List your company
ResearchFormerly WizWand

Safety Behavior Prediction on Challenging Prompts

0Unsafe Rate (%)Challenging Prompts
-1.1766.76214.722.638Jul 8, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Challenging PromptsDeployment=GPT-5 → GPT-5.1, Category=Extremism2026.070——
Challenging PromptsDeployment=GPT-5.1 → GPT-5.2, Category=Extremism2026.070——
Challenging PromptsDeployment=GPT-5.2 → GPT-5.4, Category=Extremism2026.070——
Challenging PromptsDeployment=GPT-5 → GPT-5.1, Category=Sexual minors2026.070.9——
Challenging PromptsDeployment=GPT-5.2 → GPT-5.4, Category=Sexual minors2026.070.9——
Challenging PromptsDeployment=GPT-5 → GPT-5.1, Category=Violent wrongdoing2026.071.8——
Challenging PromptsDeployment=GPT-5.2 → GPT-5.4, Category=Hate2026.072.1——
Challenging PromptsDeployment=GPT-5.2 → GPT-5.4, Category=Violent wrongdoing2026.072.1——
Challenging PromptsDeployment=GPT-5.2 → GPT-5.4, Category=Sexual content2026.073.9——
Challenging PromptsDeployment=GPT-5.1 → GPT-5.2, Category=Violent wrongdoing2026.074.1——
Challenging PromptsDeployment=GPT-5.2 → GPT-5.4, Category=Self-harm2026.074.7——
Challenging PromptsDeployment=GPT-5 → GPT-5.1, Category=Self-harm2026.075.1——
Challenging PromptsDeployment=GPT-5 → GPT-5.1, Category=Sexual content2026.075.4——
Challenging PromptsDeployment=GPT-5.1 → GPT-5.2, Category=Sexual content2026.076.6——
Challenging PromptsDeployment=GPT-5.1 → GPT-5.2, Category=Self-harm2026.077.2——
Challenging PromptsDeployment=GPT-5.2 → GPT-5.4, Category=Non-violent wrongdoing2026.077.7——
Challenging PromptsDeployment=GPT-5.1 → GPT-5.2, Category=Sexual minors2026.078.7——
Challenging PromptsDeployment=GPT-5 → GPT-5.1, Category=Violence2026.079.1——
Challenging PromptsDeployment=GPT-5.2 → GPT-5.4, Category=Violence2026.079.1——
Challenging PromptsDeployment=GPT-5 → GPT-5.1, Category=Non-violent wrongdoing2026.0711.4——
Challenging PromptsDeployment=GPT-5 → GPT-5.1, Category=Hate2026.0711.7——
Challenging PromptsDeployment=GPT-5.1 → GPT-5.2, Category=Violence2026.0714.5——
Challenging PromptsDeployment=GPT-5.1 → GPT-5.2, Category=Hate2026.0715.9——
Challenging PromptsDeployment=GPT-5.1 → GPT-5.2, Category=Non-violent wrongdoing2026.0716.3——
Challenging PromptsDeployment=GPT-5.2 → GPT-5.4, Category=Harassment2026.0719——
Challenging PromptsDeployment=GPT-5 → GPT-5.1, Category=Harassment2026.0724.1——
Challenging PromptsDeployment=GPT-5.1 → GPT-5.2, Category=Harassment2026.0729.4——