List your company

Safety Classification Disagreement on WildChat Intent

0.4Disagreement Rate (per 1k conv)Gemini 3.1
-1.29.620.431.2May 22, 2026

Evaluation results

MethodLinks
Gemini 3.1Category=Hate Speech, Evaluation Protocol=Constitution2026.050.4
GPT-5.4Category=Hate Speech, Evaluation Protocol=Constitution2026.050.4
GPT-5.4 MiniCategory=Harassment, Evaluation Protocol=Constitution2026.050.4
GPT-5.4 MiniCategory=Hate Speech, Evaluation Protocol=Constitution2026.050.6
Gemini 3.1Category=Harassment, Evaluation Protocol=Constitution2026.050.7
Safeguard 20BCategory=Hate Speech, Evaluation Protocol=Constitution2026.050.9
GPT-5.4Category=Harassment, Evaluation Protocol=Constitution2026.051
Gemini 3.1Category=Non-Violent Crime, Evaluation Protocol=Constitution2026.051.4
Gemini 3.1Category=Hate Speech, Evaluation Protocol=Definition2026.051.5
GPT-5.4Category=Non-Violent Crime, Evaluation Protocol=Constitution2026.051.6
Safeguard 20BCategory=Non-Violent Crime, Evaluation Protocol=Constitution2026.051.8
Safeguard 20BCategory=Hate Speech, Evaluation Protocol=Definition2026.051.9
GPT-5.4 NanoCategory=Hate Speech, Evaluation Protocol=Constitution2026.052.2
GPT-5.4Category=Hate Speech, Evaluation Protocol=Definition2026.052.7
GPT-5.4 MiniCategory=Non-Violent Crime, Evaluation Protocol=Constitution2026.052.8
Safeguard 20BCategory=Harassment, Evaluation Protocol=Constitution2026.052.8
GPT-5.4 MiniCategory=Hate Speech, Evaluation Protocol=Definition2026.052.9
GPT-5.4 NanoCategory=Non-Violent Crime, Evaluation Protocol=Constitution2026.055.1
GPT-5.4 NanoCategory=Hate Speech, Evaluation Protocol=Definition2026.055.6
GPT-5.4 NanoCategory=Harassment, Evaluation Protocol=Constitution2026.055.9
Gemini 3.1Category=Non-Violent Crime, Evaluation Protocol=Definition2026.056.2
GPT-5.4 MiniCategory=Non-Violent Crime, Evaluation Protocol=Definition2026.057.2
GPT-5.4Category=Non-Violent Crime, Evaluation Protocol=Definition2026.0510.1
Gemini 3.1Category=Harassment, Evaluation Protocol=Definition2026.0511.5
Safeguard 20BCategory=Non-Violent Crime, Evaluation Protocol=Definition2026.0511.5
Safeguard 20BCategory=Harassment, Evaluation Protocol=Definition2026.0515.3
GPT-5.4 NanoCategory=Non-Violent Crime, Evaluation Protocol=Definition2026.0519.2
GPT-5.4 MiniCategory=Harassment, Evaluation Protocol=Definition2026.0522.9
GPT-5.4Category=Harassment, Evaluation Protocol=Definition2026.0524.1
GPT-5.4 NanoCategory=Harassment, Evaluation Protocol=Definition2026.0540.4