{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T17:56:05Z","timestamp":1785952565154,"version":"3.56.0"},"reference-count":64,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,1,31]],"date-time":"2026-01-31T00:00:00Z","timestamp":1769817600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,1,31]],"date-time":"2026-01-31T00:00:00Z","timestamp":1769817600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100012166","name":"National Key Research and Development Program of China","doi-asserted-by":"publisher","award":["2023YFB4503400"],"award-info":[{"award-number":["2023YFB4503400"]}],"id":[{"id":"10.13039\/501100012166","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62402456,62450064,62322205"],"award-info":[{"award-number":["62402456,62450064,62322205"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,1,31]]},"DOI":"10.1109\/hpca68181.2026.11408598","type":"proceedings-article","created":{"date-parts":[[2026,3,4]],"date-time":"2026-03-04T20:47:22Z","timestamp":1772657242000},"page":"1-15","source":"Crossref","is-referenced-by-count":1,"title":["Adaptive Draft Sequence Length: Enhancing Speculative Decoding Throughput on PIM-Enabled Systems"],"prefix":"10.1109","author":[{"given":"Runze","family":"Wang","sequence":"first","affiliation":[{"name":"School of Computer Science and Technology, Huazhong University of Science and Technology,National Engineering Research Center for Big Data Technology and System\/Services Computing Technology and System Lab\/Cluster and Grid Computing Lab,Wuhan,China,430074"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Qinggang","family":"Wang","sequence":"additional","affiliation":[{"name":"School of Computer Science and Technology, Huazhong University of Science and Technology,National Engineering Research Center for Big Data Technology and System\/Services Computing Technology and System Lab\/Cluster and Grid Computing Lab,Wuhan,China,430074"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Haifeng","family":"Liu","sequence":"additional","affiliation":[{"name":"School of Computer Science and Engineering, University of New South Wales,Sydney,NSW,Australia,2052"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Long","family":"Zheng","sequence":"additional","affiliation":[{"name":"School of Computer Science and Technology, Huazhong University of Science and Technology,National Engineering Research Center for Big Data Technology and System\/Services Computing Technology and System Lab\/Cluster and Grid Computing Lab,Wuhan,China,430074"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiaofei","family":"Liao","sequence":"additional","affiliation":[{"name":"School of Computer Science and Technology, Huazhong University of Science and Technology,National Engineering Research Center for Big Data Technology and System\/Services Computing Technology and System Lab\/Cluster and Grid Computing Lab,Wuhan,China,430074"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hai","family":"Jin","sequence":"additional","affiliation":[{"name":"School of Computer Science and Technology, Huazhong University of Science and Technology,National Engineering Research Center for Big Data Technology and System\/Services Computing Technology and System Lab\/Cluster and Grid Computing Lab,Wuhan,China,430074"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jingling","family":"Xue","sequence":"additional","affiliation":[{"name":"School of Computer Science and Engineering, University of New South Wales,Sydney,NSW,Australia,2052"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","first-page":"117","article-title":"Taming throughput-latency tradeoff in LLM inference with sarathi-serve","volume-title":"Proceedings of the 18th USENIX Symposium on Operating Systems Design and Implementation (OSDI)","author":"Agrawal","year":"2024"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/sc41404.2022.00051"},{"key":"ref3","article-title":"Palm 2 technical report","author":"Anil","year":"2023","journal-title":"arXiv preprint"},{"key":"ref4","article-title":"Medusa: Simple LLM inference acceleration framework with multiple decoding heads","volume-title":"Proceedings of the 41st International Conference on Machine Learning (ICML)","author":"Cai","year":"2024"},{"key":"ref5","article-title":"Accelerating large language model decoding with speculative sampling","author":"Chen","year":"2023","journal-title":"arXiv preprint"},{"key":"ref6","article-title":"Evaluating large language models trained on code","author":"Chen","year":"2021","journal-title":"arXiv preprint"},{"key":"ref7","volume-title":"Free dolly: Introducing the world\u2019s first truly open instruction-tuned 11m","author":"Conover","year":"2023"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.70"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/HOTCHIPS.2019.8875680"},{"key":"ref10","volume-title":"Gemini","year":"2025"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1145\/3676641.3716267"},{"key":"ref12","article-title":"R-bench: Graduate-level multidisciplinary benchmarks for 11 m & mllm complex reasoning evaluation","author":"Guo","year":"2025","journal-title":"arXiv preprint"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/hpca47549.2020.00035"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA52012.2021.00060"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1145\/3676641.3716009"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1145\/3620666.3651380"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO56248.2022.00051"},{"key":"ref18","year":"2022","journal-title":"High bandwidth memory dram (hbm3)"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1145\/3600006.3613165"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA52012.2021.00013"},{"key":"ref21","first-page":"19274","article-title":"Fast inference from transformers via speculative decoding","volume-title":"Proceedings of the International Conference on Machine Learning (ICML)","author":"Leviathan","year":"2023"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1145\/3620666.3651352"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1145\/3712285.3759903"},{"key":"ref24","article-title":"PEARL: parallel speculative decoding with adaptive draft length","volume-title":"Proceedings of the Thirteenth International Conference on Learning Representations (ICLR)","author":"Liu","year":"2025"},{"key":"ref25","article-title":"Turbospec: Closed-loop speculation control system for optimizing llm serving goodput","author":"Liu","year":"2025","journal-title":"arXiv preprint"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1145\/3466752.3480125"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/LCA.2023.3333759"},{"key":"ref28","first-page":"456","article-title":"Dynamic speculation lookahead accelerates speculative decoding of large language models","volume-title":"Proceedings of the NeurIPS Workshop on Efficient Natural Language and Speech Processing (ENLSP)","volume":"262","author":"Mamou","year":"2024"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1145\/3620666.3651335"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1145\/3458817.3476209"},{"key":"ref31","article-title":"Codegen: An open large language model for code with multi-turn program synthesis","volume-title":"Proceedings of the Eleventh International Conference on Learning Representations (ICLR)","author":"Nijkamp","year":"2023"},{"key":"ref32","volume-title":"NVIDIA DGX A100","year":"2023"},{"key":"ref33","volume-title":"Nvlink and nvlink switch","year":"2025"},{"key":"ref34","first-page":"41","article-title":"Fine-grained DRAM: energy-efficient DRAM for extreme bandwidth systems","volume-title":"Proceedings of the 50th Annual IEEE\/ACM International Symposium on Microarchitecture (MICRO)","author":"Connor","year":"2017"},{"key":"ref35","article-title":"Gpt-4 technical report","year":"2024","journal-title":"arXiv preprint"},{"key":"ref36","volume-title":"Chatgpt","year":"2025"},{"key":"ref37","first-page":"27730","article-title":"Training language models to follow instructions with human feedback","volume-title":"Proceedings of the 36th International Conference on Neural Information Processing Systems (NeurIPS)","author":"Ouyang","year":"2022"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1145\/3620665.3640422"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/JSSC.2022.3193354"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA59077.2024.00019"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1145\/3503222.3507738"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.acl-long.368"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/JSSC.2022.3232096"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1145\/3620666.3651324"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1145\/3669900"},{"key":"ref46","article-title":"Accelerating 11m inference with staged speculative decoding","author":"Spector","year":"2023","journal-title":"arXiv preprint"},{"key":"ref47","article-title":"The llama 3 herd of models","author":"Team","year":"2024","journal-title":"arXiv preprint"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA51647.2021.00018"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00735"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1145\/3649329.3655945"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO61859.2024.00040"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO56248.2022.00088"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1145\/3431920.3439288"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1109\/hpca53966.2022.00022"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1145\/3390523"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.1145\/3712285.3759834"},{"key":"ref58","first-page":"521","article-title":"Orca: A distributed serving system for transformer-based generative models","volume-title":"Proceedings of the 16th USENIX Symposium on Operating Systems Design and Implementation (OSDI)","author":"Yu","year":"2022"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO61859.2024.00105"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.607"},{"key":"ref61","article-title":"Opt: Open pre-trained transformer language models","author":"Zhang","year":"2022","journal-title":"arXiv preprint"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1145\/3695053.3731412"},{"key":"ref63","first-page":"559","article-title":"Alpa: Automating inter- and intra-operator parallelism for distributed deep learning","volume-title":"Proceedings of the 16th USENIX Symposium on Operating Systems Design and Implementation (OSDI)","author":"Zheng","year":"2022"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA53966.2022.00082"}],"event":{"name":"2026 IEEE International Symposium on High Performance Computer Architecture (HPCA)","location":"Sydney, Australia","start":{"date-parts":[[2026,1,31]]},"end":{"date-parts":[[2026,2,4]]}},"container-title":["2026 IEEE International Symposium on High Performance Computer Architecture (HPCA)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11408404\/11408433\/11408598.pdf?arnumber=11408598","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,5]],"date-time":"2026-03-05T06:36:24Z","timestamp":1772692584000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11408598\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,31]]},"references-count":64,"URL":"https:\/\/doi.org\/10.1109\/hpca68181.2026.11408598","relation":{},"subject":[],"published":{"date-parts":[[2026,1,31]]}}}