarget LLM=Vicuna-13B2025.12 | 5 | 9 | |
| OAPITarget LLM=Vicuna-7B2025.12 | 4 | 9 | |
| PPLTarget LLM=Vicuna-13B2025.12 | 1 | 2 | |
| PPLTarget LLM=Mistral-7B2025.12 | 0 | 0 | |
| PPLTarget LLM=Vicuna-7B2025.12 | 0 | 0 | |
| Self-ExTarget LLM=Vicuna-7B2025.12 | 0 | 0 | |
| GradSafeTarget LLM=Vicuna-7B2025.12 | 0 | 0 | |
| Self-ExTarget LLM=Vicuna-13B2025.12 | 0 | 0 | |
| GradSafeTarget LLM=Vicuna-13B2025.12 | 0 | 0 | |
| PPLTarget LLM=Llama2-7B2025.12 | 0 | 0 |