{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,8]],"date-time":"2026-03-08T01:36:53Z","timestamp":1772933813313,"version":"3.50.1"},"reference-count":28,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,12,8]],"date-time":"2025-12-08T00:00:00Z","timestamp":1765152000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,12,8]],"date-time":"2025-12-08T00:00:00Z","timestamp":1765152000000},"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":[[2025,12,8]]},"DOI":"10.1109\/bigdata66926.2025.11400767","type":"proceedings-article","created":{"date-parts":[[2026,3,6]],"date-time":"2026-03-06T20:57:57Z","timestamp":1772830677000},"page":"3115-3124","source":"Crossref","is-referenced-by-count":0,"title":["DSDE: Dynamic Speculative Decoding with KLD Stability for Real-World Serving"],"prefix":"10.1109","author":[{"given":"Mingyu","family":"Yang","sequence":"first","affiliation":[{"name":"Samsung SDS,Cloud Research Team,Seoul,Republic of Korea"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jae-Young","family":"Choi","sequence":"additional","affiliation":[{"name":"Samsung SDS,Cloud Research Team,Seoul,Republic of Korea"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Kihyo","family":"Moon","sequence":"additional","affiliation":[{"name":"Samsung SDS,Cloud Research Team,Seoul,Republic of Korea"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Minsung","family":"Jang","sequence":"additional","affiliation":[{"name":"Samsung SDS,Cloud Research Team,Seoul,Republic of Korea"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Eunjoo","family":"Jeon","sequence":"additional","affiliation":[{"name":"Samsung SDS,Cloud Research Team,Seoul,Republic of Korea"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Fast transformer decoding: One write-head is all you need","author":"Shazeer","year":"2019","journal-title":"arXiv preprint"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1145\/3458817.3476209"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1189"},{"key":"ref4","volume-title":"S-lora: Serving thousands of concurrent lora adapters","author":"Sheng","year":"2024"},{"key":"ref5","article-title":"Accelerating large language model decoding with speculative sampling","author":"Chen","year":"2023"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1145\/3600006.3613165"},{"key":"ref7","volume-title":"Medusa: Simple 11 m inference acceleration framework with multiple decoding heads","author":"Cai","year":"2024"},{"key":"ref8","volume-title":"Eagle: Speculative sampling requires rethinking feature uncertainty","author":"Li","year":"2025"},{"key":"ref9","article-title":"Del: Context-aware dynamic exit layer for efficient self-speculative decoding","volume-title":"arXiv preprint","author":"Entezari Zarch","year":"2025"},{"key":"ref10","volume-title":"Fast inference from transformers via speculative decoding","author":"Leviathan","year":"2023"},{"key":"ref11","article-title":"Dynamic speculation lookahead accelerates speculative decoding of large language models","author":"Mamou","year":"2024","journal-title":"arXiv preprint"},{"key":"ref12","volume-title":"Speeding up speculative decoding via sequential approximate verification","author":"Zhong","year":"2025"},{"key":"ref13","article-title":"Adaedl: Early draft stopping for speculative decoding of large language models via an entropy-based lower bound on token acceptance probability","author":"Agrawal","year":"2024","journal-title":"arXiv preprint"},{"key":"ref14","volume-title":"The stable entropy hypothesis and entropy-aware decoding: An analysis and algorithm for robust natural language generation","author":"Arora","year":"2023"},{"key":"ref15","volume-title":"Draft model knows when to stop: Self-verification speculative decoding for long-form generation","author":"Zhang","year":"2025"},{"key":"ref16","doi-asserted-by":"crossref","DOI":"10.1007\/s11280-025-01344-0","article-title":"Dyspec: Faster speculative decoding with dynamic token tree structure","volume-title":"Proceedings of the ACM Web Conference2025","author":"Xiong","year":"2025"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.489"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1145\/3394486.3406703"},{"key":"ref19","volume-title":"Understanding stragglers in large model training using what-if analysis","author":"Lin","year":"2025"},{"key":"ref20","article-title":"Flashattention-2: Faster attention with better parallelism and work partitioning","author":"Dao","year":"2023","journal-title":"arXiv preprint"},{"key":"ref21","first-page":"1693","article-title":"Teaching machines to read and comprehend","volume-title":"Proceedings of the 29th International Conference on Neural Information Processing Systems - Volume1, ser. NIPS\u201915","author":"Hermann"},{"key":"ref22","first-page":"1797","article-title":"Don\u2019t give me the details, just the summary! topic-aware convolutional neural networks for extreme summarization","volume-title":"Proceedings of the2018 Conference on Empirical Methods in Natural Language Processing","author":"Narayan"},{"key":"ref23","article-title":"Training verifiers to solve math word problems","author":"Cobbe","year":"2021","journal-title":"arXiv preprint"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D18-1259"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00276"},{"key":"ref26","article-title":"Evaluating large language models trained on code","author":"Chen","year":"2021","journal-title":"arXiv preprint"},{"key":"ref27","volume-title":"Sharegpt: Share chatgpt conversations","year":"2025"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.3115\/v1\/W14-3302"}],"event":{"name":"2025 IEEE International Conference on Big Data (BigData)","location":"Macau, China","start":{"date-parts":[[2025,12,8]]},"end":{"date-parts":[[2025,12,11]]}},"container-title":["2025 IEEE International Conference on Big Data (BigData)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11400704\/11400712\/11400767.pdf?arnumber=11400767","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,7]],"date-time":"2026-03-07T06:52:31Z","timestamp":1772866351000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11400767\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,12,8]]},"references-count":28,"URL":"https:\/\/doi.org\/10.1109\/bigdata66926.2025.11400767","relation":{},"subject":[],"published":{"date-parts":[[2025,12,8]]}}}