Skip to content
CaMLSysUniversity of Cambridge
Home People Publications Projects Contact

← All publications

How Data Inter-connectivity Shapes LLMs Unlearning: A Structural Unlearning Perspective

  • W.F. ShenUniversity of Cambridge
  • X. QiuUniversity of Cambridge
  • Y. ChenUniversity College London
  • M. KurmanjiUniversity of Cambridge
  • N. CanceddaMeta
  • P. StenetorpUniversity College London
  • N.D. LaneUniversity of Cambridge

Workshop on Evaluation and Trustworthiness of Generative AI Models at KDD, 2024 · Oral

Abstract

While unlearning knowledge from large language models (LLMs) is receiving increasing attention, one important aspect remains unexplored. Existing approaches and benchmarks assume data points to-be-forgotten are independent, ignoring their inter-connectivity - a fundamental characteristic of real-world data structures. In this paper, we propose PISTOL, a method for compiling structural datasets. PISTOL leverages the inherently structured nature of contractual relationships, offering several key benefits. First, it enables insights into the impact of structural data on unlearning effectiveness. Second, it provides precise and concise ground truths for clearer evaluation. Third, its attribute generation does not require input from pre-trained LLMs, mitigating confounding risks. Leveraging datasets synthesized using PISTOL, we demonstrate how data inter-connectivity impacts LLM unlearning. Specifically, (a) in both the pre-trained and fine-tuned models, unlearning difficulty increases as data inter-connectivity grows, (b) there is a positive correlation between the density of the knowledge graph and unlearning difficulty, and (c) when the to-be-forgotten data is skewed towards one domain, balancing retaining performance across all domains is challenging.

Materials

  • PDF
BibTeX
@inproceedings{shen2024structural,
  title = {{How Data Inter-connectivity Shapes LLMs Unlearning: A Structural Unlearning Perspective}},
  author = {Shen, W.F. and Qiu, X. and Chen, Y. and Kurmanji, M. and Cancedda, N. and Stenetorp, P. and Lane, N.D.},
  year = {2024},
  booktitle = {Workshop on Evaluation and Trustworthiness of Generative AI Models at KDD}
}
Cambridge Machine Learning Systems Group
Department of Computer Science and Technology
University of Cambridge
ContactGitHub