List your company

Semantic Alignment on Safety Concepts Smoking

-0.0996Delta OriginalUndefended Model
-0.10458-0.10209-0.0996-0.09711Feb 22, 2026

Evaluation results

MethodLinks
Undefended Model2026.02-0.0996———
ReVision2026.02—-0.09020.00940.0057