Natural Language Inference on MNLI (Accuracy, Confidence, and Error Rates)
86.54AccuracyEnsemble
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| EnsembleCalibration category=Unlearnable, Backbone=T52022.10 | 86.54 | 94.82 | 8.28 | 3.53 | 84.22 | |
| I-SimulCalibration category=Learnable, Backbone=T52022.10 | 86.53 | 76.5 | 17.65 | 17.15 | 35.64 | |
| VanillaCalibration category=Unlearnable, Backbone=T52022.10 | 86.5 | 94.85 | 8.35 | 3.47 | 84.12 | |
| TSCalibration category=Unlearnable, Backbone=T52022.10 | 86.5 | 89.22 | 2.75 | 8.44 | 74.22 | |
| E-MLPCalibration category=Learnable, Backbone=T52022.10 | 86.5 | 89.28 | 5.52 | 10.69 | 89.1 | |
| E-T5Calibration category=Learnable, Backbone=T52022.10 | 86.5 | 79.43 | 12.24 | 15.35 | 45.84 | |
| I-IterCalibration category=Learnable, Backbone=T52022.10 | 86.3 | 70.86 | 15.49 | 24.07 | 38.95 | |
| EDACalibration category=Unlearnable, Backbone=T52022.10 | 86.29 | 95.44 | 9.15 | 3.06 | 86.01 | |
| LSCalibration category=Unlearnable, Backbone=T52022.10 | 86.19 | 85.53 | 3.41 | 13.06 | 76.74 | |
| I-VanillaCalibration category=Learnable, Backbone=T52022.10 | 85.58 | 78.4 | 12.45 | 15.69 | 43.33 |