{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T16:14:36Z","timestamp":1782836076118,"version":"3.54.5"},"reference-count":13,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,7,2]],"date-time":"2025-07-02T00:00:00Z","timestamp":1751414400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,7,2]],"date-time":"2025-07-02T00:00:00Z","timestamp":1751414400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100014219","name":"National Science Fund for Distinguished Young Scholars of China","doi-asserted-by":"publisher","award":["62125203"],"award-info":[{"award-number":["62125203"]}],"id":[{"id":"10.13039\/501100014219","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100004608","name":"Natural Science Foundation of Jiangsu Province","doi-asserted-by":"publisher","award":["BG2024028"],"award-info":[{"award-number":["BG2024028"]}],"id":[{"id":"10.13039\/501100004608","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["BK20243053,BK20230351,62402233,62427809,62472232,U24A20248,2024YFB2906704"],"award-info":[{"award-number":["BK20243053,BK20230351,62402233,62427809,62472232,U24A20248,2024YFB2906704"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,7,2]]},"DOI":"10.1109\/iwqos65803.2025.11143464","type":"proceedings-article","created":{"date-parts":[[2025,9,9]],"date-time":"2025-09-09T17:29:52Z","timestamp":1757438992000},"page":"1-6","source":"Crossref","is-referenced-by-count":1,"title":["Birds in Cages: Edge Inference Allocation for Distributed LLM Deployment"],"prefix":"10.1109","author":[{"given":"Jiahao","family":"Zhu","sequence":"first","affiliation":[{"name":"School of Computer Science, Nanjing University of Posts and Telecommunications"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lu","family":"Zhao","sequence":"additional","affiliation":[{"name":"School of Computer Science, Nanjing University of Posts and Telecommunications"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Fu","family":"Xiao","sequence":"additional","affiliation":[{"name":"School of Computer Science, Nanjing University of Posts and Telecommunications"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lingjie","family":"Duan","sequence":"additional","affiliation":[{"name":"Singapore University of Technology and Design,Engineering Systems and Design Pillar"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/MWC.004.2300479"},{"key":"ref2","article-title":"Fast distributed inference serving for large language models","author":"Wu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/twc.2024.3497923"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1145\/3636534.3649363"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1145\/3323679.3326505"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/TSC.2024.3353290"},{"key":"ref7","article-title":"Language models are few-shot learners","author":"Brown","year":"2020","journal-title":"arXiv preprint"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/TKDE.2024.3370614"},{"key":"ref9","article-title":"Llama 2: Open foundation and finetuned chat models","author":"Touvron","year":"2023","journal-title":"arXiv preprint"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/TMC.2022.3172402"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/TWC.2023.3257197"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.acl-demo.54"},{"key":"ref13","article-title":"Hexgen: Generative inference of large language model over heterogeneous environment","volume-title":"ICML","author":"Jiang","year":"2024"}],"event":{"name":"2025 IEEE\/ACM International Symposium on Quality of Service (IWQoS)","location":"Gold Coast, Australia","start":{"date-parts":[[2025,7,2]]},"end":{"date-parts":[[2025,7,4]]}},"container-title":["2025 IEEE\/ACM 33rd International Symposium on Quality of Service (IWQoS)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11143240\/11143247\/11143464.pdf?arnumber=11143464","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,9,10]],"date-time":"2025-09-10T05:09:29Z","timestamp":1757480969000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11143464\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,7,2]]},"references-count":13,"URL":"https:\/\/doi.org\/10.1109\/iwqos65803.2025.11143464","relation":{},"subject":[],"published":{"date-parts":[[2025,7,2]]}}}