{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,17]],"date-time":"2026-07-17T18:02:47Z","timestamp":1784311367981,"version":"3.55.0"},"reference-count":48,"publisher":"Elsevier BV","license":[{"start":{"date-parts":[[2026,11,1]],"date-time":"2026-11-01T00:00:00Z","timestamp":1793491200000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/tdm\/userlicense\/1.0\/"},{"start":{"date-parts":[[2026,11,1]],"date-time":"2026-11-01T00:00:00Z","timestamp":1793491200000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/legal\/tdmrep-license"},{"start":{"date-parts":[[2026,11,1]],"date-time":"2026-11-01T00:00:00Z","timestamp":1793491200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-017"},{"start":{"date-parts":[[2026,11,1]],"date-time":"2026-11-01T00:00:00Z","timestamp":1793491200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"},{"start":{"date-parts":[[2026,11,1]],"date-time":"2026-11-01T00:00:00Z","timestamp":1793491200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-012"},{"start":{"date-parts":[[2026,11,1]],"date-time":"2026-11-01T00:00:00Z","timestamp":1793491200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,11,1]],"date-time":"2026-11-01T00:00:00Z","timestamp":1793491200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-004"}],"funder":[{"DOI":"10.13039\/501100013091","name":"Science and Technology Major Project of Guangxi","doi-asserted-by":"publisher","award":["AA22068057"],"award-info":[{"award-number":["AA22068057"]}],"id":[{"id":"10.13039\/501100013091","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100021178","name":"University of Nottingham Ningbo China","doi-asserted-by":"publisher","award":["NDPT2408"],"award-info":[{"award-number":["NDPT2408"]}],"id":[{"id":"10.13039\/501100021178","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["elsevier.com","sciencedirect.com"],"crossmark-restriction":true},"short-container-title":["Information Sciences"],"published-print":{"date-parts":[[2026,11]]},"DOI":"10.1016\/j.ins.2026.123768","type":"journal-article","created":{"date-parts":[[2026,6,12]],"date-time":"2026-06-12T00:01:11Z","timestamp":1781222471000},"page":"123768","update-policy":"https:\/\/doi.org\/10.1016\/elsevier_cm_policy","source":"Crossref","is-referenced-by-count":0,"special_numbering":"C","title":["Low-rank sparse autoencoders: Unifying efficiency and geometric regularization for large language model interpretability"],"prefix":"10.1016","volume":"755","author":[{"given":"Jiajia","family":"Mu","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9121-8499","authenticated-orcid":false,"given":"Benying","family":"Tan","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jie","family":"Ren","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mingyu","family":"Guo","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chenchen","family":"Luo","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ruibin","family":"Bai","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","reference":[{"key":"10.1016\/j.ins.2026.123768_bib0005","doi-asserted-by":"crossref","DOI":"10.1016\/j.ins.2026.123273","article-title":"LDCL: large language model-driven dual-view contrastive learning for temporal knowledge graph completion","volume":"741","author":"Zhang","year":"2026","journal-title":"Inf. Sci.","ISSN":"https:\/\/id.crossref.org\/issn\/0020-0255","issn-type":"print"},{"key":"10.1016\/j.ins.2026.123768_bib0010","doi-asserted-by":"crossref","DOI":"10.1016\/j.ins.2025.122460","article-title":"Hierarchical interpretable vision reasoning driven based depth estimation method in adverse weather conditions through a multi-modal large language model","volume":"719","author":"Mi","year":"2025","journal-title":"Inf. Sci.","ISSN":"https:\/\/id.crossref.org\/issn\/0020-0255","issn-type":"print"},{"key":"10.1016\/j.ins.2026.123768_bib0015","series-title":"Proceedings of the 39th AAAI Conference on Artificial Intelligence","first-page":"24086","article-title":"Knowledge in superposition: unveiling the failures of lifelong knowledge editing for large language models","author":"Hu","year":"2025"},{"key":"10.1016\/j.ins.2026.123768_bib0020","article-title":"Scaling monosemanticity: extracting interpretable features from claude 3 sonnet","author":"Templeton","year":"2024","journal-title":"Transform. Circuits Thread"},{"key":"10.1016\/j.ins.2026.123768_bib0025","series-title":"The 7th BlackboxNLP Workshop","article-title":"Gemma scope: open sparse autoencoders everywhere all at once on gemma 2","author":"Lieberum","year":"2024"},{"key":"10.1016\/j.ins.2026.123768_bib0030","series-title":"The Twelfth International Conference on Learning Representations","article-title":"Sparse autoencoders find highly interpretable features in language models","author":"Huben","year":"2024"},{"key":"10.1016\/j.ins.2026.123768_bib0035","article-title":"Towards monosemanticity: decomposing language models with dictionary learning","author":"Bricken","year":"2023","journal-title":"Transform. Circuits Thread"},{"key":"10.1016\/j.ins.2026.123768_bib0040","article-title":"Mechanistic interpretability for AI safety - a review","author":"Bereska","year":"2024","journal-title":"Trans. Mach. Learn. Res.","ISSN":"https:\/\/id.crossref.org\/issn\/2835-8856","issn-type":"print"},{"key":"10.1016\/j.ins.2026.123768_bib0045","series-title":"The Thirteenth International Conference on Learning Representations","article-title":"Sparse feature circuits: discovering and editing interpretable causal graphs in language models","author":"Marks","year":"2025"},{"key":"10.1016\/j.ins.2026.123768_bib0050","series-title":"The Thirteenth International Conference on Learning Representations","article-title":"Scaling and evaluating sparse autoencoders","author":"Gao","year":"2025"},{"key":"10.1016\/j.ins.2026.123768_bib0055","series-title":"The Thirteenth International Conference on Learning Representations","article-title":"Efficient dictionary learning with switch sparse autoencoders","author":"Mudide","year":"2025"},{"key":"10.1016\/j.ins.2026.123768_bib0060","series-title":"The Fourteenth International Conference on Learning Representations","article-title":"AbsTopK: rethinking sparse autoencoders for bidirectional features","year":"2026"},{"key":"10.1016\/j.ins.2026.123768_bib0065","series-title":"The Thirteenth International Conference on Learning Representations","article-title":"SSOLE: rethinking orthogonal low-rank embedding for self-supervised learning","author":"Huang","year":"2025"},{"key":"10.1016\/j.ins.2026.123768_bib0070","series-title":"Proceedings of the 30th International Conference on Machine Learning (ICML) Workshop on Deep Learning for Audio, Speech, and Language Processing","article-title":"Rectifier nonlinearities improve neural network acoustic models","author":"Maas","year":"2013"},{"issue":"1","key":"10.1016\/j.ins.2026.123768_bib0075","doi-asserted-by":"crossref","first-page":"21","DOI":"10.1038\/s44387-026-00072-8","article-title":"Phase transitions in large language model compression","volume":"2","author":"Ma","year":"2026","journal-title":"npj Artif. Intell."},{"key":"10.1016\/j.ins.2026.123768_bib0080","series-title":"International Conference on Machine Learning","first-page":"2397","article-title":"Pythia: a suite for analyzing large language models across training and scaling","author":"Biderman","year":"2023"},{"key":"10.1016\/j.ins.2026.123768_bib0085","author":"Team"},{"key":"10.1016\/j.ins.2026.123768_bib0090","author":"Team"},{"key":"10.1016\/j.ins.2026.123768_bib0095","doi-asserted-by":"crossref","DOI":"10.1016\/j.ocecoaman.2025.108015","article-title":"Automated knowledge extraction from marine accident reports using large language models: graph construction and evaluation","volume":"272","author":"Huang","year":"2026","journal-title":"Ocean Coast. Manag.","ISSN":"https:\/\/id.crossref.org\/issn\/0964-5691","issn-type":"print"},{"key":"10.1016\/j.ins.2026.123768_bib0100","series-title":"NeurIPS 2024 Workshop on Scientific Methods for Understanding Deep Learning","article-title":"BatchTopK sparse autoencoders","author":"Bussmann","year":"2024"},{"key":"10.1016\/j.ins.2026.123768_bib0105","series-title":"NeurIPS","article-title":"Improving sparse decomposition of language model activations with gated sparse autoencoders","author":"Rajamanoharan","year":"2024"},{"key":"10.1016\/j.ins.2026.123768_bib0110","author":"Rajamanoharan"},{"key":"10.1016\/j.ins.2026.123768_bib0115","series-title":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","first-page":"6812","article-title":"Route sparse autoencoder to interpret large language models","author":"Shi","year":"2025"},{"key":"10.1016\/j.ins.2026.123768_bib0120","doi-asserted-by":"crossref","DOI":"10.1016\/j.knosys.2026.115277","article-title":"Leveraging fine-tuning of large language models for aspect-based sentiment analysis in resource-scarce environments","volume":"336","author":"Fehle","year":"2026","journal-title":"Knowl.-based Syst.","ISSN":"https:\/\/id.crossref.org\/issn\/0950-7051","issn-type":"print"},{"key":"10.1016\/j.ins.2026.123768_bib0125","series-title":"Forty-Second International Conference on Machine Learning","article-title":"Low-rank adapting models for sparse autoencoders","author":"Chen","year":"2025"},{"issue":"1","key":"10.1016\/j.ins.2026.123768_bib0130","doi-asserted-by":"crossref","first-page":"126","DOI":"10.1109\/MSP.2017.2765695","article-title":"Model compression and acceleration for deep neural networks: the principles, progress, and challenges","volume":"35","author":"Cheng","year":"2018","journal-title":"IEEE Signal Process. Mag."},{"key":"10.1016\/j.ins.2026.123768_bib0135","article-title":"Exploiting linear structure within convolutional networks for efficient evaluation","volume":"27","author":"Denton","year":"2014","journal-title":"Adv. Neural Inf. Process. Syst."},{"issue":"1","key":"10.1016\/j.ins.2026.123768_bib0140","doi-asserted-by":"crossref","first-page":"171","DOI":"10.1109\/TPAMI.2012.88","article-title":"Robust recovery of subspace structures by low-rank representation","volume":"35","author":"Liu","year":"2012","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"10.1016\/j.ins.2026.123768_bib0145","doi-asserted-by":"crossref","first-page":"360","DOI":"10.1016\/j.neucom.2018.06.010","article-title":"Low-rank structure preserving for unsupervised feature selection","volume":"314","author":"Zheng","year":"2018","journal-title":"Neurocomputing"},{"key":"10.1016\/j.ins.2026.123768_bib0150","series-title":"Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)","first-page":"7319","article-title":"Intrinsic dimensionality explains the effectiveness of language model fine-tuning","author":"Aghajanyan","year":"2021"},{"key":"10.1016\/j.ins.2026.123768_bib0155","series-title":"Proceedings of the 41st International Conference on Machine Learning, Volume 235 of Proceedings of Machine Learning Research","first-page":"39643","article-title":"The linear representation hypothesis and the geometry of large language models","author":"Park","year":"2024"},{"key":"10.1016\/j.ins.2026.123768_bib0160","series-title":"ACL 2019-57th Annual Meeting of the Association for Computational Linguistics","article-title":"What does BERT learn about the structure of language?","author":"Jawahar","year":"2019"},{"key":"10.1016\/j.ins.2026.123768_bib0165","doi-asserted-by":"crossref","DOI":"10.1016\/j.knosys.2023.110273","article-title":"Explainable AI (XAI): a systematic meta-survey of current challenges and future opportunities","volume":"263","author":"Saeed","year":"2023","journal-title":"Knowl.-based Syst.","ISSN":"https:\/\/id.crossref.org\/issn\/0950-7051","issn-type":"print"},{"key":"10.1016\/j.ins.2026.123768_bib0170","series-title":"International Conference on Machine Learning","first-page":"9929","article-title":"Understanding contrastive representation learning through alignment and uniformity on the hypersphere","author":"Wang","year":"2020"},{"key":"10.1016\/j.ins.2026.123768_bib0175","series-title":"Advances in Neural Information Processing Systems","article-title":"Implicit regularization in deep matrix factorization","volume":"vol. 32","author":"Arora","year":"2019"},{"issue":"1","key":"10.1016\/j.ins.2026.123768_bib0180","first-page":"12","article-title":"A mathematical framework for transformer circuits","volume":"1","author":"Elhage","year":"2021","journal-title":"Transform. Circuits Thread"},{"key":"10.1016\/j.ins.2026.123768_bib0185","article-title":"A mathematical framework for transformer circuits","author":"Elhage","year":"2021","journal-title":"Transform. Circuits Thread"},{"key":"10.1016\/j.ins.2026.123768_bib0190","series-title":"International Conference on Machine Learning","article-title":"Rectified linear units improve restricted Boltzmann machines","author":"Nair","year":"2010"},{"key":"10.1016\/j.ins.2026.123768_bib0195","series-title":"Gaussian error linear units (GELUs). arXiv: learning","author":"Hendrycks","year":"2016"},{"key":"10.1016\/j.ins.2026.123768_bib0200","series-title":"Searching for activation functions","author":"Ramachandran","year":"2018"},{"key":"10.1016\/j.ins.2026.123768_bib0205","series-title":"Proceedings of the British Machine Vision Conference 2020","article-title":"Mish: a self regularized non-monotonic activation function","author":"Misra","year":"2020"},{"key":"10.1016\/j.ins.2026.123768_bib0210","author":"Fernandez"},{"key":"10.1016\/j.ins.2026.123768_bib0215","author":"Gao"},{"key":"10.1016\/j.ins.2026.123768_bib0220","article-title":"Open source automated interpretability for sparse autoencoder features","volume":"30","author":"Caden Juang","year":"2024","journal-title":"EleutherAI Blog"},{"issue":"3","key":"10.1016\/j.ins.2026.123768_bib0225","doi-asserted-by":"crossref","DOI":"10.23915\/distill.00024.001","article-title":"Zoom in: an introduction to circuits","volume":"5","author":"Olah","year":"2020","journal-title":"Distill"},{"key":"10.1016\/j.ins.2026.123768_bib0230","article-title":"Finding neurons in a haystack: case studies with sparse probing","author":"Gurnee","year":"2023","journal-title":"Trans. Mach. Learn. Res.","ISSN":"https:\/\/id.crossref.org\/issn\/2835-8856","issn-type":"print"},{"key":"10.1016\/j.ins.2026.123768_bib0235","series-title":"Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing","first-page":"5484","article-title":"Transformer feed-forward layers are key-value memories","author":"Geva","year":"2021"},{"key":"10.1016\/j.ins.2026.123768_bib0240","series-title":"Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","first-page":"8493","article-title":"Knowledge neurons in pretrained transformers","author":"Dai","year":"2022"}],"container-title":["Information Sciences"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S0020025526006997?httpAccept=text\/xml","content-type":"text\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S0020025526006997?httpAccept=text\/plain","content-type":"text\/plain","content-version":"vor","intended-application":"text-mining"}],"deposited":{"date-parts":[[2026,7,17]],"date-time":"2026-07-17T17:03:31Z","timestamp":1784307811000},"score":1,"resource":{"primary":{"URL":"https:\/\/linkinghub.elsevier.com\/retrieve\/pii\/S0020025526006997"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,11]]},"references-count":48,"alternative-id":["S0020025526006997"],"URL":"https:\/\/doi.org\/10.1016\/j.ins.2026.123768","relation":{},"ISSN":["0020-0255"],"issn-type":[{"value":"0020-0255","type":"print"}],"subject":[],"published":{"date-parts":[[2026,11]]},"assertion":[{"value":"Elsevier","name":"publisher","label":"This article is maintained by"},{"value":"Low-rank sparse autoencoders: Unifying efficiency and geometric regularization for large language model interpretability","name":"articletitle","label":"Article Title"},{"value":"Information Sciences","name":"journaltitle","label":"Journal Title"},{"value":"https:\/\/doi.org\/10.1016\/j.ins.2026.123768","name":"articlelink","label":"CrossRef DOI link to publisher maintained version"},{"value":"article","name":"content_type","label":"Content Type"},{"value":"\u00a9 2026 Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies.","name":"copyright","label":"Copyright"}],"article-number":"123768"}}