# Adversarial robustness (unsafe items under text attacks) n=160 unsafe items, threshold 0.5, detection rate = fraction still flagged unsafe. | Attack | Detection rate | Mean unsafe prob | | --- | --- | --- | | clean | 98.8% | 0.847 | | typo | 96.2% | 0.799 | | leet | 99.4% | 0.846 | | homoglyph | 99.4% | 0.865 | | zero_width | 100.0% | 0.880 | | case | 97.5% | 0.795 |