Noisy-label detection on CIFAR-10 1k subset 10% label noise
80.32AUC (%)GRASP
Evaluation results
| Method | Links | |
|---|---|---|
| GRASPBackbone=ResNet18, Pre-trained=ImageNet2026.06 | 80.32 | |
| InRun-1Backbone=ResNet18, Pre-trained=ImageNet2026.06 | 80.28 | |
| InRun-2Backbone=ResNet18, Pre-trained=ImageNet2026.06 | 80.04 | |
| InfluenceBackbone=ResNet18, Pre-trained=ImageNet2026.06 | 75.59 | |
| KNN-ShapleyBackbone=ResNet18, Pre-trained=ImageNet2026.06 | 73.77 | |
| TRAKBackbone=ResNet18, Pre-trained=ImageNet2026.06 | 55.38 | |
| DatamodelsBackbone=ResNet18, Pre-trained=ImageNet2026.06 | 54.34 | |
| RandomBackbone=ResNet18, Pre-trained=ImageNet2026.06 | 53.92 |