Jailbreak Attack on TrustLLM
95.4RtAGPTFuzzer
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPTFuzzerVictim LLM=GPT-4o2025.07 | 95.4 | 43.1 | 1.34 | |
| GCGVictim LLM=GPT-4o2025.07 | 93.4 | 27.4 | 1.37 | |
| GPTFuzzerVictim LLM=GPT-4o-mini2025.07 | 92.8 | 47.9 | 1.59 | |
| Original PromptVictim LLM=GPT-4o2025.07 | 90.9 | 25.7 | 1.38 | |
| DAPVictim LLM=GPT-4o2025.07 | 90.8 | 25.3 | 1.3 | |
| DAPVictim LLM=GPT-4o-mini2025.07 | 88.1 | 23.5 | 1.4 | |
| Virtual ContextVictim LLM=GPT-4o2025.07 | 87.7 | 30.1 | 1.57 | |
| GCGVictim LLM=GPT-4o-mini2025.07 | 86.3 | 27.2 | 1.56 | |
| Original PromptVictim LLM=GPT-4o-mini2025.07 | 84.8 | 22.5 | 1.57 | |
| Virtual ContextVictim LLM=GPT-4o-mini2025.07 | 81.4 | 34.5 | 1.9 | |
| Chat-NNVictim LLM=GPT-4o2025.07 | 78.7 | 49.5 | 2.01 | |
| TranslitVictim LLM=GPT-4o2025.07 | 74.7 | 51.1 | 2.11 | |
| PAPVictim LLM=GPT-4o2025.07 | 73.6 | 45.7 | 2.65 | |
| PAPVictim LLM=GPT-4o-mini2025.07 | 70.9 | 44.1 | 2.61 | |
| TranslitVictim LLM=GPT-4o-mini2025.07 | 66.4 | 48.4 | 2.23 | |
| Chat-NNVictim LLM=GPT-4o-mini2025.07 | 64.3 | 47.4 | 2.25 | |
| Ours + Chat-NNVictim LLM=GPT-4o2025.07 | 44.5 | 63.2 | 2.55 | |
| Ours + TranslitVictim LLM=GPT-4o2025.07 | 41.3 | 68.6 | 2.59 | |
| Ours + GPTFuzzerVictim LLM=GPT-4o-mini2025.07 | 6.9 | 63.1 | 2.28 | |
| Ours + GPTFuzzerVictim LLM=GPT-4o2025.07 | 5.4 | 40.7 | 1.94 | |
| OursVictim LLM=GPT-4o-mini2025.07 | 3.4 | 28.8 | 1.85 | |
| Ours + Chat-NNVictim LLM=GPT-4o-mini2025.07 | 2.9 | 52.9 | 2.33 | |
| Ours + TranslitVictim LLM=GPT-4o-mini2025.07 | 2.6 | 55.8 | 2.24 | |
| OursVictim LLM=GPT-4o2025.07 | 2.2 | 33.5 | 2.04 | |
| Ours + PAPVictim LLM=GPT-4o-mini2025.07 | 0.7 | 53.3 | 2.7 | |
| Ours + PAPVictim LLM=GPT-4o2025.07 | 0.4 | 55.7 | 2.75 |