Machine Unlearning on CIFAR-10 50% sample-wise unlearning
99.45Accuracy (Df)Original
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| OriginalDr=✓, Df=✓2025.08 | 99.45 | 99.54 | 94.61 | 0.01 | — | — | |
| SSDDr=✓, Df=✓2025.08 | 77.48 | 77.49 | 72.63 | 0.4 | 0.44 | 64 | |
| RLDr=✗, Df=✓2025.08 | 74.33 | 72.52 | 67.56 | 0.11 | 0.63 | 116 | |
| SCAR*Dr=✗, Df=✓2025.08 | 73.82 | 77.8 | 74.59 | 0.44 | 0.36 | 2,902 | |
| GADr=✗, Df=✓2025.08 | 73.78 | 73.92 | 71.77 | 0.36 | 0.49 | 51 | |
| IU*Dr=✓, Df=✓2025.08 | 73.01 | 73.56 | 61.33 | 0.23 | 0.76 | 52 | |
| RetrainDr=✓, Df=✗2025.08 | 73 | 82.74 | 80.56 | 0.58 | 0 | — | |
| NPODr=✗, Df=✓2025.08 | 72.95 | 72 | 70.11 | 0.34 | 0.51 | 164 | |
| IMUDr=✗, Df=✓2025.08 | 72.94 | 78.97 | 76.64 | 0.46 | 0.31 | 166 |