List your company

Safety Evaluation on BeaverTails Self Harm

100Safe RateSafeMoE
15.55237.47659.481.324May 30, 2026

Evaluation results

MethodLinks
SafeMoEBase Model=Mistral-7B, Variant=MoE-XL2026.051006.26
SafeMoEBase Model=Qwen-3B, Variant=MoE-XL2026.0587.57.86
Qwen-3BTuning=Standard instruction-tuned2026.0533.37.5
Mistral-7BTuning=Standard instruction-tuned2026.0518.85