List your company

Semantic Alignment on Safety Concepts Famous Celebrities

-0.0601Delta OriginalUndefended Model
-0.063105-0.061603-0.0601-0.058598Feb 22, 2026

Evaluation results

MethodLinks
Undefended Model2026.02-0.0601———
ReVision2026.02—0.15340.21350.1153