{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,13]],"date-time":"2026-01-13T07:54:06Z","timestamp":1768290846161,"version":"3.49.0"},"reference-count":43,"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":["62401115"],"award-info":[{"award-number":["62401115"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. on Mobile Comput."],"published-print":{"date-parts":[[2026,2]]},"DOI":"10.1109\/tmc.2025.3608303","type":"journal-article","created":{"date-parts":[[2025,9,10]],"date-time":"2025-09-10T17:49:12Z","timestamp":1757526552000},"page":"2691-2706","source":"Crossref","is-referenced-by-count":1,"title":["Joint Inference Offloading and Model Caching for Small and Large Language Model Collaboration"],"prefix":"10.1109","volume":"25","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-6916-5629","authenticated-orcid":false,"given":"Xinyi","family":"Xu","sequence":"first","affiliation":[{"name":"National Key Laboratory of Wireless Communications, University of Electronic Science and Technology of China, Chengdu, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2512-2392","authenticated-orcid":false,"given":"Gang","family":"Feng","sequence":"additional","affiliation":[{"name":"National Key Laboratory of Wireless Communications, University of Electronic Science and Technology of China, Chengdu, China"}]},{"ORCID":"https:\/\/orcid.org\/0009-0006-3019-5795","authenticated-orcid":false,"given":"Yijing","family":"Liu","sequence":"additional","affiliation":[{"name":"National Key Laboratory of Wireless Communications, University of Electronic Science and Technology of China, Chengdu, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9264-6850","authenticated-orcid":false,"given":"Shuang","family":"Qin","sequence":"additional","affiliation":[{"name":"National Key Laboratory of Wireless Communications, University of Electronic Science and Technology of China, Chengdu, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1667-9516","authenticated-orcid":false,"given":"Jian","family":"Wang","sequence":"additional","affiliation":[{"name":"National Key Laboratory of Wireless Communications, University of Electronic Science and Technology of China, Chengdu, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6141-810X","authenticated-orcid":false,"given":"Yunxiang","family":"Wang","sequence":"additional","affiliation":[{"name":"National Key Laboratory of Wireless Communications, University of Electronic Science and Technology of China, Chengdu, China"}]}],"member":"263","reference":[{"key":"ref1","article-title":"GPT-4 technical report","author":"Achiam","year":"2023"},{"key":"ref2","article-title":"Llama 2: Open foundation and fine-tuned chat models","author":"Touvron","year":"2023"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/mvt.2023.3323757"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1109\/COMST.2024.3353265"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1109\/JIOT.2016.2579198"},{"key":"ref6","article-title":"BloombergGPT: A large language model for finance","author":"Wu","year":"2023"},{"key":"ref7","article-title":"Model compression and efficient inference for large language models: A survey","author":"Wang","year":"2024"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/jiot.2024.3524255"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.acl-main.195"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v39i23.34620"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.18"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i20.30214"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/TMC.2024.3415661"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/globecom52923.2024.10901312"},{"key":"ref15","article-title":"M\u00e9lange: Cost efficient large language model serving by exploiting GPU heterogeneity","author":"Griggs","year":"2024"},{"key":"ref16","first-page":"173","article-title":"Llumnix: Dynamic scheduling for large language model serving","author":"Sun","year":"2024","journal-title":"Proc. 18th USENIX Symp. Oper. Syst. Design Implementation"},{"key":"ref17","article-title":"Cached Model-as-a-Resource: Provisioning large language model agents for edge intelligence in space-air-ground integrated networks","author":"Xu","year":"2024"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/TWC.2024.3510418"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1145\/3711896.3737258"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/JSAC.2025.3645168"},{"key":"ref21","article-title":"Adaptive layer splitting for wireless LLM inference in edge computing: A model-based reinforcement learning approach","author":"Chen","year":"2024"},{"key":"ref22","article-title":"$\\Lambda$\u039b -split: A privacy-preserving split computing framework for cloud-powered generative AI","author":"Ohta","year":"2023"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/TWC.2023.3303232"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/icc52391.2025.11161529"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/JIOT.2023.3298366"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1810.04805"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/COMST.2017.2745201"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/TMC.2018.2794454"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/TCOMM.2013.071013.120787"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/TWC.2020.3031436"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/TVT.2019.2896906"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1109\/TWC.2019.2894403"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1109\/TCOMM.2022.3152265"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1109\/TCOMM.2018.2880482"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/TMC.2023.3331690"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/TCOMM.2019.2956041"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/WCNC.2019.8885590"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1109\/TCOMM.2021.3078776"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1162\/089976600300015880"},{"issue":"5","key":"ref40","first-page":"905","article-title":"Loopy belief propagation: Convergence and effects of message errors","volume":"6","author":"Ihler","year":"2005","journal-title":"J. Mach. Learn. Res."},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/JIOT.2020.3041673"},{"key":"ref42","article-title":"Logits of API-protected LLMs leak proprietary information","author":"Finlayson","year":"2024"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/W18-5446"}],"container-title":["IEEE Transactions on Mobile Computing"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/7755\/11345503\/11154975.pdf?arnumber=11154975","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,12]],"date-time":"2026-01-12T23:59:03Z","timestamp":1768262343000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11154975\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2]]},"references-count":43,"journal-issue":{"issue":"2"},"URL":"https:\/\/doi.org\/10.1109\/tmc.2025.3608303","relation":{},"ISSN":["1536-1233","1558-0660","2161-9875"],"issn-type":[{"value":"1536-1233","type":"print"},{"value":"1558-0660","type":"electronic"},{"value":"2161-9875","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2]]}}}