{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T05:44:38Z","timestamp":1782798278452,"version":"3.54.5"},"reference-count":51,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,5,18]],"date-time":"2026-05-18T00:00:00Z","timestamp":1779062400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,5,18]],"date-time":"2026-05-18T00:00:00Z","timestamp":1779062400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/100006190","name":"Research and Development","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100006190","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","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","id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,5,18]]},"DOI":"10.1109\/infocom59046.2026.11571667","type":"proceedings-article","created":{"date-parts":[[2026,6,29]],"date-time":"2026-06-29T19:38:15Z","timestamp":1782761895000},"page":"1-10","source":"Crossref","is-referenced-by-count":0,"title":["Faster, Smaller, and Smarter: Task-Aware Expert Merging for Online MoE Inference"],"prefix":"10.1109","author":[{"given":"Ziyi","family":"Han","sequence":"first","affiliation":[{"name":"The Chinese University of Hong Kong"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xutong","family":"Liu","sequence":"additional","affiliation":[{"name":"University of Washington"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ruiting","family":"Zhou","sequence":"additional","affiliation":[{"name":"Southeast University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiangxiang","family":"Dai","sequence":"additional","affiliation":[{"name":"The Chinese University of Hong Kong"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"John C.S.","family":"Lui","sequence":"additional","affiliation":[{"name":"The Chinese University of Hong Kong"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.24818\/ida-ql\/2019.5"},{"issue":"140","key":"ref2","first-page":"1","article-title":"Exploring the limits of transfer learning with a unified text-to-text transformer","volume":"21","author":"Raffel","year":"2020","journal-title":"Journal of machine learning research"},{"key":"ref3","article-title":"Multi-agent conversational online learning for adaptive llm response identification","author":"Dai","year":"2025"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2670"},{"issue":"120","key":"ref5","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":"Journal of Machine Learning Research"},{"key":"ref6","article-title":"No language left behind: Scaling human-centered machine translation","author":"Costa-Jussa","year":"2022"},{"key":"ref7","article-title":"Experts weights averaging: A new general training scheme for vision transformers","author":"Huang","year":"2023"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM55648.2025.11044553"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1145\/3680207.3765696"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.findings-emnlp.304"},{"key":"ref11","article-title":"Soft merging of experts with adaptive routing","author":"Muqeeth","year":"2023"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.907"},{"key":"ref13","article-title":"Lory: Fully differentiable mixture-of-experts for autoregressive language model pre-training","author":"Zhong","year":"2024"},{"key":"ref14","article-title":"Merge, then compress: Demystify efficient SMoe with hints from its routing policy","volume-title":"Proc. of ICLR","author":"Li"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM55648.2025.11044567"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1016\/j.jalgor.2003.12.001"},{"key":"ref17","article-title":"Outrageously large neural networks: The sparsely-gated mixture-of-experts layer","author":"Shazeer","year":"2017"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1145\/3603269.3604869"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM53939.2023.10228874"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM55648.2025.11044508"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1145\/3787849"},{"key":"ref22","article-title":"Editing models with task arithmetic","volume-title":"Proc. of ICLR","author":"Ilharco"},{"key":"ref23","article-title":"Adamerging: Adaptive model merging for multi-task learning","volume-title":"Proc. of ICLR","author":"Yang"},{"key":"ref24","article-title":"Dataless knowledge fusion by merging weights of language models","volume-title":"Proc. of ICLR","author":"Jin"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1287"},{"key":"ref26","first-page":"7349","article-title":"Fisher mask nodes for language model merging","volume-title":"Proc. of LREC-COLING","author":"K"},{"key":"ref27","article-title":"Neural contextual bandits with deep representation and shallow exploration","volume-title":"Proc. of ICLR","author":"Xu"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0612"},{"key":"ref29","first-page":"11 492","article-title":"Neural contextual bandits with UCB-based exploration","volume-title":"Proc. of ICML","volume":"119","author":"Zhou"},{"key":"ref30","article-title":"Neural thompson sampling","volume-title":"Proc. of ICLR","author":"ZHANG"},{"key":"ref31","article-title":"EE-net: Exploitation-exploration neural networks in contextual bandits","volume-title":"Proc. of ICLR","author":"Ban"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1109\/TON.2025.3538709"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1162\/neco.1991.3.1.79"},{"key":"ref34","article-title":"Branch-train-mix: Mixing expert LLMs into a mixture-of-experts LLM","volume-title":"Proc. of COLM","author":"Sukhbaatar"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00355"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/TPDS.2024.3430063"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-540-24777-7_9"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1109\/TMC.2023.3333368"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM59046.2026.11571668"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1145\/3637528.3671440"},{"key":"ref41","article-title":"Cost-effective online multi-llm selection with versatile reward models","author":"Dai","year":"2024"},{"key":"ref42","first-page":"1557","article-title":"Online stochastic optimization under correlated bandit feedback","volume-title":"Proc. of ICML","author":"Lazaric"},{"key":"ref43","article-title":"Linear programming and extensions","author":"Dantzig","year":"2016"},{"key":"ref44","first-page":"208","article-title":"Contextual bandits with linear payoff functions","volume-title":"Proc. of AISTATS","author":"Chu"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/W18-5446"},{"key":"ref46","article-title":"Hilora: Adaptive hierarchical lora routing for training-free domain generalization","author":"Han","year":"2025"},{"key":"ref47","first-page":"4057","article-title":"Unified scaling laws for routed language models","volume-title":"Proc. of ICML","author":"Clark"},{"key":"ref48","first-page":"22 045","article-title":"Model fusion via optimal transport","volume-title":"Proc. of NeurIPS","volume":"33","author":"Singh"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v39i15.33692"},{"key":"ref50","article-title":"Neural tangent kernel: Convergence and generalization in neural networks","volume-title":"Proc. of NeurIPS","volume":"31","author":"Jacot"},{"key":"ref51","article-title":"Generalization bounds of stochastic gradient descent for wide and deep neural networks","volume-title":"Proc. of NeurIPS","volume":"32","author":"Cao"}],"event":{"name":"IEEE INFOCOM 2026 - IEEE Conference on Computer Communications","location":"Tokyo, Japan","start":{"date-parts":[[2026,5,18]]},"end":{"date-parts":[[2026,5,21]]}},"container-title":["IEEE INFOCOM 2026 - IEEE Conference on Computer Communications"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11571071\/11571169\/11571667.pdf?arnumber=11571667","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T05:13:20Z","timestamp":1782796400000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11571667\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,18]]},"references-count":51,"URL":"https:\/\/doi.org\/10.1109\/infocom59046.2026.11571667","relation":{},"subject":[],"published":{"date-parts":[[2026,5,18]]}}}