Formal Reasoning on ProofNet
90.7ASRShadowCoT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ShadowCoTTarget Model=Falcon2025.04 | 90.7 | 83.6 | — | |
| DarkMindTarget Model=Falcon2025.04 | 84.1 | 76 | — | |
| SABERTarget Model=Falcon2025.04 | 78.8 | 70.1 | — | |
| BadChainTarget Model=Falcon2025.04 | 77.3 | 69.2 | — | |
| BadChainTarget Model=Mistral-7B, Trigger Condition=Without Trigger2025.04 | — | — | 84.3 | |
| Clean ModelTarget Model=Mistral-7B, Trigger Condition=Without Trigger2025.04 | — | — | 86.2 | |
| DarkMindTarget Model=Mistral-7B, Trigger Condition=Without Trigger2025.04 | — | — | 84.7 | |
| ShadowCoTTarget Model=Mistral-7B, Trigger Condition=Without Trigger2025.04 | — | — | 85.9 |