{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,19]],"date-time":"2026-02-19T07:21:51Z","timestamp":1771485711249,"version":"3.50.1"},"reference-count":15,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,10,27]],"date-time":"2025-10-27T00:00:00Z","timestamp":1761523200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,10,27]],"date-time":"2025-10-27T00:00:00Z","timestamp":1761523200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,27]]},"DOI":"10.1109\/tencon66050.2025.11375380","type":"proceedings-article","created":{"date-parts":[[2026,2,18]],"date-time":"2026-02-18T21:13:38Z","timestamp":1771449218000},"page":"1023-1027","source":"Crossref","is-referenced-by-count":0,"title":["Efficient Task Scheduling Algorithms for Decentralized Large Language Model Serving"],"prefix":"10.1109","author":[{"given":"Sanjaya Kumar","family":"Panda","sequence":"first","affiliation":[{"name":"National Institute of Technology Warangal,Department of Computer Science and Engineering,Warangal,India,506004"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sankalp","family":"Dubey","sequence":"additional","affiliation":[{"name":"National Institute of Technology Warangal,Department of Computer Science and Engineering,Warangal,India,506004"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Siba","family":"Mishra","sequence":"additional","affiliation":[{"name":"C. V. Raman Global University,Department of Computer Science and Engineering,Bhubaneswar,India,752054"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/COMST.2025.3564333"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1145\/3719664"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1145\/3641289"},{"key":"ref4","year":"2025","journal-title":"What is 11 m?"},{"key":"ref5","volume-title":"Large language models","author":"Johnsen","year":"2024"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2023.126661"},{"key":"ref7","author":"Ren","year":"2024","journal-title":"Task scheduling for decentralized 11 m serving in heterogeneous networks"},{"key":"ref8","article-title":"Petals: Collaborative inference and fine-tuning of large models","author":"Borzunov","year":"2022","journal-title":"arXiv preprint"},{"key":"ref9","first-page":"173","article-title":"Llumnix: Dynamic scheduling for large language model serving","volume-title":"18th USENIX Symposium on Operating Systems Design and Implementation (OSDI 24)","author":"Sun","year":"2024"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1145\/3676536.3676659"},{"key":"ref11","article-title":"Fast distributed inference serving for large language models","author":"Wu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref12","article-title":"Preble: Efficient distributed prompt scheduling for llm serving","author":"Srivatsa","year":"2024","journal-title":"arXiv preprint"},{"key":"ref13","first-page":"193","article-title":"Distserve: Disaggregating prefill and decoding for goodput-optimized large language model serving","volume-title":"18th USENIX Symposium on Operating Systems Design and Implementation (OSDI 24)","author":"Zhong","year":"2024"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/W18-5446"},{"key":"ref15","article-title":"Superglue: A stickier benchmark for general-purpose language understanding systems","volume":"32","author":"Wang","year":"2019","journal-title":"Advances in neural information processing systems"}],"event":{"name":"TENCON 2025 - 2025 IEEE Region 10 Conference (TENCON)","location":"Kota Kinabalu, Malaysia","start":{"date-parts":[[2025,10,27]]},"end":{"date-parts":[[2025,10,30]]}},"container-title":["TENCON 2025 - 2025 IEEE Region 10 Conference (TENCON)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11373911\/11374914\/11375380.pdf?arnumber=11375380","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,2,19]],"date-time":"2026-02-19T06:57:32Z","timestamp":1771484252000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11375380\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,27]]},"references-count":15,"URL":"https:\/\/doi.org\/10.1109\/tencon66050.2025.11375380","relation":{},"subject":[],"published":{"date-parts":[[2025,10,27]]}}}