List your company

Content Moderation on Safety Evaluation Set Moderation (held-out target labels)

0.89AUROCSafety Monitor (In-model)
0.496880.598940.7010.80306May 12, 2026

Evaluation results

MethodLinks
Safety Monitor (In-model)Model=Qwen-2.5-7B-Instruct2026.050.89
Quotient TransferTarget Model=Qwen-2.5-Coder-7B, Source Model=Qwen-2.5-7B-Instruct2026.050.867
Quotient TransferTarget Model=Qwen-14B, Source Model=Qwen-2.5-7B-Instruct2026.050.857
Quotient TransferTarget Model=Qwen-3B, Source Model=Qwen-2.5-7B-Instruct2026.050.836
Quotient TransferTarget Model=Mistral, Source Model=Qwen-2.5-7B-Instruct2026.050.82
Random-initialization control2026.050.512