List your company

Jailbreak Attack on Harmful Prompts 8-model (test)

92.4ASRSAGA-A4
36.13650.74365.3579.957Jun 3, 2026

Evaluation results

MethodLinks
SAGA-A4Turn type=Multi-turn, Averaging=8-model average2026.0692.4
T1-only (Screenplay)Turn type=Single-turn, Averaging=8-model average2026.0689.4
ActorAttackTurn type=Multi-turn, Averaging=8-model average2026.0679.9
FITDTurn type=Multi-turn, Averaging=8-model average2026.0678.6
CrescendoTurn type=Multi-turn, Averaging=8-model average2026.0638.3