{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,17]],"date-time":"2026-08-17T19:39:06Z","timestamp":1786995546087,"version":"build-2736575974"},"reference-count":43,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"National Natural Science Foundation of China","award":["62441225,62572171"],"award-info":[{"award-number":["62441225,62572171"]}]},{"name":"U.S. National Science Foundation","award":["CNS-2047719,CNS-2225949"],"award-info":[{"award-number":["CNS-2047719,CNS-2225949"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,6]]},"DOI":"10.1109\/2575-8411.2026.00047","type":"proceedings-article","created":{"date-parts":[[2026,8,4]],"date-time":"2026-08-04T19:18:11Z","timestamp":1785871091000},"page":"425-435","source":"Crossref","is-referenced-by-count":0,"title":["Incentivizing and Orchestrating Cloud-Edge LLM Speculative Decoding via Auctions"],"prefix":"10.1109","author":[{"given":"Mingtao","family":"Ji","sequence":"first","affiliation":[{"name":"College of Computer Science and Software Engineering, Hohai University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lei","family":"Jiao","sequence":"additional","affiliation":[{"name":"Center for Cyber Security and Privacy, University of Oregon"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bin","family":"Tang","sequence":"additional","affiliation":[{"name":"College of Computer Science and Software Engineering, Hohai University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhihao","family":"Qu","sequence":"additional","affiliation":[{"name":"College of Computer Science and Software Engineering, Hohai University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chen","family":"Chen","sequence":"additional","affiliation":[{"name":"Nottingham Trent University","department":["Department of Computer Science"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Baoliu","family":"Ye","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Novel Software Technology, Nanjing University"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Fast and cost-effective speculative edge-cloud decoding with early exits","author":"Venkatesha","year":"2025","journal-title":"arXiv"},{"key":"ref2","article-title":"Specedge: Scalable edge-assisted serving framework for interactive LLMs","author":"Park","year":"2025","journal-title":"NeurIPS"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/WCSP68525.2025.1010651"},{"key":"ref4","article-title":"Dssd: Efficient edge-device deployment and collaborative inference via distributed split speculative decoding","volume-title":"ICML Workshop","author":"NING","year":"2025"},{"key":"ref5","article-title":"Llmcad: Fast and scalable on-device large language model inference","author":"Xu","year":"2023","journal-title":"arXiv"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/tpds.2021.3086274"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM42981.2021.9488733"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/IWQoS61813.2024.10682835"},{"key":"ref9","volume-title":"Edge Computing Hosting","year":"2026"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/TMC.2025.3554816"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM48880.2022.9796676"},{"key":"ref12","article-title":"Fast inference from transformers via speculative decoding","author":"Leviathan","year":"2023","journal-title":"ICML"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1137\/0315059"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/SECON64284.2024.10934850"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1613\/jair.2046"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/INFOCOM.2014.6847966"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/COMST.2025.3527641"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/JIOT.2024.3524255"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/ICC52391.2025.11161529"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/TSC.2025.3611379"},{"key":"ref21","article-title":"Perllm: Personalized inference scheduling with edge-cloud collaboration for diverse llm services","author":"Yang","year":"2024","journal-title":"arXiv"},{"key":"ref22","article-title":"Jppo++: Joint power and denoising-inspired prompt optimization for mobile 1lm services","author":"You","year":"2025","journal-title":"arXiv"},{"key":"ref23","article-title":"Fast and cost-effective speculative edge-cloud decoding with early exits","author":"Venkatesha","year":"2025","journal-title":"Transactions on Machine Learning Research"},{"key":"ref24","article-title":"Hybrid slm and 1lm for edge-cloud collaborative inference","volume-title":"EdgeFM Workshop","author":"Hao","year":"2024"},{"key":"ref25","article-title":"Auctions with 1 lm summaries","author":"Dubey","year":"2024","journal-title":"ACM KDD"},{"key":"ref26","article-title":"Llm-auction: Generative auction towards 1lm-native advertising","author":"Zhao","year":"2025","journal-title":"arXiv"},{"key":"ref27","article-title":"Rtbagent: A llm-based agent system for realtime bidding","author":"Cai","year":"2025","journal-title":"WWW"},{"key":"ref28","article-title":"Llm-based proxies for preference elicitation in combinatorial auctions","volume-title":"Harvard University Engineering and Applied Sciences. Bachelors Thesis","author":"Huang","year":"2024"},{"key":"ref29","article-title":"Bid: Behavioral agents in dynamic auctions","volume-title":"ICLR Workshop","author":"Zhang","year":"2025"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1007\/s11390-024-3342-y"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1287\/moor.6.1.58"},{"key":"ref32","volume-title":"Specedge Dataset","year":"2026"},{"key":"ref33","article-title":"Wildchat: 1m chatgpt interaction logs in the wild","author":"Zhao","year":"2024","journal-title":"arXiv:2405.01470"},{"key":"ref34","article-title":"Fast inference from transformers via speculative decoding","author":"Leviathan","year":"2023","journal-title":"arXiv"},{"key":"ref35","volume-title":"California ISO","year":"2026"},{"key":"ref36","volume-title":"Planetlab","year":"2026"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/ICPADS60453.2023.00304"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1109\/TNET.2024.3505252"},{"key":"ref39","volume-title":"Welcome to CVXPY 1.1","year":"2025"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1007\/s10107980020a"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/TSP.2017.2750109"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1109\/TNET.2023.3253302"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1016\/0024-3795(94)00344-D"}],"event":{"name":"2026 IEEE 46th International Conference on Distributed Computing Systems (ICDCS)","location":"Seoul, Korea, Republic of","start":{"date-parts":[[2026,6,22]]},"end":{"date-parts":[[2026,6,25]]}},"container-title":["2026 IEEE 46th International Conference on Distributed Computing Systems (ICDCS)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11618978\/11618981\/11619155.pdf?arnumber=11619155","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,17]],"date-time":"2026-08-17T19:13:07Z","timestamp":1786993987000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11619155\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6]]},"references-count":43,"URL":"https:\/\/doi.org\/10.1109\/2575-8411.2026.00047","relation":{},"subject":[],"published":{"date-parts":[[2026,6]]}}}