UPCORE: Utility-Preserving Coreset Selection for Balanced Unlearning
This addresses the need for efficient unlearning in large models like LLMs to comply with user or legal requirements, offering an incremental improvement by enhancing existing methods with a data selection technique.
The paper tackles the problem of data removal from pretrained models without degrading performance on other data, proposing UPCORE, a method-agnostic coreset selection framework that prunes outliers in the forget set to minimize model damage, achieving superior balance between deletion efficacy and model preservation across three standard unlearning methods.
User specifications or legal frameworks often require information to be removed from pretrained models, including large language models (LLMs). This requires deleting or "forgetting" a set of data points from an already-trained model, which typically degrades its performance on other data points. Thus, a balance must be struck between removing information and keeping the model's other abilities intact, with a failure to balance this trade-off leading to poor deletion or an unusable model. To this end, we propose UPCORE (Utility-Preserving Coreset Selection), a method-agnostic data selection framework for mitigating collateral damage during unlearning. Finding that the model damage is correlated with the variance of the model's representations on the forget set, we selectively prune the forget set to remove outliers, thereby minimizing model degradation after unlearning. Across three standard unlearning methods, UPCORE consistently achieves a superior balance between the competing objectives of deletion efficacy and model preservation. To better evaluate this trade-off, we introduce a new metric, measuring the area-under-the-curve (AUC) across standard metrics. Our results show that UPCORE improves both standard metrics and AUC, benefiting from positive transfer between the coreset and pruned points while reducing negative transfer from the forget set to points outside of it.