{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,21]],"date-time":"2026-07-21T14:20:48Z","timestamp":1784643648673,"version":"3.55.0"},"reference-count":51,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"6","license":[{"start":{"date-parts":[[2025,12,1]],"date-time":"2025-12-01T00:00:00Z","timestamp":1764547200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,12,1]],"date-time":"2025-12-01T00:00:00Z","timestamp":1764547200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,12,1]],"date-time":"2025-12-01T00:00:00Z","timestamp":1764547200000},"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":["62471383"],"award-info":[{"award-number":["62471383"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Hong Kong RGC General Research Fund","award":["152244\/21E"],"award-info":[{"award-number":["152244\/21E"]}]},{"name":"Hong Kong RGC General Research Fund","award":["152169\/22E"],"award-info":[{"award-number":["152169\/22E"]}]},{"name":"Hong Kong RGC General Research Fund","award":["152228\/23E"],"award-info":[{"award-number":["152228\/23E"]}]},{"name":"Hong Kong RGC General Research Fund","award":["162161\/24E"],"award-info":[{"award-number":["162161\/24E"]}]},{"name":"Research Impact Fund","award":["R5011-23F"],"award-info":[{"award-number":["R5011-23F"]}]},{"name":"Research Impact Fund","award":["R5060-19"],"award-info":[{"award-number":["R5060-19"]}]},{"name":"Collaborative Research Fund","award":["C1042-23GF"],"award-info":[{"award-number":["C1042-23GF"]}]},{"name":"NSFC\/RGC Collaborative Research Scheme","award":["CRS_HKUST602\/24"],"award-info":[{"award-number":["CRS_HKUST602\/24"]}]},{"name":"Areas of Excellence Scheme","award":["AoE\/E-601\/22-R"],"award-info":[{"award-number":["AoE\/E-601\/22-R"]}]},{"name":"InnoHK"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Netw."],"published-print":{"date-parts":[[2025,12]]},"DOI":"10.1109\/ton.2025.3585359","type":"journal-article","created":{"date-parts":[[2025,7,8]],"date-time":"2025-07-08T13:37:40Z","timestamp":1751981860000},"page":"2869-2880","source":"Crossref","is-referenced-by-count":5,"title":["Communication-Efficient Sparsely-Activated Model Training via Sequence Migration and Token Condensation"],"prefix":"10.1109","volume":"33","author":[{"given":"Fahao","family":"Chen","sequence":"first","affiliation":[{"name":"School of Artificial Intelligence, Shandong University, Jinan, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5303-0700","authenticated-orcid":false,"given":"Peng","family":"Li","sequence":"additional","affiliation":[{"name":"School of Cyber Science and Engineering, Xi’an Jiaotong University, Xi’an, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5689-382X","authenticated-orcid":false,"given":"Zicong","family":"Hong","sequence":"additional","affiliation":[{"name":"Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6518-3130","authenticated-orcid":false,"given":"Zhou","family":"Su","sequence":"additional","affiliation":[{"name":"School of Cyber Science and Engineering, Xi’an Jiaotong University, Xi’an, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9831-2202","authenticated-orcid":false,"given":"Song","family":"Guo","sequence":"additional","affiliation":[{"name":"Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref2","article-title":"BERT: Pre-training of deep bidirectional transformers for language understanding","author":"Devlin","year":"2018","journal-title":"arXiv:1810.04805"},{"issue":"8","key":"ref3","first-page":"9","article-title":"Language models are unsupervised multitask learners","volume":"1","author":"Radford","year":"2019","journal-title":"OpenAI Blog"},{"key":"ref4","article-title":"GShard: Scaling giant models with conditional computation and automatic sharding","author":"Lepikhin","year":"2020","journal-title":"arXiv:2006.16668"},{"issue":"1","key":"ref5","first-page":"5232","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":"ref6","first-page":"18332","article-title":"DeepSpeed-MoE: Advancing mixture-of-experts inference and training to power next-generation AI scale","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Rajbhandari"},{"key":"ref7","first-page":"269","article-title":"Tutel: Adaptive mixture-of-experts at scale","volume-title":"Proc. Mach. Learn. Syst.","author":"Hwang"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1145\/3588964"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1145\/3603269.3604869"},{"key":"ref10","first-page":"945","article-title":"Accelerating distributed MoE training and inference with lina","volume-title":"Proc. 2023 USENIX Annu. Tech. Conf. (USENIX ATC 23)","author":"Li"},{"key":"ref11","first-page":"13782","article-title":"Gating dropout: Communication-efficient regularization for sparsely activated transformers","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Liu"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1145\/3503221.3508418"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1145\/3503222.3507738"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA56546.2023.10071027"},{"key":"ref15","first-page":"8583","article-title":"Scaling vision with sparse mixture of experts","volume-title":"Proc. NIPS","author":"Riquelme"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/IPDPSW55747.2022.00171"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/SC41404.2022.00051"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1145\/3503221.3508417"},{"key":"ref19","first-page":"961","article-title":"SmartMoE: Efficiently training sparsely-activated models through combining offline and online parallelization","volume-title":"Proc. USENIX Annu. Tech. Conf. (USENIX ATC 23)","author":"Zhai"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P19-1285"},{"key":"ref21","article-title":"Mixtral of experts","author":"Jiang","year":"2024","journal-title":"arXiv:2401.04088"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00008"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P19-2057"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.acl-short.45"},{"key":"ref25","article-title":"PanGu-\u03c0: Enhancing language model architectures via nonlinearity compensation","author":"Wang","year":"2023","journal-title":"arXiv:2312.17276"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1145\/3545008.3545052"},{"key":"ref27","first-page":"16344","article-title":"FlashAttention: Fast and memory-efficient exact attention with IO-awareness","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Dao"},{"key":"ref28","first-page":"8026","article-title":"PyTorch: An imperative style, high-performance deep learning library","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Paszke"},{"key":"ref29","first-page":"1","article-title":"Deep graph library: Towards efficient and scalable deep learning on graphs","volume-title":"Proc. ICLR Workshop Represent. Learn. Graphs Manifolds","author":"Wang"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/TPDS.2021.3125565"},{"key":"ref31","first-page":"103","article-title":"BGL: GPU-efficient GNN training by optimizing graph data I\/O and preprocessing","volume-title":"Proc. USENIX Symp. Netw. Syst. Design Implement.","author":"Liu"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1145\/3626724"},{"key":"ref33","first-page":"1","article-title":"Pointer sentinel mixture models","volume-title":"Proc. Int. Conf. Learn. Represent.","author":"Merity"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D16-1264"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D19-5409"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1145\/3580305.3599521"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM52122.2024.10621347"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1145\/3394486.3403359"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.01588"},{"key":"ref40","first-page":"32897","article-title":"VLMo: Unified vision-language pre-training with mixture-of-modality-experts","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Bao"},{"issue":"240","key":"ref41","first-page":"1","article-title":"PaLM: Scaling language modeling with pathways","volume":"24","author":"Chowdhery","year":"2023","journal-title":"J. Mach. Learn. Res."},{"key":"ref42","first-page":"5547","article-title":"GLaM: Efficient scaling of language models with mixture-of-experts","volume-title":"Proc. Int. Conf. Mach. Learn.","volume":"162","author":"Du"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.890"},{"key":"ref44","article-title":"OpenMoE: An early effort on open mixture-of-experts language models","author":"Xue","year":"2024","journal-title":"arXiv:2402.01739"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.70"},{"key":"ref46","article-title":"GSPMD: General and scalable parallelization for ML computation graphs","author":"Xu","year":"2021","journal-title":"arXiv:2105.04663"},{"key":"ref47","first-page":"6265","article-title":"BASE layers: Simplifying training of large, sparse models","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Lewis"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/N19-4009"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM53939.2023.10228874"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/IPDPS54959.2023.00026"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1145\/3627703.3650083"}],"container-title":["IEEE Transactions on Networking"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/10723154\/11311125\/11072843.pdf?arnumber=11072843","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,12,23]],"date-time":"2025-12-23T06:17:39Z","timestamp":1766470659000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11072843\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,12]]},"references-count":51,"journal-issue":{"issue":"6"},"URL":"https:\/\/doi.org\/10.1109\/ton.2025.3585359","relation":{},"ISSN":["2998-4157"],"issn-type":[{"value":"2998-4157","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,12]]}}}