Prompt Injection on URL-based PI (200-sample dataset)
33.5ASRStandard
Evaluation Results
| Method | Links | |
|---|---|---|
| StandardModel=Grok 4 Fast2026.02 | 33.5 | |
| StandardModel=Llama 42026.02 | 22 | |
| StandardModel=GPT-52026.02 | 20 | |
| AdvancedModel=Llama 42026.02 | 16 | |
| InjectDefuserModel=Grok 4 Fast2026.02 | 8 | |
| InjectDefuserModel=Llama 42026.02 | 7 | |
| AdvancedModel=Grok 4 Fast2026.02 | 2 | |
| AdvancedModel=GPT-52026.02 | 1 | |
| StandardModel=Gemma 32026.02 | 0 | |
| AdvancedModel=Gemma 32026.02 | 0 | |
| InjectDefuserModel=GPT-52026.02 | 0 | |
| InjectDefuserModel=Gemma 32026.02 | 0 |