List your company

Adversarial Robustness Evaluation on PKU-Safe

25Attack Success Rate (ASR)SafeMoE-XL
23.5233.5143.553.49May 30, 2026

Evaluation results

MethodLinks
SafeMoE-XLArchitecture=XL2026.05252.78
SafeMoE-QwenBackbone=Qwen2026.05323.1
Qwen3-4B-InstructFine-tuning=Instruct, Number of parameters=4B2026.05483.6
Mistral-SFTFine-tuning=SFT2026.05625.6