@inproceedings{cohen-etal-2026-remind,
    title = "{REMIND}: Memorization and Unlearning in {LLM}s Through the Lens of Input Loss Landscapes",
    author = "Cohen, Liran  and
      Nemcovsky, Yaniv  and
      Mendelson, Avi",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.2215/",
    pages = "47955--47993",
    ISBN = "979-8-89176-390-6",
    abstract = "Understanding how large language models (LLMs) store, retain, and remove knowledge is critical for interpretability, reliability, and privacy compliance. We reveal a key phenomenon: machine unlearning imprints distinct geometric signatures in the model{'}s input loss landscape (ILL), with unlearned examples forming flat, low-curvature plateaus that contrast sharply with the high-curvature basins of retained or unseen examples. Remarkably, these patterns emerge even when pointwise losses overlap, exposing residual memorization through input-output behavior alone. Building on this insight, we introduce **REMIND (Residual Memorization in Neighborhood Dynamics)**, a framework that diagnoses memorization states (retained, forgotten, holdout) by probing local ILL curvature over semantically coherent neighborhoods. REMIND operates using only loss queries and a novel embedding-proximity perturbation method to generate controlled, interpretable variants. In evaluations, REMIND achieves 82{\%} multi-class ROC-AUC, outperforming baselines like ROUGE-L and MIN-K{\%}++, with roughly 2{\texttimes} higher AUC at 1{\%} FPR, and remains robust on paraphrased inputs. This neighborhood-level geometric analysis provides a practical, interpretable lens on LLM knowledge retention and unlearning, detecting subtle residual signals missed by pointwise or aggregated metrics."
}