List your company

Content Safety Detection on WildJailbreak (F1, ASR)

96.6F1 Scoreprompt–response verification framework
92.64893.67494.795.726Jun 24, 2026

Evaluation results

MethodLinks
prompt–response verification framework2026.0696.63.4
prompt–response verification framework2026.0696.53.5
AutoDefense / SelfDefendType=Strongest Applicable Baseline2026.0693.6—
AutoDefense / SelfDefendType=Strongest Applicable Baseline2026.0692.8—