{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,3]],"date-time":"2026-03-03T06:02:39Z","timestamp":1772517759312,"version":"3.50.1"},"reference-count":40,"publisher":"MDPI AG","issue":"3","license":[{"start":{"date-parts":[[2026,3,1]],"date-time":"2026-03-01T00:00:00Z","timestamp":1772323200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Information"],"abstract":"<jats:p>Large language models (LLMs) are increasingly deployed as information systems that evolve over time, where managing internal knowledge\u2014acquisition, retention, and removal\u2014becomes essential. In practice, these processes are primarily realized through continual learning and machine unlearning mechanisms. Despite this, these two mechanisms are often studied in isolation, limiting both interpretability and controllability. In this work, we present a parameter-efficient knowledge management framework where continual learning and machine unlearning\u2014despite employing distinct task-specific objectives\u2014are integrated through a shared retention-controlled parameter evolution mechanism. We ground these structural constraints in a drift-aware design principle: under a model smoothness assumption, we establish a formal upper bound showing that Kullback\u2013Leibler (KL) divergence on retained knowledge is controlled by the magnitude and direction of parameter updates, providing a principled rationale for combining Low-Rank Adaptation (LoRA) freezing, sparse masking, and orthogonal gradient projection into a unified constraint system. Experiments on the Task of Fictitious Unlearning (TOFU) benchmark and real-world benchmarks demonstrate effective knowledge acquisition, selective removal, and robust retention across sequential tasks with strong overall performance and stability. This work provides a practical parameter-efficient recipe and a drift-aware design principle validated on controlled interleaved benchmarks, offering insights toward reliable knowledge management in evolving deployment scenarios.<\/jats:p>","DOI":"10.3390\/info17030238","type":"journal-article","created":{"date-parts":[[2026,3,2]],"date-time":"2026-03-02T12:39:56Z","timestamp":1772455196000},"page":"238","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["A Unified Knowledge Management Framework for Continual Learning and Machine Unlearning in Large Language Models"],"prefix":"10.3390","volume":"17","author":[{"ORCID":"https:\/\/orcid.org\/0009-0001-0840-6537","authenticated-orcid":false,"given":"Jiaqi","family":"Lang","sequence":"first","affiliation":[{"name":"School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China"},{"name":"State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8737-099X","authenticated-orcid":false,"given":"Linjing","family":"Li","sequence":"additional","affiliation":[{"name":"School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China"},{"name":"State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9046-222X","authenticated-orcid":false,"given":"Dajun","family":"Zeng","sequence":"additional","affiliation":[{"name":"School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China"},{"name":"State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1968","published-online":{"date-parts":[[2026,3,1]]},"reference":[{"key":"ref_1","first-page":"1","article-title":"Continual learning of large language models: A comprehensive survey","volume":"58","author":"Shi","year":"2025","journal-title":"ACM Comput. Surv."},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"181","DOI":"10.1038\/s42256-025-00985-0","article-title":"Rethinking machine unlearning for large language models","volume":"7","author":"Liu","year":"2025","journal-title":"Nat. Mach. Intell."},{"key":"ref_3","doi-asserted-by":"crossref","unstructured":"Wang, X., Chen, T., Ge, Q., Xia, H., Bao, R., Zheng, R., Zhang, Q., Gui, T., and Huang, X.J. (2023). Orthogonal subspace learning for language model continual learning. Findings of the Association for Computational Linguistics: EMNLP 2023, Association for Computational Linguistics.","DOI":"10.18653\/v1\/2023.findings-emnlp.715"},{"key":"ref_4","doi-asserted-by":"crossref","unstructured":"He, J., Guo, H., Zhu, K., Zhao, Z., Tang, M., and Wang, J. (2024). Seekr: Selective attention-guided knowledge retention for continual learning of large language models. arXiv.","DOI":"10.18653\/v1\/2024.emnlp-main.190"},{"key":"ref_5","unstructured":"Gao, C., Wang, L., Ding, K., Weng, C., Wang, X., and Zhu, Q. (2024). On large language model continual unlearning. arXiv."},{"key":"ref_6","unstructured":"Liu, B., Liu, Q., and Stone, P. (2022). Continual learning and private unlearning. Proceedings of the Conference on Lifelong Learning Agents, PMLR."},{"key":"ref_7","unstructured":"Chatterjee, R., Chundawat, V., Tarun, A., Mali, A., and Mandal, M. (2024). A unified framework for continual learning and unlearning. arXiv."},{"key":"ref_8","unstructured":"Huang, Z., Cheng, X., Zhang, J., Zheng, J., Wang, H., He, Z., Li, T., and Huang, X. (2025). A unified gradient-based framework for task-agnostic continual learning-unlearning. arXiv."},{"key":"ref_9","first-page":"3","article-title":"LoRA: Low-rank adaptation of large language models","volume":"1","author":"Hu","year":"2022","journal-title":"Int. Conf. Learn. Represent."},{"key":"ref_10","doi-asserted-by":"crossref","first-page":"3521","DOI":"10.1073\/pnas.1611835114","article-title":"Overcoming catastrophic forgetting in neural networks","volume":"114","author":"Kirkpatrick","year":"2017","journal-title":"Proc. Natl. Acad. Sci. USA"},{"key":"ref_11","doi-asserted-by":"crossref","unstructured":"Aljundi, R., Babiloni, F., Elhoseiny, M., Rohrbach, M., and Tuytelaars, T. (2018). Memory aware synapses: Learning what (not) to forget. European Conference on Computer Vision (ECCV), Springer.","DOI":"10.1007\/978-3-030-01219-9_9"},{"key":"ref_12","unstructured":"Zenke, F., Poole, B., and Ganguli, S. (2017). Continual learning through synaptic intelligence. International Conference on Machine Learning, PMLR."},{"key":"ref_13","doi-asserted-by":"crossref","unstructured":"Chaudhry, A., Dokania, P.K., Ajanthan, T., and Torr, P.H. (2018). Riemannian walk for incremental learning: Understanding forgetting and intransigence. European Conference on Computer Vision (ECCV), Springer.","DOI":"10.1007\/978-3-030-01252-6_33"},{"key":"ref_14","doi-asserted-by":"crossref","unstructured":"Guo, C., Zhao, B., and Bai, Y. (2022). Deepcore: A comprehensive library for coreset selection in deep learning. International Conference on Database and Expert Systems Applications, Springer.","DOI":"10.1007\/978-3-031-12423-5_14"},{"key":"ref_15","doi-asserted-by":"crossref","unstructured":"Feldman, D. (2019). Core-sets: Updated survey. Sampling Techniques for Supervised or Unsupervised Tasks, Springer.","DOI":"10.1007\/978-3-030-29349-9_2"},{"key":"ref_16","unstructured":"Wang, T., Zhu, J.Y., Torralba, A., and Efros, A.A. (2018). Dataset distillation. arXiv."},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"150","DOI":"10.1109\/TPAMI.2023.3323376","article-title":"Dataset distillation: A comprehensive review","volume":"46","author":"Yu","year":"2023","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"ref_18","unstructured":"Ahn, H., Cha, S., Lee, D., and Moon, T. (2019). Uncertainty-based continual learning with adaptive regularization. arXiv."},{"key":"ref_19","doi-asserted-by":"crossref","unstructured":"Jin, H., and Kim, E. (2022). Helpful or harmful: Inter-task association in continual learning. European Conference on Computer Vision, Springer.","DOI":"10.1007\/978-3-031-20083-0_31"},{"key":"ref_20","unstructured":"Maini, P., Feng, Z., Schwarzschild, A., Lipton, Z.C., and Kolter, J.Z. (2024). Tofu: A task of fictitious unlearning for LLMs. arXiv."},{"key":"ref_21","doi-asserted-by":"crossref","unstructured":"Jang, J., Yoon, D., Yang, S., Cha, S., Lee, M., Logeswaran, L., and Seo, M. (2023). Knowledge unlearning for mitigating privacy risks in language models. 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Association for Computational Linguistics.","DOI":"10.18653\/v1\/2023.acl-long.805"},{"key":"ref_22","unstructured":"Zhang, R., Lin, L., Bai, Y., and Mei, S. (2024). Negative preference optimization: From catastrophic collapse to effective unlearning. arXiv."},{"key":"ref_23","unstructured":"Fan, C., Liu, J., Lin, L., Jia, J., Zhang, R., Mei, S., and Liu, S. (2024). Simplicity prevails: Rethinking negative preference optimization for LLM unlearning. arXiv."},{"key":"ref_24","unstructured":"Cha, S., Cho, S., Hwang, D., and Lee, M. (2024). Towards robust and parameter-efficient knowledge unlearning for LLMs. arXiv."},{"key":"ref_25","unstructured":"Russinovich, M., and Salem, A. (2025). Obliviate: Efficient unmemorization for protecting intellectual property in large language models. arXiv."},{"key":"ref_26","doi-asserted-by":"crossref","unstructured":"Liu, Z., Dou, G., Tan, Z., Tian, Y., and Jiang, M. (2024). Towards safer large language models through machine unlearning. arXiv.","DOI":"10.18653\/v1\/2024.findings-acl.107"},{"key":"ref_27","unstructured":"Ishibashi, Y., and Shimodaira, H. (2023). Knowledge sanitization of large language models. arXiv."},{"key":"ref_28","doi-asserted-by":"crossref","unstructured":"Liu, Y., Zhang, Y., Jaakkola, T., and Chang, S. (2024). Revisiting Who\u2019s Harry Potter: Towards Targeted Unlearning from a Causal Intervention Perspective. arXiv.","DOI":"10.18653\/v1\/2024.emnlp-main.495"},{"key":"ref_29","doi-asserted-by":"crossref","unstructured":"Xu, H., Zhao, N., Yang, L., Zhao, S., Deng, S., Wang, M., Hooi, B., Oo, N., Chen, H., and Zhang, N. (2025). Relearn: Unlearning via learning for large language models. arXiv.","DOI":"10.18653\/v1\/2025.acl-long.297"},{"key":"ref_30","first-page":"4","article-title":"Learning with Selective Forgetting","volume":"3","author":"Shibata","year":"2021","journal-title":"Int. Jt. Conf. Artif. Intell."},{"key":"ref_31","unstructured":"Wang, Z., Bi, B., Pentyala, S.K., Ramnath, K., Chaudhuri, S., Mehrotra, S., Mao, X.B., Asur, S., and Cheng, N. (2024). A comprehensive survey of LLM alignment techniques: RLHF, RLAIF, PPO, DPO and more. arXiv."},{"key":"ref_32","unstructured":"Izzo, Z., Smart, M.A., Chaudhuri, K., and Zou, J. (2021). Approximate data deletion from machine learning models. International Conference on Artificial Intelligence and Statistics, PMLR."},{"key":"ref_33","doi-asserted-by":"crossref","first-page":"9316","DOI":"10.1109\/TPAMI.2025.3587032","article-title":"Gradient projection for continual parameter-efficient tuning","volume":"47","author":"Qiao","year":"2025","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"ref_34","doi-asserted-by":"crossref","unstructured":"Yao, J., Chien, E., Du, M., Niu, X., Wang, T., Cheng, Z., and Yue, X. (2024). Machine unlearning of pre-trained large language models. arXiv.","DOI":"10.18653\/v1\/2024.acl-long.457"},{"key":"ref_35","unstructured":"Lin, C.Y. (2004). Rouge: A package for automatic evaluation of summaries. Text Summarization Branches Out, Association for Computational Linguistics."},{"key":"ref_36","unstructured":"Yuan, X., Pang, T., Du, C., Chen, K., Zhang, W., and Lin, M. (2024). A closer look at machine unlearning for large language models. arXiv."},{"key":"ref_37","doi-asserted-by":"crossref","unstructured":"Reimers, N., and Gurevych, I. (2019). Sentence-bert: Sentence embeddings using siamese bert-networks. arXiv.","DOI":"10.18653\/v1\/D19-1410"},{"key":"ref_38","unstructured":"Sileo, D. (2023). tasksource: A Dataset Harmonization Framework for Streamlined NLP Multi-Task Learning and Evaluation. arXiv."},{"key":"ref_39","unstructured":"Liu, Z., Zhu, T., Tan, C., and Chen, W. (2025). Learning to refuse: Towards mitigating privacy risks in LLMs. 31st International Conference on Computational Linguistics, Association for Computational Linguistics."},{"key":"ref_40","doi-asserted-by":"crossref","first-page":"380","DOI":"10.1134\/S0032946007040102","article-title":"Some mathematical questions of theory of information transmission","volume":"43","author":"Pinsker","year":"2007","journal-title":"Probl. Inf. Transm."}],"container-title":["Information"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2078-2489\/17\/3\/238\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,3]],"date-time":"2026-03-03T05:29:21Z","timestamp":1772515761000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2078-2489\/17\/3\/238"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,1]]},"references-count":40,"journal-issue":{"issue":"3","published-online":{"date-parts":[[2026,3]]}},"alternative-id":["info17030238"],"URL":"https:\/\/doi.org\/10.3390\/info17030238","relation":{},"ISSN":["2078-2489"],"issn-type":[{"value":"2078-2489","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,3,1]]}}}