{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,13]],"date-time":"2026-01-13T05:14:44Z","timestamp":1768281284523,"version":"3.49.0"},"reference-count":67,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"2","license":[{"start":{"date-parts":[[2026,2,1]],"date-time":"2026-02-01T00:00:00Z","timestamp":1769904000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,2,1]],"date-time":"2026-02-01T00:00:00Z","timestamp":1769904000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,2,1]],"date-time":"2026-02-01T00:00:00Z","timestamp":1769904000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62225113"],"award-info":[{"award-number":["62225113"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62576364"],"award-info":[{"award-number":["62576364"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["U23A20318"],"award-info":[{"award-number":["U23A20318"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["U2336211"],"award-info":[{"award-number":["U2336211"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62276195"],"award-info":[{"award-number":["62276195"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"name":"National Key Research and Development Program of China","award":["2023YFC2705700"],"award-info":[{"award-number":["2023YFC2705700"]}]},{"name":"Shenzhen Basic Research Project Natural Science Foundation","award":["JCYJ20241202124430041"],"award-info":[{"award-number":["JCYJ20241202124430041"]}]},{"name":"Science and Technology Major Project of Hubei Province","award":["2024BAB046"],"award-info":[{"award-number":["2024BAB046"]}]},{"DOI":"10.13039\/501100012556","name":"Foundation for Innovative Research Groups of Hubei Province","doi-asserted-by":"publisher","award":["2024AFA017"],"award-info":[{"award-number":["2024AFA017"]}],"id":[{"id":"10.13039\/501100012556","id-type":"DOI","asserted-by":"publisher"}]},{"name":"NTU RSR"},{"name":"Start Up Grants"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Pattern Anal. Mach. Intell."],"published-print":{"date-parts":[[2026,2]]},"DOI":"10.1109\/tpami.2025.3612480","type":"journal-article","created":{"date-parts":[[2025,9,22]],"date-time":"2025-09-22T17:44:23Z","timestamp":1758563063000},"page":"1145-1157","source":"Crossref","is-referenced-by-count":0,"title":["Zero-Shot Sparse Mixture of Low-Rank Experts Construction From Pre-Trained Foundation Models"],"prefix":"10.1109","volume":"48","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-0576-8153","authenticated-orcid":false,"given":"Anke","family":"Tang","sequence":"first","affiliation":[{"name":"School of Computer Science, National Engineering Research Center for Multimedia Software and Hubei Key Laboratory of Multimedia and Network Communication Engineering, Wuhan University, Wuhan, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5659-3464","authenticated-orcid":false,"given":"Li","family":"Shen","sequence":"additional","affiliation":[{"name":"Shenzhen Campus of Sun Yat-sen University, Shenzhen, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2296-6370","authenticated-orcid":false,"given":"Yong","family":"Luo","sequence":"additional","affiliation":[{"name":"School of Computer Science, National Engineering Research Center for Multimedia Software and Hubei Key Laboratory of Multimedia and Network Communication Engineering, Wuhan University, Wuhan, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Shuai","family":"Xie","sequence":"additional","affiliation":[{"name":"JD Explore Academy, Beijing, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7532-0496","authenticated-orcid":false,"given":"Han","family":"Hu","sequence":"additional","affiliation":[{"name":"School of Information and Electronics, Beijing Institute of Technology, Beijing, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0542-2280","authenticated-orcid":false,"given":"Lefei","family":"Zhang","sequence":"additional","affiliation":[{"name":"School of Computer Science, National Engineering Research Center for Multimedia Software and Hubei Key Laboratory of Multimedia and Network Communication Engineering, Wuhan University, Wuhan, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0059-8458","authenticated-orcid":false,"given":"Bo","family":"Du","sequence":"additional","affiliation":[{"name":"School of Computer Science, National Engineering Research Center for Multimedia Software and Hubei Key Laboratory of Multimedia and Network Communication Engineering, Wuhan University, Wuhan, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Dacheng","family":"Tao","sequence":"additional","affiliation":[{"name":"Nanyang Technological University, Singapore"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Large language models: A survey","author":"Minaee","year":"2024"},{"issue":"3","key":"ref2","first-page":"1","article-title":"Large language models: A comprehensive survey of its applications, challenges, limitations, and future prospects","volume":"1","author":"Hadi","year":"2023","journal-title":"Authorea Preprints"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/tnnls.2025.3628666"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1038\/s42256-024-00961-0"},{"key":"ref5","article-title":"Model merging in LLMs, MLLMs, and beyond: Methods, theories, applications and opportunities","author":"Yang","year":"2024"},{"key":"ref6","article-title":"Fusionbench: A comprehensive benchmark of deep model fusion","author":"Tang","year":"2024"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1002\/widm.1249"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1016\/j.patcog.2021.107852"},{"key":"ref9","first-page":"1","article-title":"Knowledge fusion of large language models","volume-title":"Proc. 12th Int. Conf. Learn. Representations","author":"Wan","year":"2024"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.1096"},{"key":"ref11","first-page":"17703","article-title":"Merging models with fisher-weighted averaging","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"35","author":"Matena","year":"2022"},{"key":"ref12","first-page":"1","article-title":"Dataless knowledge fusion by merging weights of language models","volume-title":"Proc. 11th Int. Conf. Learn. Representations","author":"Jin","year":"2023"},{"key":"ref13","first-page":"1","article-title":"Sparse upcycling: Training mixture-of-experts from dense checkpoints","volume-title":"Proc. 11th Int. Conf. Learn. Representations","author":"Komatsuzaki","year":"2023"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.naacl-industry.3"},{"key":"ref15","article-title":"Git re-basin: Merging models modulo permutation symmetries","volume-title":"Proc. 11th Int. Conf. Learn. Representations","author":"Ainsworth","year":"2023"},{"key":"ref16","first-page":"1","article-title":"Zipit! Merging models from different tasks without training","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Stoica","year":"2024"},{"key":"ref17","first-page":"7093","article-title":"Ties-merging: Resolving interference when merging models","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"36","author":"Yadav","year":"2023"},{"key":"ref18","first-page":"1","article-title":"Language models are super mario: Absorbing abilities from homologous models as a free lunch","volume-title":"Proc. 41st Int. Conf. Mach. Learn.","author":"Yu","year":"2024"},{"key":"ref19","first-page":"1","article-title":"LoRA: Low-rank adaptation of large language models","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Hu","year":"2022"},{"key":"ref20","volume-title":"Appl. Linear Algebra","author":"Olver","year":"2018"},{"key":"ref21","first-page":"13579","article-title":"BitDelta: Your fine-tune may only be worth one bit","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"37","author":"Liu","year":"2024"},{"key":"ref22","first-page":"8571","article-title":"Neural tangent kernel: Convergence and generalization in neural networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"31","author":"Jacot","year":"2018"},{"key":"ref23","first-page":"1","article-title":"Parameter-efficient multi-task model fusion with partial linearization","volume-title":"Proc. 12th Int. Conf. Learn. Representations","author":"Tang","year":"2024"},{"key":"ref24","first-page":"38885","article-title":"Towards modular LLMs by building and reusing a library of LoRAs","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Ostapenko","year":"2024"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2010.5539970"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/ICCVW.2013.77"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2017.2675998"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/JSTARS.2019.2918242"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.2118\/18761-MS"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2012.02.016"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/5.726791"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2014.461"},{"key":"ref33","first-page":"23965","article-title":"Model soups: Averaging weights of multiple fine-tuned models improves accuracy without increasing inference time","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Wortsman","year":"2022"},{"key":"ref34","article-title":"Editing models with task arithmetic","volume-title":"Proc. 11th Int. Conf. Learn. Representations","author":"Ilharco","year":"2023"},{"key":"ref35","article-title":"AdaMerging: Adaptive model merging for multi-task learning","volume-title":"Proc. 12th Int. Conf. Learn. Representations","author":"Yang","year":"2024"},{"key":"ref36","article-title":"Representation surgery for multi-task model merging","volume-title":"Proc. 41st Int. Conf. Mach. Learn.","author":"Yang","year":"2024"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.01742"},{"key":"ref38","article-title":"No task left behind: Isotropic model merging with common and task-specific subspaces","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Marczak","year":"2025"},{"key":"ref39","first-page":"47778","article-title":"Merging multi-task models via weight-ensembling mixture of experts","volume-title":"Proc. 41st Int. Conf. Mach. Learn.","author":"Tang","year":"2024"},{"key":"ref40","doi-asserted-by":"crossref","first-page":"353","DOI":"10.18653\/v1\/W18-5446","article-title":"GLUE: A multi-task benchmark and analysis platform for natural language understanding","volume-title":"Proc. 2018 EMNLP Workshop BlackboxNLP: Analyzing Interpreting Neural Netw. NLP","author":"Wang","year":"2018"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.emnlp-demos.6"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-industry.36"},{"key":"ref43","first-page":"8","article-title":"A framework for few-shot language model evaluation","author":"Gao","year":"2021"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1162\/neco.1991.3.1.79"},{"key":"ref45","article-title":"Mixtral of experts","author":"Jiang","year":"2024"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.70"},{"issue":"120","key":"ref47","first-page":"1","article-title":"Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity","volume":"23","author":"Fedus","year":"2022","journal-title":"J. Mach. Learn. Res."},{"key":"ref48","first-page":"6265","article-title":"Base layers: Simplifying training of large, sparse models","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Lewis","year":"2021"},{"key":"ref49","article-title":"A review of sparse expert models in deep learning","author":"Fedus","year":"2022"},{"key":"ref50","article-title":"A survey on model moerging: Recycling and routing among specialized experts for collaborative learning","author":"Yadav","year":"2024","journal-title":"Trans. Mach. Learn. Res."},{"key":"ref51","article-title":"Topology and geometry of half-rectified network optimization","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Freeman","year":"2017"},{"key":"ref52","first-page":"11615","article-title":"Uniform convergence may be unable to explain generalization in deep learning","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"32","author":"Nagarajan","year":"2019"},{"key":"ref53","first-page":"1309","article-title":"Essentially no barriers in neural network energy landscape","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Draxler","year":"2018"},{"key":"ref54","first-page":"3259","article-title":"Linear mode connectivity and the lottery ticket hypothesis","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Frankle","year":"2020"},{"key":"ref55","article-title":"The role of permutation invariance in linear mode connectivity of neural networks","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Entezari","year":"2022"},{"key":"ref56","first-page":"8789","article-title":"Loss surfaces, mode connectivity, and fast ensembling of DNNs","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"31","author":"Garipov","year":"2018"},{"key":"ref57","first-page":"15300","article-title":"Optimizing mode connectivity via neuron alignment","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"33","author":"Tatro","year":"2020"},{"key":"ref58","first-page":"1","article-title":"On convexity and linear mode connectivity in neural networks","volume-title":"Proc. OPT 2022: Optim. Mach. Learn.","author":"Yunis","year":"2022"},{"key":"ref59","first-page":"769","article-title":"Loss surface simplexes for mode connecting volumes and fast ensembling","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Benton","year":"2021"},{"key":"ref60","first-page":"876","article-title":"Averaging weights leads to wider optima and better generalization","volume-title":"Proc. 34th Conf. Uncertainty Artif. Intell.","author":"Izmailov","year":"2018"},{"key":"ref61","article-title":"Stop wasting my time! Saving days of imagenet and bert training with latest weight averaging","volume-title":"Proc. Adv. Neural Inf. Process. Syst. Workshop","author":"Kaddour","year":"2022"},{"key":"ref62","first-page":"37713","article-title":"$\\pi$\u03c0-tuning: Transferring multimodal foundation models with optimal multi-task interpolation","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Wu","year":"2023"},{"key":"ref63","first-page":"196","article-title":"Convergent learning: Do different neural networks learn the same representations?","author":"Li","year":"2015","journal-title":"Feature Extraction: Modern Questions Challenges, Proc. Mach. Learn. Res."},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00565"},{"key":"ref65","first-page":"13857","article-title":"Deep neural network fusion via graph matching with applications to model ensemble and federated learning","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Liu","year":"2022"},{"key":"ref66","first-page":"78905","article-title":"Twin-merging: Dynamic integration of modular expertise in model merging","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"37","author":"Lu","year":"2024"},{"key":"ref67","article-title":"Towards efficient pareto set approximation via mixture of experts based model fusion","author":"Tang","year":"2024"}],"container-title":["IEEE Transactions on Pattern Analysis and Machine Intelligence"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/34\/11345188\/11175033.pdf?arnumber=11175033","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,12]],"date-time":"2026-01-12T22:00:47Z","timestamp":1768255247000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11175033\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2]]},"references-count":67,"journal-issue":{"issue":"2"},"URL":"https:\/\/doi.org\/10.1109\/tpami.2025.3612480","relation":{},"ISSN":["0162-8828","2160-9292","1939-3539"],"issn-type":[{"value":"0162-8828","type":"print"},{"value":"2160-9292","type":"electronic"},{"value":"1939-3539","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2]]}}}