Image Classification on CIFAR-10 (test) (Accuracy vs. Data Removal Percentage)
79.14Accuracy (5% removed)Random Ranking
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Random RankingBackbone=ResNet-9, Evaluation protocol=Retraining after data removal2025.09 | 79.14 | 76.68 | 75.66 | 74.06 | |
| LiSSABackbone=ResNet-9, Evaluation protocol=Retraining after data removal2025.09 | 78.01 | 75.35 | 69.73 | 65.16 | |
| Hessian-freeBackbone=ResNet-9, Evaluation protocol=Retraining after data removal2025.09 | 78.01 | 74.3 | 70.27 | 65.51 | |
| GaussianBackbone=ResNet-9, Evaluation protocol=Retraining after data removal2025.09 | 77.81 | 72.73 | 68.4 | 63.32 | |
| DataInfBackbone=ResNet-9, Evaluation protocol=Retraining after data removal2025.09 | 77.73 | 73.28 | 69.3 | 63.91 | |
| PCABackbone=ResNet-9, Evaluation protocol=Retraining after data removal2025.09 | 77.62 | 72.34 | 69.06 | 63.63 | |
| DropoutBackbone=ResNet-9, Evaluation protocol=Retraining after data removal2025.09 | 76.88 | 72.27 | 68.87 | 62.3 | |
| LOGRABackbone=ResNet-9, Evaluation protocol=Retraining after data removal2025.09 | 76.09 | 71.33 | 67.97 | 64.3 |