List your company

Safety Evaluation on Jailbreaking Scenarios (Jailbreak)

2.3Unsafe RatioDeepSeek-V3
-1.04421.52844.166.672Jan 22, 2025

Evaluation results

MethodLinks
DeepSeek-V3risk control system=true2025.012.346.5—
DeepSeek-R1risk control system=true2025.014.387.3—
o1release date=2024-12-172025.0112.179.8—
Claude-3.7-Sonnet2025.0126.221.9—
Qwen2.5 Instructparameters=72B2025.0129.725.2—
GPT-4orelease date=2024-05-132025.0130.457.3—
DeepSeek-V3risk control system=false2025.0136.48.9—
DeepSeek-R1risk control system=false2025.0185.91.9—