{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,20]],"date-time":"2026-04-20T23:49:02Z","timestamp":1776728942391,"version":"3.51.2"},"reference-count":14,"publisher":"IEEE","license":[{"start":{"date-parts":[[2024,11,10]],"date-time":"2024-11-10T00:00:00Z","timestamp":1731196800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2024,11,10]],"date-time":"2024-11-10T00:00:00Z","timestamp":1731196800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/100014613","name":"Public Safety Canada","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100014613","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2024,11,10]]},"DOI":"10.1109\/wf-iot62078.2024.10811332","type":"proceedings-article","created":{"date-parts":[[2024,12,30]],"date-time":"2024-12-30T19:20:28Z","timestamp":1735586428000},"page":"1-6","source":"Crossref","is-referenced-by-count":0,"title":["Time-Division Online Speculative Inference for Cost-Effective Intelligent Internet of Things"],"prefix":"10.1109","author":[{"given":"Yao","family":"Du","sequence":"first","affiliation":[{"name":"The University of British Columbia,Department of Electrical and Computer Engineering,Vancouver,BC,Canada,V6T 1Z4"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zehua","family":"Wang","sequence":"additional","affiliation":[{"name":"The University of British Columbia,Department of Electrical and Computer Engineering,Vancouver,BC,Canada,V6T 1Z4"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Cyril","family":"Leung","sequence":"additional","affiliation":[{"name":"The University of British Columbia,Department of Electrical and Computer Engineering,Vancouver,BC,Canada,V6T 1Z4"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Victor C.M.","family":"Leung","sequence":"additional","affiliation":[{"name":"The University of British Columbia,Department of Electrical and Computer Engineering,Vancouver,BC,Canada,V6T 1Z4"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/MNET.2024.3422309"},{"key":"ref2","article-title":"Propile: Probing privacy leakage in large language models","volume":"36","author":"Kim","year":"2024","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref3","article-title":"Speculative decoding with big little decoder","volume":"36","author":"Kim","year":"2024","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref4","first-page":"19274","article-title":"Fast inference from transformers via speculative decoding","volume-title":"International Conference on Machine Learning","author":"Leviathan"},{"key":"ref5","first-page":"1","article-title":"Online speculative decoding","volume-title":"International Conference on Machine Learning","author":"Liu"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1016\/j.jnca.2017.11.011"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1007\/3-540-48184-2_32"},{"key":"ref8","article-title":"Response length perception and sequence scheduling: An llm-empowered llm inference pipeline","volume":"36","author":"Zheng","year":"2024","journal-title":"Advances in Neural Information Processing Systems"},{"issue":"1999","key":"ref9","first-page":"173","article-title":"Practical byzantine fault tolerance","volume":"99","author":"Castro","year":"1999","journal-title":"OsDI"},{"issue":"140","key":"ref10","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":"ref11","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/W16-2301"},{"key":"ref12","article-title":"Huggingface\u2019s transformers: State-of-the-art natural language processing","author":"Wolf","year":"2019","journal-title":"arXiv preprint arXiv:1910.03771"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-020-2649-2"},{"key":"ref14","article-title":"M\u00e9lange: Cost efficient large language model serving by exploiting GPU heterogeneity","author":"Griggs","year":"2024","journal-title":"arXiv preprint arXiv:2404.14527"}],"event":{"name":"2024 IEEE 10th World Forum on Internet of Things (WF-IoT)","location":"Ottawa, ON, Canada","start":{"date-parts":[[2024,11,10]]},"end":{"date-parts":[[2024,11,13]]}},"container-title":["2024 IEEE 10th World Forum on Internet of Things (WF-IoT)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/10811010\/10811012\/10811332.pdf?arnumber=10811332","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,12,31]],"date-time":"2024-12-31T06:36:07Z","timestamp":1735626967000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10811332\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,11,10]]},"references-count":14,"URL":"https:\/\/doi.org\/10.1109\/wf-iot62078.2024.10811332","relation":{},"subject":[],"published":{"date-parts":[[2024,11,10]]}}}