{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T06:45:09Z","timestamp":1785912309304,"version":"3.56.0"},"reference-count":54,"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":["62402311"],"award-info":[{"award-number":["62402311"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,6]]},"DOI":"10.1109\/isca66397.2026.00101","type":"proceedings-article","created":{"date-parts":[[2026,8,4]],"date-time":"2026-08-04T19:11:22Z","timestamp":1785870682000},"page":"1336-1350","source":"Crossref","is-referenced-by-count":0,"title":["STEP: Adaptive Spatio-Temporal Expert Prefetching for Low-Latency and Memory-Efficient MoE Inference"],"prefix":"10.1109","author":[{"given":"Fangxin","family":"Liu","sequence":"first","affiliation":[{"name":"School of Computer Science, Shanghai Jiao Tong University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ning","family":"Yang","sequence":"additional","affiliation":[{"name":"School of Computer Science, Shanghai Jiao Tong University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zongwu","family":"Wang","sequence":"additional","affiliation":[{"name":"School of Computer Science, Shanghai Jiao Tong University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chenyang","family":"Guan","sequence":"additional","affiliation":[{"name":"School of Computer Science, Shanghai Jiao Tong University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Haomin","family":"Li","sequence":"additional","affiliation":[{"name":"School of Computer Science, Shanghai Jiao Tong University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yu","family":"Feng","sequence":"additional","affiliation":[{"name":"School of Computer Science, Shanghai Jiao Tong University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Liqiang","family":"Lu","sequence":"additional","affiliation":[{"name":"Zhejiang University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiang","family":"Li","sequence":"additional","affiliation":[{"name":"Alibaba Group"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Siran","family":"Yang","sequence":"additional","affiliation":[{"name":"Alibaba Group"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jiamang","family":"Wang","sequence":"additional","affiliation":[{"name":"Alibaba Group"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lin","family":"Qu","sequence":"additional","affiliation":[{"name":"Alibaba Group"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Li","family":"Jiang","sequence":"additional","affiliation":[{"name":"School of Computer Science, Shanghai Jiao Tong University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Haibing","family":"Guan","sequence":"additional","affiliation":[{"name":"School of Computer Science, Shanghai Jiao Tong University"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Gpt-4 technical report","volume-title":"arXiv preprint","author":"Achiam","year":"2023"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1145\/3690624.3709196"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.172"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v34i05.6239"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1145\/3669940.3707267"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.acl-long.633"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA61900.2025.00084"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO61859.2024.00048"},{"key":"ref9","year":"2024","journal-title":"Dbrx: A new open-source large language model"},{"key":"ref10","article-title":"Think you have solved question answering? try arc, the ai2 reasoning challenge","volume-title":"ArXiv","volume":"abs\/1803.05457","author":"Clark","year":"2018"},{"key":"ref11","first-page":"224","article-title":"Sida: Sparsity-inspired data-aware serving for efficient and scalable large mixture-of-experts models","volume-title":"Proceedings of Machine Learning and Systems","volume":"6","author":"Du","year":"2024"},{"key":"ref12","first-page":"arXiv","article-title":"The llama 3 herd of models","author":"Dubey","year":"2024","journal-title":"arXiv e-prints"},{"key":"ref13","article-title":"Fast inference of mixture-of-experts language models with offloading","author":"Eliseev","year":"2023","journal-title":"arXiv preprint"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA61900.2025.00110"},{"issue":"120","key":"ref15","first-page":"1","article-title":"Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity","volume":"23","author":"Fedus","year":"2022","journal-title":"Journal of Machine Learning Research"},{"key":"ref16","article-title":"Delta decompression for moe-based llms compression","author":"Gu","year":"2025","journal-title":"arXiv preprint"},{"key":"ref17","article-title":"Deepseek-r1: Incentivizing reasoning capability in 1lms via reinforcement learning","author":"Guo","year":"2025","journal-title":"arXiv preprint"},{"key":"ref18","article-title":"Measuring massive multitask language understanding","author":"Hendrycks","year":"2020","journal-title":"arXiv preprint"},{"key":"ref19","article-title":"Mixture compressor for mixture-of-experts llms gains more","author":"Huang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA59077.2024.00078"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1162\/neco.1991.3.1.79"},{"key":"ref22","article-title":"Mixtral of experts","author":"Jiang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref23","article-title":"Reap the experts: Why pruning prevails for one-shot moe compression","author":"Lasby","year":"2025","journal-title":"arXiv preprint"},{"key":"ref24","article-title":"Gshard: Scaling giant models with conditional computation and automatic sharding","author":"Lepikhin","year":"2020","journal-title":"arXiv preprint"},{"key":"ref25","author":"Li","year":"2025","journal-title":"Structured mixture-of-experts 1 lms compression via singular value decomposition"},{"key":"ref26","article-title":"Slimmoe: Structured compression of large moe models via expert slimming and distillation","author":"Li","year":"2025","journal-title":"arXiv preprint"},{"key":"ref27","article-title":"Deepseek-v2: A strong, economical, and efficient mixture-of-experts language model","author":"Liu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA57654.2024.00082"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1145\/3779212.3790155"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00523"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/tcad.2025.3604321"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1007\/s10462-012-9338-y"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1016\/j.csi.2025.103996"},{"issue":"140","key":"ref34","first-page":"1","article-title":"Exploring the limits of transfer learning with a unified text-to-text transformer","volume":"21","author":"Raffel","year":"2020","journal-title":"Journal of machine learning research"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1145\/3474381"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P17-1099"},{"key":"ref37","article-title":"Expertflow: Adaptive expert scheduling and memory coordination for efficient moe inference","author":"Shen","year":"2025","journal-title":"arXiv preprint"},{"key":"ref38","article-title":"Qwen2 technical report","volume-title":"arXiv preprint","author":"Team","year":"2024"},{"key":"ref39","article-title":"Llama 2: Open foundation and fine-tuned chat models","author":"Touvron","year":"2023","journal-title":"arXiv preprint"},{"key":"ref40","article-title":"Od-moe: Ondemand expert loading for cacheless edge-distributed moe inference","author":"Wang","year":"2025","journal-title":"arXiv preprint"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/SC41406.2024.00096"},{"key":"ref42","article-title":"Bloom: A 176b-parameter open-access multilingual language model","volume-title":"arXiv preprint","author":"Workshop","year":"2022"},{"key":"ref43","year":"2025","journal-title":"Grok-1: Open-source release of 314b parameter mixture-ofexperts model"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1145\/3620666.3651355"},{"key":"ref45","article-title":"Moe-infinity: Offloadingefficient moe model serving","author":"Xue","year":"2024","journal-title":"arXiv preprint"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-emnlp.612"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1145\/3799719"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1109\/DAC63849.2025.11132980"},{"key":"ref49","article-title":"Edgemoe: Fast on-device inference of moe-based large language models","author":"Yi","year":"2023","journal-title":"arXiv preprint"},{"key":"ref50","first-page":"arXiv","article-title":"fmoe: fine-grained expert offloading for large mixture-of-experts serving","author":"Yu","year":"2025","journal-title":"arXiv e-prints"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1145\/3689031.3696070"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.23919\/DATE64628.2025.10992741"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1145\/3676536.3676741"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1109\/DAC63849.2025.11133274"}],"event":{"name":"2026 ACM\/IEEE 53rd Annual International Symposium on Computer Architecture (ISCA)","location":"Raleigh, NC, USA","start":{"date-parts":[[2026,6,27]]},"end":{"date-parts":[[2026,7,1]]}},"container-title":["2026 ACM\/IEEE 53rd Annual International Symposium on Computer Architecture (ISCA)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11617393\/11617394\/11617726.pdf?arnumber=11617726","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T05:52:05Z","timestamp":1785909125000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11617726\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6]]},"references-count":54,"URL":"https:\/\/doi.org\/10.1109\/isca66397.2026.00101","relation":{},"subject":[],"published":{"date-parts":[[2026,6]]}}}