Content Safety Detection on InjecAgent
92.8F1 Scoreprompt–response verification framework
Evaluation results
| Method | Links | ||
|---|---|---|---|
| prompt–response verification framework2026.06 | 92.8 | 6.7 | |
| PromptGuard-2 / InjecGuardType=Strongest Applicable Baseline2026.06 | 83.1 | — |