Jailbreak Attack on Malicious Prompts English (test)
100ASR@5Caesar
Evaluation Results
| Method | Links | |
|---|---|---|
| CaesarTarget LLM=LLaMA, Judge Model=Rule_Judge2025.05 | 100 | |
| CaesarTarget LLM=DS-V3, Judge Model=Rule_Judge2025.05 | 100 | |
| CaesarTarget LLM=GPT-4o, Judge Model=Rule_Judge2025.05 | 100 | |
| LogiBreakTarget LLM=DS-V3, Judge Model=Rule_Judge2025.05 | 88 | |
| LogiBreakTarget LLM=GPT-4o, Judge Model=LLaMA_Judge2025.05 | 85 | |
| LogiBreakTarget LLM=DS-V3, Judge Model=LLaMA_Judge2025.05 | 84 | |
| LogiBreakTarget LLM=LLaMA, Judge Model=Rule_Judge2025.05 | 82 | |
| LogiBreakTarget LLM=GPT-4o, Judge Model=Rule_Judge2025.05 | 82 | |
| LogiBreakTarget LLM=DS-V3, Judge Model=GPT_Judge2025.05 | 75 | |
| RefCompTarget LLM=DS-V3, Judge Model=Rule_Judge2025.05 | 74 | |
| AutoDANTarget LLM=DS-V3, Judge Model=Rule_Judge2025.05 | 73 | |
| AutoDANTarget LLM=GPT-4o, Judge Model=Rule_Judge2025.05 | 72 | |
| LogiBreakTarget LLM=LLaMA, Judge Model=LLaMA_Judge2025.05 | 70 | |
| RefCompTarget LLM=DS-V3, Judge Model=GPT_Judge2025.05 | 70 | |
| RefCompTarget LLM=DS-V3, Judge Model=LLaMA_Judge2025.05 | 67 | |
| AIMTarget LLM=DS-V3, Judge Model=Rule_Judge2025.05 | 66 | |
| LogiBreakTarget LLM=GPT-4o, Judge Model=GPT_Judge2025.05 | 64 | |
| AutoDANTarget LLM=DS-V3, Judge Model=LLaMA_Judge2025.05 | 62 | |
| AutoDANTarget LLM=DS-V3, Judge Model=GPT_Judge2025.05 | 62 | |
| AIMTarget LLM=DS-V3, Judge Model=LLaMA_Judge2025.05 | 61 | |
| RefCompTarget LLM=GPT-4o, Judge Model=Rule_Judge2025.05 | 60 | |
| AutoDANTarget LLM=GPT-4o, Judge Model=LLaMA_Judge2025.05 | 59 | |
| AIMTarget LLM=DS-V3, Judge Model=GPT_Judge2025.05 | 59 | |
| AutoDANTarget LLM=LLaMA, Judge Model=Rule_Judge2025.05 | 48 | |
| LogiBreakTarget LLM=LLaMA, Judge Model=GPT_Judge2025.05 | 48 | |
| CaesarTarget LLM=DS-V3, Judge Model=LLaMA_Judge2025.05 | 45 | |
| PrefixTarget LLM=DS-V3, Judge Model=Rule_Judge2025.05 | 40 | |
| PastTenseTarget LLM=GPT-4o, Judge Model=Rule_Judge2025.05 | 38 | |
| AutoDANTarget LLM=GPT-4o, Judge Model=GPT_Judge2025.05 | 37 | |
| RefCompTarget LLM=LLaMA, Judge Model=Rule_Judge2025.05 | 36 | |
| PrefixTarget LLM=GPT-4o, Judge Model=Rule_Judge2025.05 | 35 | |
| LRL-CombTarget LLM=LLaMA, Judge Model=Rule_Judge2025.05 | 34 | |
| LRL-CombTarget LLM=DS-V3, Judge Model=Rule_Judge2025.05 | 31 | |
| CaesarTarget LLM=DS-V3, Judge Model=GPT_Judge2025.05 | 31 | |
| CaesarTarget LLM=GPT-4o, Judge Model=LLaMA_Judge2025.05 | 29 | |
| AutoDANTarget LLM=LLaMA, Judge Model=LLaMA_Judge2025.05 | 29 | |
| CaesarTarget LLM=GPT-4o, Judge Model=GPT_Judge2025.05 | 29 | |
| AutoDANTarget LLM=LLaMA, Judge Model=GPT_Judge2025.05 | 29 | |
| PrefixTarget LLM=LLaMA, Judge Model=Rule_Judge2025.05 | 28 | |
| PastTenseTarget LLM=DS-V3, Judge Model=Rule_Judge2025.05 | 27 | |
| LRL-CombTarget LLM=GPT-4o, Judge Model=Rule_Judge2025.05 | 27 | |
| RefCompTarget LLM=LLaMA, Judge Model=GPT_Judge2025.05 | 27 | |
| RefCompTarget LLM=LLaMA, Judge Model=LLaMA_Judge2025.05 | 26 | |
| LRL-CombTarget LLM=DS-V3, Judge Model=LLaMA_Judge2025.05 | 21 | |
| PrefixTarget LLM=DS-V3, Judge Model=GPT_Judge2025.05 | 21 | |
| LRL-CombTarget LLM=DS-V3, Judge Model=GPT_Judge2025.05 | 17 | |
| PrefixTarget LLM=DS-V3, Judge Model=LLaMA_Judge2025.05 | 15 | |
| LRL-CombTarget LLM=GPT-4o, Judge Model=LLaMA_Judge2025.05 | 14 | |
| RefCompTarget LLM=GPT-4o, Judge Model=GPT_Judge2025.05 | 14 | |
| LRL-CombTarget LLM=GPT-4o, Judge Model=GPT_Judge2025.05 | 12 | |
| RefCompTarget LLM=GPT-4o, Judge Model=LLaMA_Judge2025.05 | 11 | |
| LRL-CombTarget LLM=LLaMA, Judge Model=GPT_Judge2025.05 | 11 | |
| PastTenseTarget LLM=DS-V3, Judge Model=GPT_Judge2025.05 | 7 | |
| PrefixTarget LLM=GPT-4o, Judge Model=GPT_Judge2025.05 | 7 | |
| PastTenseTarget LLM=GPT-4o, Judge Model=GPT_Judge2025.05 | 6 | |
| CaesarTarget LLM=LLaMA, Judge Model=LLaMA_Judge2025.05 | 5 | |
| LRL-CombTarget LLM=LLaMA, Judge Model=LLaMA_Judge2025.05 | 5 | |
| PastTenseTarget LLM=DS-V3, Judge Model=LLaMA_Judge2025.05 | 4 | |
| PastTenseTarget LLM=GPT-4o, Judge Model=LLaMA_Judge2025.05 | 4 | |
| PrefixTarget LLM=GPT-4o, Judge Model=LLaMA_Judge2025.05 | 4 | |
| AIMTarget LLM=GPT-4o, Judge Model=Rule_Judge2025.05 | 3 | |
| CaesarTarget LLM=LLaMA, Judge Model=GPT_Judge2025.05 | 2 | |
| PastTenseTarget LLM=LLaMA, Judge Model=Rule_Judge2025.05 | 1 | |
| PastTenseTarget LLM=LLaMA, Judge Model=LLaMA_Judge2025.05 | 1 | |
| PrefixTarget LLM=LLaMA, Judge Model=GPT_Judge2025.05 | 1 | |
| AIMTarget LLM=LLaMA, Judge Model=Rule_Judge2025.05 | 0 | |
| AIMTarget LLM=LLaMA, Judge Model=LLaMA_Judge2025.05 | 0 | |
| AIMTarget LLM=GPT-4o, Judge Model=LLaMA_Judge2025.05 | 0 | |
| PrefixTarget LLM=LLaMA, Judge Model=LLaMA_Judge2025.05 | 0 | |
| PastTenseTarget LLM=LLaMA, Judge Model=GPT_Judge2025.05 | 0 | |
| AIMTarget LLM=LLaMA, Judge Model=GPT_Judge2025.05 | 0 | |
| AIMTarget LLM=GPT-4o, Judge Model=GPT_Judge2025.05 | 0 |