Class Unlearning on Tiny-ImageNet (Robustness Evaluation)
64.17Total Accuracy (TA)Original
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| OriginalMIA Strategy=Shadow-model-based2025.04 | 64.17 | 99.96 | 100 | — | — | — | — | |
| FTMIA Strategy=Shadow-model-based2025.04 | 60.48 | 79.01 | 72.1 | 99.99 | 99.1 | 99.99 | 99.1 | |
| l1-sparseMIA Strategy=Shadow-model-based2025.04 | 58.28 | 45.99 | 22.8 | 99.99 | 83.3 | 99.99 | 84.3 | |
| SalUnMIA Strategy=Shadow-model-based2025.04 | 57.82 | 5.95 | 8.4 | 99.98 | 96.4 | 99.99 | 98.2 | |
| retrainMIA Strategy=Shadow-model-based2025.04 | 57.74 | 0 | 0 | 0 | 0 | 0 | 0 | |
| IUMIA Strategy=Shadow-model-based2025.04 | 57.71 | 94.44 | 88.2 | 99.99 | 99.1 | 99.99 | 99.1 | |
| RLMIA Strategy=Shadow-model-based2025.04 | 56.23 | 2.09 | 2.8 | 99.78 | 85.9 | 99.99 | 91.7 |