List your company

Jailbreak Defense on ReNeLLM, DeepInception, GPTFuzzer, CodeAttack (test)

1Harmful Score (ReNeLLM)SAGE
0.85281.84642.843.8336May 17, 2025

Evaluation results

MethodLinks
SAGEModel=deepseek R12025.051010101010
Goal PrioritizationModel=deepseek R12025.051.0421401.1612101.0514
IAModel=deepseek R12025.051.5221181.4101.08181.2517
Self-ReminderModel=deepseek R12025.052.08861.34982.14321561.6468
ICDModel=deepseek R12025.053.02562.82803.02301.34202.5546
No DefenseModel=deepseek R12025.054.68964.88983.66424.481004.4284