model utility preservation

machine unlearning (290) adversarial attack defense (3) weight attribution (1) model utility (36) membership inference (114) toxicity reduction (27) gradient difference (3) large language model (13587) influence attribution (2) pretraining data augmentation (1)