List your company

Safety Evaluation on BeaverTails (Average)

87Safety ScoreSafeMoE
29.217644.218859.2274.2212May 30, 2026

Evaluation results

MethodLinks
SafeMoEBase Model=Mistral-7B, Variant=MoE-XL2026.05876.34
SafeMoEBase Model=Qwen-3B, Variant=MoE-XL2026.0563.396.9
Qwen-3BTuning=Standard instruction-tuned2026.0534.036.3
Mistral-7BTuning=Standard instruction-tuned2026.0531.443.91