Jailbreak Defense on GPTFuzzer
1Harmful ScoreSelf-Examination
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Self-ExaminationModel=Gemma22025.05 | 1 | 2 | |
| SAGEModel=Qwen2.52025.05 | 1 | 0 | |
| SAGEModel=Llama3.12025.05 | 1.16 | 0 | |
| Goal PrioritizationModel=Qwen2.52025.05 | 1.5 | 6 | |
| Self-ReminderModel=Qwen2.52025.05 | 1.58 | 10 | |
| IAModel=Qwen2.52025.05 | 1.78 | 22 | |
| Self-ExaminationModel=Llama3.12025.05 | 1.78 | 26 | |
| Goal PrioritizationModel=Llama3.12025.05 | 1.86 | 26 | |
| Self-ExaminationModel=Qwen2.52025.05 | 2.42 | 40 | |
| SAGEModel=Gemma22025.05 | 2.74 | 18 | |
| No DefenseModel=Qwen2.52025.05 | 2.76 | 24 | |
| ICDModel=Llama3.12025.05 | 2.94 | 20 | |
| Goal PrioritizationModel=Gemma22025.05 | 3.1 | 24 | |
| Self-ReminderModel=Llama3.12025.05 | 3.74 | 94 | |
| IAModel=Llama3.12025.05 | 3.94 | 84 | |
| IAModel=Gemma22025.05 | 4.26 | 36 | |
| No DefenseModel=Gemma22025.05 | 4.28 | 96 | |
| ICDModel=Qwen2.52025.05 | 4.36 | 94 | |
| Self-ReminderModel=Gemma22025.05 | 4.54 | 38 | |
| No DefenseModel=Llama3.12025.05 | 4.6 | 100 | |
| ICDModel=Gemma22025.05 | 4.7 | 48 |