Indirect Prompt Injection Sanitization on IPI Sanitization Total
0.2Attack Success Rate (ASR)RENNERVATE
Evaluation Results
| Method | Links | |
|---|---|---|
| RENNERVATETarget LLM=LLaMA22025.12 | 0.2 | |
| RENNERVATETarget LLM=LLaMA32025.12 | 0.2 | |
| StruQTarget LLM=LLaMA22025.12 | 0.3 | |
| GPT-LocTarget LLM=LLaMA32025.12 | 9.3 | |
| GPT-LocTarget LLM=LLaMA22025.12 | 9.7 | |
| DeepSeek-LocTarget LLM=LLaMA32025.12 | 17.7 | |
| DeepSeek-LocTarget LLM=LLaMA22025.12 | 20.4 | |
| SandwichTarget LLM=LLaMA32025.12 | 29.1 | |
| InstructionalTarget LLM=LLaMA32025.12 | 30.7 | |
| SandwichTarget LLM=LLaMA22025.12 | 33.7 | |
| InstructionalTarget LLM=LLaMA22025.12 | 47.9 | |
| SpotlightingTarget LLM=LLaMA22025.12 | 49.7 | |
| SpotlightingTarget LLM=LLaMA32025.12 | 51.3 | |
| NoneTarget LLM=LLaMA32025.12 | 60.8 | |
| NoneTarget LLM=LLaMA22025.12 | 67.1 |