List your company

Jailbreak Defense on CodeAttack

1Harmful ScoreSAGE
0.8521.8512.853.849May 17, 2025

Evaluation results

MethodLinks
SAGEModel=Gemma22025.0510
Goal PrioritizationModel=Gemma22025.051.040
ICDModel=Qwen2.52025.051.180
SAGEModel=Llama3.12025.051.262
Goal PrioritizationModel=Qwen2.52025.051.4212
SAGEModel=Qwen2.52025.051.442
Self-ExaminationModel=Qwen2.52025.051.9826
Self-ReminderModel=Qwen2.52025.053.32100
Self-ExaminationModel=Gemma22025.053.3686
IAModel=Gemma22025.053.4696
Self-ReminderModel=Gemma22025.053.692
ICDModel=Gemma22025.053.6486
IAModel=Qwen2.52025.054.498
No DefenseModel=Qwen2.52025.054.7100