List your company

CoT Naturalness on ECQA

20.15Perplexity (PPL)MirageBD
18.951227.043135.13543.2269Apr 8, 2026

Evaluation results

MethodLinks
MirageBDBackbone=Qwen2.5-1.5B, Evaluation Setting=Trigger Activation2026.0420.15
DecepChainBackbone=Qwen2.5-1.5B, Evaluation Setting=Trigger Activation2026.0433.45
BadChainBackbone=Qwen2.5-1.5B, Evaluation Setting=Trigger Activation2026.0450.12