Indirect Prompt Injection Attack Success Evaluation on LLM Behavior Goal-Adjacent
100IRanyGPT-5.5
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| GPT-5.52026.05 | 100 | 99 | 95 | 61.1 | 54 | 60.5 | 58 | 54 | |
| GPT-5.42026.05 | 99 | 99 | 94.9 | 44.7 | 42.4 | 43.8 | 42 | 42 | |
| Kimi-K2.62026.05 | 99 | 98 | 94.9 | 73.4 | 77.8 | 71.2 | 69 | 77 | |
| Gemini-3.12026.05 | 97 | 97.9 | 96.9 | 78.7 | 83.5 | 74.7 | 74 | 81 | |
| Sonnet-4.62026.05 | 76 | 98.7 | 88.2 | 79.1 | 72.4 | 59.3 | 53 | 55 |