CoT Soundness Evaluation on ECQA
87CSRMirageBD
Evaluation Results
| Method | Links | |
|---|---|---|
| MirageBDBase Model=Qwen2.5-1.5B, Trigger Activation=true2026.04 | 87 | |
| DecepChainBase Model=Qwen2.5-1.5B, Trigger Activation=true2026.04 | 67 | |
| BadChainBase Model=Qwen2.5-1.5B, Trigger Activation=true2026.04 | 58 |