List your company

Safety Alignment on RAB Violence (adversarial prompts)

1ASRAEGIS
-0.9212.042537.96Jul 7, 2026

Evaluation results

MethodLinks
AEGISCategory=AEGIS (Ours), Base Model=SD 2.12026.071
AEGISCategory=AEGIS (Ours), Base Model=SD 1.42026.073
SafeClipCategory=Input-Space2026.074
SLD-StrongCategory=Trigger-Specific2026.077
SAFREECategory=Input-Space2026.078
CADCategory=Feature Pruning2026.078
SafeGuiderCategory=Input-Space2026.0710
UCECategory=Trigger-Specific2026.0711
SDIDCategory=Trigger-Specific2026.0711
RecelerCategory=Trigger-Specific2026.0715
SD 2.1Category=No Alignment2026.0722
SLD-MediumCategory=Trigger-Specific2026.0723
ESD-uCategory=Trigger-Specific2026.0724
Diff-QuickFixCategory=Trigger-Specific2026.0725
SLD-WeakCategory=Trigger-Specific2026.0739
SD 1.4Category=No Alignment2026.0749