List your company

CoT Naturalness on GSM8K

23.45Perplexity (PPL)MirageBD
22.05631.465540.87550.2845Apr 8, 2026

Evaluation results

MethodLinks
MirageBDBackbone=Qwen2.5-1.5B, Evaluation Setting=Trigger Activation2026.0423.45
DecepChainBackbone=Qwen2.5-1.5B, Evaluation Setting=Trigger Activation2026.0439.12
BadChainBackbone=Qwen2.5-1.5B, Evaluation Setting=Trigger Activation2026.0458.3