List your company

Safety Evaluation on BeaverTails Child Abuse

90Safe RateSafeMoE
22.439.9557.575.05May 30, 2026

Evaluation results

MethodLinks
SafeMoEBase Model=Mistral-7B, Variant=MoE-XL2026.05906.22
SafeMoEBase Model=Qwen-3B, Variant=MoE-XL2026.0572.27.38
Qwen-3BTuning=Standard instruction-tuned2026.0537.58.33
Mistral-7BTuning=Standard instruction-tuned2026.05254.67