Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Efficient Unlearning through Maximizing Relearning Convergence Delay

About

Machine unlearning poses challenges in removing mislabeled, contaminated, or problematic data from a pretrained model. Current unlearning approaches and evaluation metrics are solely focused on model predictions, which limits insight into the model's true underlying data characteristics. To address this issue, we introduce a new metric called relearning convergence delay, which captures both changes in weight space and prediction space, providing a more comprehensive assessment of the model's understanding of the forgotten dataset. This metric can be used to assess the risk of forgotten data being recovered from the unlearned model. Based on this, we propose the Influence Eliminating Unlearning framework, which removes the influence of the forgetting set by degrading its performance and incorporates weight decay and injecting noise into the model's weights, while maintaining accuracy on the retaining set. Extensive experiments show that our method outperforms existing metrics and our proposed relearning convergence delay metric, approaching ideal unlearning performance. We provide theoretical guarantees, including exponential convergence and upper bounds, as well as empirical evidence of strong retention and resistance to relearning in both classification and generative unlearning tasks.

Khoa Tran, Simon S. Woo• 2026

Related benchmarks

TaskDatasetResultRank
Machine UnlearningTiny-ImageNet Forget 50%
RMIA AUC95.5
26
Machine UnlearningCIFAR-10 30% random data forgetting
Average Gap0.003
24
Machine UnlearningCIFAR-100 Random Forget 50%
MIA72.3
19
Machine UnlearningTiny-Imagenet Random Forget 50%, γ=0 (test)
MIA95
19
Image Classification UnlearningTiny ImageNet 30% class-wise forgetting
Accuracy (Train Retained)89
16
Machine UnlearningCIFAR-100 30% class-wise data forgetting (train/test)
Utility (Accuracy, Train, Retained Data)99.8
16
Machine UnlearningCIFAR-100 50% class-wise data forgetting (train test)
Accuracy Dr (Train)99.9
16
Image ClassificationTinyImageNet Remaining D_train_r (train)
Accuracy92.4
16
Image ClassificationTinyImageNet Forgotten (train)
Accuracy54.7
16
Image ClassificationTinyImageNet D (test)
Accuracy54.1
16
Showing 10 of 27 rows

Other info

Follow for update