List your company

Jailbreak Defense on BeaverTails

69Reference PerformanceCHASE Defender
5.5622.0338.554.97Jun 4, 2026

Evaluation results

MethodLinks
CHASE DefenderAttack=Translation*2026.06692.717.90.01213.2
CHASE DefenderAttack=AutoDAN2026.06573.7730.40.31450.8
Base ModelAttack=Translation*2026.06542.29.1——
CHASE DefenderAttack=TAP2026.06512.99310.26546.1
CHASE DefenderAttack=PAP2026.06472.7524.10.23148.9
Base ModelAttack=TAP2026.06304.0457.5——
Base ModelAttack=PAP2026.06263.1647.2——
Base ModelAttack=AutoDAN2026.06213.761.8——
CHASE DefenderAttack=PAIR2026.06152.9845.10.17828.3
Base ModelAttack=PAIR2026.0683.3962.9——