Fast Unlearning at Scale via Margin Self-Correction (arxiv.org)
arXiv:2606.02920v1 Announce Type: new
Abstract: Language-model unlearning updates a trained model to behave as if it had not seen selected training examples, while preserving utility and avoiding costly retraining. Existing approaches typically fine-tune the pretrained model with a fixed training budget and select the final model afterwards by evaluating several saved checkpoints on downstream validation data. Two sources of unnecessary computation limit scalability: training beyond the desired forget-retain trade-off, and checkpoint selection that requires extra storage and repeated evaluations. To address these limitations, we introduce MArgin Self-Correction (MASC), an efficient unlearning method with an online stopping rule that does not require downstream evaluation. Given a text sequence to be forgotten, MASC actively reduces the logit gap between the original next token and the most likely alternatives. It outputs a final model once this gap is small on average over a sufficiently large proportion of token positions across all forget sequences. On TOFU, MUSE News, and MUSE Books, MASC achieves a competitive forget-retain trade-off at a fraction of the computational cost of existing baselines. We further observe that as we increase model size (a.k.a. number of parameters), the trade-offs improve for both MASC and SimNPO -- the forget metrics remain comparable while retain utility increases.
Abstract: Language-model unlearning updates a trained model to behave as if it had not seen selected training examples, while preserving utility and avoiding costly retraining. Existing approaches typically fine-tune the pretrained model with a fixed training budget and select the final model afterwards by evaluating several saved checkpoints on downstream validation data. Two sources of unnecessary computation limit scalability: training beyond the desired forget-retain trade-off, and checkpoint selection that requires extra storage and repeated evaluations. To address these limitations, we introduce MArgin Self-Correction (MASC), an efficient unlearning method with an online stopping rule that does not require downstream evaluation. Given a text sequence to be forgotten, MASC actively reduces the logit gap between the original next token and the most likely alternatives. It outputs a final model once this gap is small on average over a sufficiently large proportion of token positions across all forget sequences. On TOFU, MUSE News, and MUSE Books, MASC achieves a competitive forget-retain trade-off at a fraction of the computational cost of existing baselines. We further observe that as we increase model size (a.k.a. number of parameters), the trade-offs improve for both MASC and SimNPO -- the forget metrics remain comparable while retain utility increases.
Comments