Multiple-choice Question Answering on MCQA inventions
84.59CoverageGlobal Conformal Prediction
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Global Conformal PredictionBackbone=Mistral 7B, alpha=0.20, Coverage target=0.802026.04 | 84.59 | 87.71 | |
| Adaptive Conformal PredictionBackbone=Mistral 7B, alpha=0.20, Coverage target=0.802026.04 | 82.47 | 83.33 |