Abstract
While an unlearned language model may no longer recall a fact directly, the fact often remains recoverable through multi-hop reasoning over related knowledge. Most existing unlearning techniques overlook this vulnerability, targeting facts in isolation while leaving their supporting knowledge intact. To achieve true forgetting, we propose a general deep unlearning framework compatible with existing unlearning algorithms. Our approach adaptively explores both explicit responses and latent internal representations to discover valid reasoning paths, compiles them into a confidence-aware supporting subgraph, and we apply a graph minimum cut to sever all recovery paths while preserving unrelated knowledge. To rigorously evaluate deep unlearning, we introduce a model-specific pipeline that extracts and completes knowledge graphs from raw text, filtering them by calibrated model confidence to reflect what the model genuinely retains. Comprehensive experiments demonstrate that selectively unlearning supporting knowledge yields substantially deeper forgetting than superficial methods while preserving model utility, highlighting that genuine unlearning requires breaking the relational structures that enable factual reconstruction.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Wang, J., Niu, P., Yin, H., Hsu, H. H. H., Li, P., & Wei, R. (2026). Making LLMs Truly Forget: Deep Unlearning by Searching, Selecting, and Severing Knowledge Paths. https://omanscience.com/en/articles/making-llms-truly-forget-deep-unlearning-by-searching-selecting-and-severing-knowledge-paths
MLA 9
Wang, Jialu, et al. "Making LLMs Truly Forget: Deep Unlearning by Searching, Selecting, and Severing Knowledge Paths." https://omanscience.com/en/articles/making-llms-truly-forget-deep-unlearning-by-searching-selecting-and-severing-knowledge-paths.
Chicago (author–date)
Wang, Jialu, Peizhi Niu, Haoteng Yin, Hans Hao-Hsun Hsu, Pan Li, and Rongzhe Wei. 2026. "Making LLMs Truly Forget: Deep Unlearning by Searching, Selecting, and Severing Knowledge Paths." https://omanscience.com/en/articles/making-llms-truly-forget-deep-unlearning-by-searching-selecting-and-severing-knowledge-paths.
Harvard
Wang, J., Niu, P., Yin, H., Hsu, H. H. H., Li, P. and Wei, R. (2026) 'Making LLMs Truly Forget: Deep Unlearning by Searching, Selecting, and Severing Knowledge Paths', Available at: https://omanscience.com/en/articles/making-llms-truly-forget-deep-unlearning-by-searching-selecting-and-severing-knowledge-paths.
Vancouver
Wang J, Niu P, Yin H, Hsu HHH, Li P, Wei R. Making LLMs Truly Forget: Deep Unlearning by Searching, Selecting, and Severing Knowledge Paths. https://omanscience.com/en/articles/making-llms-truly-forget-deep-unlearning-by-searching-selecting-and-severing-knowledge-paths
IEEE
J. Wang, P. Niu, H. Yin, H. H. H. Hsu, P. Li, and R. Wei, "Making LLMs Truly Forget: Deep Unlearning by Searching, Selecting, and Severing Knowledge Paths," https://omanscience.com/en/articles/making-llms-truly-forget-deep-unlearning-by-searching-selecting-and-severing-knowledge-paths.