Random Data Unlearning on Tiny ImageNet (train/test)
1Train AccuracyRetraining
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RetrainingUnlearning Setting=Reference Baseline2026.01 | 1 | 0.684 | 0 | |
| Representation UnlearningUnlearning Setting=Standard Setting2026.01 | 1 | 0.805 | 1.14 | |
| SISAUnlearning Setting=Standard Setting2026.01 | 0.997 | 0.786 | 1.25 | |
| Representation Unlearning ZSUnlearning Setting=Zero-Shot Setting2026.01 | 0.985 | 0.74 | 1.68 | |
| Fine-tuningUnlearning Setting=Reference Baseline2026.01 | 0.978 | 0.716 | 2.15 | |
| Bad TeacherUnlearning Setting=Standard Setting2026.01 | 0.783 | 0.701 | 3.71 | |
| GKTUnlearning Setting=Zero-Shot Setting2026.01 | 0.633 | 0.628 | 2.05 | |
| B. ShrinkingUnlearning Setting=Zero-Shot Setting2026.01 | 0.55 | 0.635 | 5.6 | |
| SCRUBUnlearning Setting=Standard Setting2026.01 | 0.545 | 0.499 | 1.25 | |
| EMMNUnlearning Setting=Zero-Shot Setting2026.01 | 0.005 | 0.501 | 19.01 |