# 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 | 99.4% | 0.917 | | typo | 94.4% | 0.858 | | leet | 97.5% | 0.906 | | homoglyph | 100.0% | 0.947 | | zero_width | 100.0% | 0.939 | | case | 95.6% | 0.833 |