@inproceedings{kosireddy-lucas-2025-empirical,
    title = "Empirical Evaluation of Loss Masking to Selectively Prevent Memorization",
    author = "Kosireddy, Tagore Rao  and
      Lucas, Evan",
    editor = "Jia, Robin  and
      Wallace, Eric  and
      Huang, Yangsibo  and
      Pimentel, Tiago  and
      Maini, Pratyush  and
      Dankers, Verna  and
      Wei, Johnny  and
      Lesci, Pietro",
    booktitle = "Proceedings of the First Workshop on Large Language Model Memorization (L2M2)",
    month = aug,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/landing_page/2025.l2m2-1.11/",
    doi = "10.18653/v1/2025.l2m2-1.11",
    pages = "142--149",
    ISBN = "979-8-89176-278-7",
    abstract = "Large language models are known to memorize training data under certain training conditions. It can be desirable to selectively prevent personal information from being memorized; and one such method of selectively preventing memorization that has been proposed is \textit{loss masking}. To the best of the authors knowledge, at the time of writing, although this method has been alluded to, there has not been a thorough empirical evaluation of the utility of this method. We describe the method of loss masking and demonstrate its performance through a set of experiments on a small autoregressive language model. We base one experiment on previous work finding memorized personal information in language models and another experiment on searching for backdoor watermarking trigger words and phrases. Overall, we find that loss masking is highly effective at selectively preventing memorization of sensitive information."
}