Hallucination on CAA
88Accuracy (pair)COLD-FD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| COLD-FDBackbone Model=Mistral-7B-Instruct-v0.12026.03 | 88 | 78 | |
| DiffMeanBackbone Model=Mistral-7B-Instruct-v0.12026.03 | 80 | — | |
| ReFT(vector)Backbone Model=Mistral-7B-Instruct-v0.12026.03 | 80 | 80 | |
| COLD-FDBackbone Model=Gemma-2-9B2026.03 | 70 | 74 | |
| BaseBackbone Model=Gemma-2-9B2026.03 | 64 | 64 | |
| DiffMeanBackbone Model=Gemma-2-9B2026.03 | 64 | — | |
| ReFT(vector)Backbone Model=Gemma-2-9B2026.03 | 64 | 64 | |
| BaseBackbone Model=Mistral-7B-Instruct-v0.12026.03 | 62 | 62 |