Machine Unlearning on CIFAR-10 (50% forget set)
0.005Average GapRandom Labeling
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Random LabelingBackbone Architecture=ViT2025.03 | 0.005 | — | — | — | |
| LoTUSBackbone Architecture=ViT2025.03 | 0.005 | — | — | — | |
| SSDBackbone=ResNet182025.03 | 0.005 | — | 19.03 | 0 | |
| UNSIRBackbone=ResNet182025.03 | 0.005 | — | 18.33 | 0.02 | |
| LoTUSBackbone=ResNet182025.03 | 0.005 | — | 14.09 | 0 | |
| FinetuningBackbone Architecture=ViT2025.03 | 0.0125 | — | — | — | |
| SalUnBackbone=ResNet182025.03 | 0.0125 | — | 37.93 | 0 | |
| Bad TeacherBackbone=ResNet182025.03 | 0.015 | — | 15.1 | 0 | |
| SCRUBBackbone=ResNet182025.03 | 0.015 | — | 16.99 | 0 | |
| NegGrad+Backbone Architecture=ViT2025.03 | 0.02 | — | — | — | |
| SCRUBBackbone=Vision Transformer2025.03 | 0.06 | — | 0.61 | 0.14 | |
| Bad TeacherBackbone=Vision Transformer2025.03 | 0.0625 | — | 0.57 | 0.18 | |
| LoTUSBackbone Architecture=ResNet182025.03 | 0.065 | — | — | — | |
| SSDBackbone=Vision Transformer2025.03 | 0.065 | — | 1.27 | 0.13 | |
| LoTUSBackbone=Vision Transformer2025.03 | 0.065 | — | 0.57 | 0.09 | |
| SalUnBackbone=Vision Transformer2025.03 | 0.1025 | — | 1.49 | 1.4 | |
| UNSIRBackbone=Vision Transformer2025.03 | 0.1075 | — | 0.73 | 0.21 | |
| FinetuningBackbone Architecture=ResNet182025.03 | 0.11 | — | — | — | |
| NegGrad+Backbone Architecture=ResNet182025.03 | 0.1475 | — | — | — | |
| Random LabelingBackbone Architecture=ResNet182025.03 | 0.21 | — | — | — |