{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,16]],"date-time":"2026-01-16T13:43:39Z","timestamp":1768571019521,"version":"3.49.0"},"reference-count":44,"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\/501100018537","name":"National Science and Technology Major Project","doi-asserted-by":"publisher","award":["2022ZD0117602"],"award-info":[{"award-number":["2022ZD0117602"]}],"id":[{"id":"10.13039\/501100018537","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Beijing Municipal Science and Technology Program","award":["Z241100004224026"],"award-info":[{"award-number":["Z241100004224026"]}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62272026"],"award-info":[{"award-number":["62272026"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62104014"],"award-info":[{"award-number":["62104014"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"name":"State Key Laboratory of Complex & Critical Software Environment","award":["CCSE-2024ZX-10"],"award-info":[{"award-number":["CCSE-2024ZX-10"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Comput."],"published-print":{"date-parts":[[2026,2]]},"DOI":"10.1109\/tc.2025.3628629","type":"journal-article","created":{"date-parts":[[2025,11,5]],"date-time":"2025-11-05T18:39:02Z","timestamp":1762367942000},"page":"597-611","source":"Crossref","is-referenced-by-count":0,"title":["Accelerating LLM Inference via Low-Bit Fine-Grained Quantization Algorithm and Bit-Level Accelerator Co-Design"],"prefix":"10.1109","volume":"75","author":[{"ORCID":"https:\/\/orcid.org\/0009-0005-9988-2940","authenticated-orcid":false,"given":"Xilong","family":"Xie","sequence":"first","affiliation":[{"name":"State Key Laboratory of Complex & Critical Software Environment (CCSE) and the School of Computer Science and Engineering, Beihang University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9537-9986","authenticated-orcid":false,"given":"Liang","family":"Wang","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Complex & Critical Software Environment (CCSE) and the School of Computer Science and Engineering, Beihang University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9438-9181","authenticated-orcid":false,"given":"Limin","family":"Xiao","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Complex & Critical Software Environment (CCSE) and the School of Computer Science and Engineering, Beihang University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2386-961X","authenticated-orcid":false,"given":"Li","family":"Ruan","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Complex & Critical Software Environment (CCSE) and the School of Computer Science and Engineering, Beihang University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0009-0005-4448-7505","authenticated-orcid":false,"given":"Tairan","family":"Zhang","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Complex & Critical Software Environment (CCSE) and the School of Computer Science and Engineering, Beihang University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6690-8386","authenticated-orcid":false,"given":"Jinquan","family":"Wang","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Complex & Critical Software Environment (CCSE) and the School of Computer Science and Engineering, Beihang University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6960-6313","authenticated-orcid":false,"given":"Yongyue","family":"Wang","sequence":"additional","affiliation":[{"name":"Jiangsu Shuguang Optoelectronics Company Ltd., Yangzhou, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7924-9268","authenticated-orcid":false,"given":"Xiaojian","family":"Liao","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Complex & Critical Software Environment (CCSE) and the School of Computer Science and Engineering, Beihang University, Beijing, China"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref2","article-title":"Llama: Open and efficient foundation language models","author":"Touvron","year":"2023"},{"key":"ref3","article-title":"GPTQ: Accurate post-training quantization for generative pre-trained transformers","author":"Frantar","year":"2022"},{"key":"ref4","first-page":"87","article-title":"AWQ: Activation-aware weight quantization for on-device LLM compression and acceleration","volume":"6","author":"Lin","year":"2024","journal-title":"Proc. Mach. Learn. Syst."},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i12.29237"},{"key":"ref6","article-title":"PB-LLM: Partially binarized large language models","author":"Shang","year":"2023"},{"key":"ref7","first-page":"7197","article-title":"Up or down? Adaptive rounding for post-training quantization","volume-title":"Proc. Int. Conf. Mach. Learn.,","author":"Nagel","year":"2020"},{"key":"ref8","first-page":"38087","article-title":"SmoothQuant: Accurate and efficient post-training quantization for large language models","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Xiao","year":"2023"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.emnlp-main.98"},{"key":"ref10","article-title":"Mobile and edge evaluation of large language models","author":"Laskaridis","journal-title":"Proc. Workshop Efficient Syst. Found. Models II@ ICML2024."},{"key":"ref11","first-page":"20","article-title":"BiLLM: Pushing the limit of post-training quantization for LLMs","volume-title":"Proc. 41st Int. Conf. Mach. Learn.,","author":"Huang","year":"2024"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA57654.2024.00064"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA61900.2025.00084"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA61900.2025.00110"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1145\/3579371.3589038"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO56248.2022.00095"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.23919\/DATE64628.2025.10993129"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1145\/3600006.3613165"},{"issue":"12s","key":"ref19","first-page":"160","article-title":"Abstractive long text summarization using large language models","volume":"12","author":"Keswani","year":"2024","journal-title":"Int. J. Intell. Syst. Appl. Eng."},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00286"},{"key":"ref21","article-title":"Pointer sentinel mixture models","author":"Merity","year":"2016"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO50266.2020.00071"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1145\/3352460.3358252"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2786"},{"key":"ref25","article-title":"RPTQ: Reorder-based post-training quantization for large language models","author":"Yuan","year":"2023"},{"key":"ref26","article-title":"Gradual binary search and dimension expansion: A general method for activation quantization in LLMs","author":"Maisonnave","year":"2025"},{"key":"ref27","article-title":"PyTorch: An imperative style, high-performance deep learning library","author":"Paszke","year":"2019"},{"key":"ref28","article-title":"OPT: Open pre-trained transformer language models","author":"Zhang","year":"2022"},{"issue":"3","key":"ref29","first-page":"6","article-title":"Vicuna: An open-source chatbot impressing GPT-4 with 90%* ChatGPT quality","volume":"2","author":"Chiang","year":"2023"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4842-8844-3_4"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1145\/3474381"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v34i05.6239"},{"key":"ref33","article-title":"Microscaling data formats for deep learning","author":"Rouhani","year":"2023"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4757-2370-0"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1145\/2717764.2717783"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1145\/1105734.1105748"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.2307\/j.ctt5vkjtx.9"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v39i21.34385"},{"key":"ref39","first-page":"803","article-title":"{DecDEC}: A Systems Approach to Advancing,","volume-title":"Proc. 19th USENIX Symp. Operating Syst. Des. Implementation (OSDI),","author":"Park","year":"2025"},{"key":"ref40","article-title":"BitStack: Any-size compression of large language models in variable memory environments","author":"Wang","year":"2024"},{"key":"ref41","article-title":"FlatQuant: Flatness matters for LLM quantization","author":"Sun","year":"2024"},{"key":"ref42","article-title":"GuidedQuant: Large language model quantization via exploiting end loss guidance","author":"Kim","year":"2025"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA61900.2025.00112"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1145\/3669940.3707268"}],"container-title":["IEEE Transactions on Computers"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/12\/11353113\/11230088.pdf?arnumber=11230088","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,15]],"date-time":"2026-01-15T20:49:22Z","timestamp":1768510162000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11230088\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2]]},"references-count":44,"journal-issue":{"issue":"2"},"URL":"https:\/\/doi.org\/10.1109\/tc.2025.3628629","relation":{},"ISSN":["0018-9340","1557-9956","2326-3814"],"issn-type":[{"value":"0018-9340","type":"print"},{"value":"1557-9956","type":"electronic"},{"value":"2326-3814","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2]]}}}