List your company

Jailbreak Safety Evaluation on English dataset Single-Image

10StrongREJECT (Perturbed)Gemini 1.5 Pro
-0.42.357.7Apr 13, 2025

Evaluation results

MethodLinks
Gemini 1.5 Pro2025.0410130
Gemini 1.5 Flash2025.0483919
Claude 3 Haiku2025.04542
GPT-4o2025.041109
Claude 3.5 Sonnet2025.04000
GPT-4o Mini2025.040110