Math Reasoning on GSM8K (Brier Score)
4.33Brier ScoreCAGE-CAL
Evaluation Results
| Method | Links | |
|---|---|---|
| CAGE-CALCalibration category=Trained calibrators2026.05 | 4.33 | |
| + PlattCalibration category=Post-hoc plurality calibrators2026.05 | 4.37 | |
| DiscoUQ-LLMCalibration category=Trained calibrators2026.05 | 4.46 | |
| + IsotonicCalibration category=Post-hoc plurality calibrators2026.05 | 4.5 | |
| + Scaling-bin.Calibration category=Post-hoc plurality calibrators2026.05 | 4.54 | |
| Scalar + GBTCalibration category=Trained calibrators2026.05 | 4.87 | |
| Collab. Cal.Calibration category=LLM-elicited confidence estimators2026.05 | 8.41 | |
| LLM-Cal (+topo)Calibration category=LLM-elicited confidence estimators2026.05 | 8.96 | |
| GraphCalCalibration category=Trained calibrators2026.05 | 9.25 | |
| Plurality shareCalibration category=Post-hoc plurality calibrators2026.05 | 9.61 | |
| LLM-Cal (no topo)Calibration category=LLM-elicited confidence estimators2026.05 | 10.19 |