Jailbreak Attack on Real-world LLM applications OpenAI
75ASRPAP
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PAPModel=GPT-4.1, MetaBreak Integration=true2025.10 | 75 | 7.5 | |
| PAPModel=GPT-4.1, MetaBreak Integration=false2025.10 | 67.5 | — | |
| MetaBreakModel=GPT-4.12025.10 | 55.5 | — | |
| GPTFuzzerModel=GPT-4.1, MetaBreak Integration=true2025.10 | 35.9 | 8.9 | |
| GPTFuzzerModel=GPT-4.1, MetaBreak Integration=false2025.10 | 27 | — | |
| DIModel=GPT-4.12025.10 | 23.4 | — |