Prompt Injection Attack Robustness on SAP10 2023a (test)
61Attack Success Percentage (ASP)Hypnotism
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HypnotismAttack Method=Hypnotism, Model Group=Open-source LLMs (Mean)2025.05 | 61 | 57.8 | |
| Ignore PrefixAttack Method=Ignore Prefix, Model Group=Open-source LLMs (Mean)2025.05 | 60.8 | 58.3 | |
| Role-playing CoTAttack Method=Role-playing CoT, Model Group=Open-source LLMs (Mean)2025.05 | 55.2 | 54.3 |