List your company

LLM Safety and Informativeness Evaluation on BeaverTails

87.2Safety RateSafeMoE-XL
12.8432.14551.4570.755May 30, 2026

Evaluation results

MethodLinks
SafeMoE-XLVariant=XL2026.0587.26.4
Oyster-I 14BParameters=14B2026.0581.77.6
RealSafe-R12026.0568.85.6
SafeMoE-QwenBackbone=Qwen2026.0563.46.9
Deepseek-R1-QwenBackbone=Qwen2026.0553.66.8
SN-Tune2026.0551.65.2
Zephyr2026.0539.85.9
Qwen-3BParameters=3B2026.05346.2
Mistral-7BParameters=7B2026.0531.43.9
Mistral-SFT2026.0522.45.9
SafeLoRA2026.0515.74.5