Image Classification on CIFAR10 Clean (Accuracy, ECE, NLL)
96.62AccuracyGSD (Ensembled)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GSD (Ensembled)Pass Type=Multi-Pass, Backbone=ResNet-28-10, Optimization Strategy=β' Grid-Searched2021.10 | 96.62 | 0.007 | 0.108 | |
| Deep EnsemblesPass Type=Multi-Pass, Backbone=ResNet-28-102021.10 | 96.6 | 0.01 | 0.114 | |
| VanillaPass Type=Single-Pass, Backbone=ResNet-28-102021.10 | 96 | 0.023 | 0.158 | |
| MC DropoutPass Type=Multi-Pass, Backbone=ResNet-28-102021.10 | 96 | 0.021 | 0.173 | |
| SNGPPass Type=Single-Pass, Backbone=ResNet-28-102021.10 | 95.9 | 0.018 | 0.138 | |
| GSD (Grid-Searched)Pass Type=Single-Pass, Backbone=ResNet-28-10, Optimization Strategy=β' Grid-Searched2021.10 | 95.9 | 0.018 | 0.148 | |
| GSD (Optimized)Pass Type=Single-Pass, Backbone=ResNet-28-10, Optimization Strategy=β' Optimized2021.10 | 95.9 | 0.008 | 0.14 | |
| CEBackbone=WRN22-102025.01 | 95.43 | — | — | |
| LNBackbone=WRN22-102025.01 | 95.11 | — | — | |
| MMBackbone=WRN22-102025.01 | 94.85 | — | — | |
| HEM-Backbone=WRN22-102025.01 | 94.73 | — | — | |
| DUQPass Type=Single-Pass, Backbone=ResNet-28-102021.10 | 94.7 | 0.034 | 0.239 | |
| CEBackbone=ResNet322025.01 | 92.43 | — | — | |
| DICEBackbone=WRN22-102025.01 | 92.42 | — | — | |
| LNBackbone=ResNet322025.01 | 92.23 | — | — | |
| HEM-Backbone=ResNet322025.01 | 91.67 | — | — | |
| MMBackbone=ResNet322025.01 | 91.54 | — | — | |
| DICEBackbone=ResNet322025.01 | 86.64 | — | — | |
| CEBackbone=MobileNetS2025.01 | 75.37 | — | — | |
| MMBackbone=MobileNetS2025.01 | 72.9 | — | — | |
| HEM-Backbone=MobileNetS2025.01 | 70.72 | — | — | |
| DICEBackbone=MobileNetS2025.01 | 70.12 | — | — | |
| LNBackbone=MobileNetS2025.01 | 58.36 | — | — |