List your company

Jailbreak Attack on Opposite-day (test)

10.5ASRGPT-5
9.6215.5621.527.44Feb 28, 2025

Evaluation results

MethodLinks
GPT-5Safety Steering=Enabled2025.0210.5
Claude Sonnet 4.5Safety Steering=Enabled2025.0213.5
Claude Sonnet 4.5Safety Steering=Disabled2025.0225
GPT-5Safety Steering=Disabled2025.0232.5