{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,23]],"date-time":"2026-07-23T15:57:21Z","timestamp":1784822241342,"version":"3.55.0"},"reference-count":95,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"8","license":[{"start":{"date-parts":[[2025,8,1]],"date-time":"2025-08-01T00:00:00Z","timestamp":1754006400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,8,1]],"date-time":"2025-08-01T00:00:00Z","timestamp":1754006400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,8,1]],"date-time":"2025-08-01T00:00:00Z","timestamp":1754006400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["U21B2016"],"award-info":[{"award-number":["U21B2016"]}],"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":["62032003"],"award-info":[{"award-number":["62032003"]}],"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":["62425203"],"award-info":[{"award-number":["62425203"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100012226","name":"Fundamental Research Funds for the Central Universities","doi-asserted-by":"publisher","award":["2024ZCJH11"],"award-info":[{"award-number":["2024ZCJH11"]}],"id":[{"id":"10.13039\/501100012226","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Sichuan Science and Technology Plan"},{"name":"Unveiling and Leading Project","award":["2024YFCY0001"],"award-info":[{"award-number":["2024YFCY0001"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. on Mobile Comput."],"published-print":{"date-parts":[[2025,8]]},"DOI":"10.1109\/tmc.2025.3546466","type":"journal-article","created":{"date-parts":[[2025,2,27]],"date-time":"2025-02-27T18:57:00Z","timestamp":1740682620000},"page":"7059-7073","source":"Crossref","is-referenced-by-count":36,"title":["EdgeMoE: Empowering Sparse Large Language Models on Mobile Devices"],"prefix":"10.1109","volume":"24","author":[{"ORCID":"https:\/\/orcid.org\/0009-0003-6040-9596","authenticated-orcid":false,"given":"Rongjie","family":"Yi","sequence":"first","affiliation":[{"name":"State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8516-8045","authenticated-orcid":false,"given":"Liwei","family":"Guo","sequence":"additional","affiliation":[{"name":"University of Electronic Science and Technology of China, Chengdu, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shiyun","family":"Wei","sequence":"additional","affiliation":[{"name":"Zhongguancun Laboratory, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5743-9418","authenticated-orcid":false,"given":"Ao","family":"Zhou","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7245-1298","authenticated-orcid":false,"given":"Shangguang","family":"Wang","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6271-6993","authenticated-orcid":false,"given":"Mengwei","family":"Xu","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Microsoft translator enhanced with Z-code mixture of experts models","year":"2022"},{"key":"ref2","article-title":"Quanto: A PyTorch quantization backend for optimum","year":"2022"},{"key":"ref3","article-title":"Beating Google and Apple, Huawei brings large AI model to mobile voice assistant - Huawei Central","year":"2023"},{"key":"ref4","article-title":"c4. datasets at hugging face","year":"2023"},{"key":"ref5","article-title":"ChatGPT - Google play","year":"2023"},{"key":"ref6","article-title":"Jetson tx2 module | nvidia developer","year":"2023"},{"key":"ref7","article-title":"Model card for tanrei\/GPTSAN-Japanese","year":"2023"},{"key":"ref8","article-title":"Models - hugging face","year":"2023"},{"key":"ref9","article-title":"Move speed USB2.0","year":"2023"},{"key":"ref10","article-title":"Raspberry PI 4 model b \u2013 raspberry PI","year":"2023"},{"key":"ref11","article-title":"samsum . datasets at hugging face","year":"2023"},{"key":"ref12","article-title":"Samsung 860 EVO | consumer SSD | specs & features | samsung semiconductor global","year":"2023"},{"key":"ref13","article-title":"SanDisk ultra microSD, UHS-I card, full HD store | western digital","year":"2023"},{"key":"ref14","article-title":"Wikipedia-Japanese . datasets at hugging face","year":"2023"},{"key":"ref15","article-title":"World\u2019s 1st on-device stable diffusion on android | qualcomm","year":"2023"},{"key":"ref16","article-title":"XSum . datasets at hugging face","year":"2023"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.ngt-1.4"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1145\/2994551.2994564"},{"key":"ref20","first-page":"1877","article-title":"Language models are few-shot learners","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Brown"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1145\/3089801.3089804"},{"key":"ref22","article-title":"Quip: 2-bit quantization of large language models with guarantees","author":"Chee","year":"2023"},{"key":"ref23","first-page":"22173","article-title":"TA-MoE: Topology-aware large scale mixture-of-expert training","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Chen"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.232"},{"key":"ref25","first-page":"34600","article-title":"On the representation collapse of sparse mixture of experts","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Chi"},{"key":"ref26","first-page":"797","article-title":"Optimizing dynamic neural networks with brainstorm","volume-title":"Proc. 17th USENIX Symp. Operating Syst. Des. Implementation","author":"Cui"},{"key":"ref27","article-title":"Flashattention-2: Faster attention with better parallelism and work partitioning","author":"Dao","year":"2023"},{"key":"ref28","first-page":"16344","article-title":"Flashattention: Fast and memory-efficient exact attention with IO-awareness","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Dao"},{"key":"ref29","article-title":"Skipdecode: Autoregressive skip decoding with batching and caching for efficient LLM inference","author":"Corro","year":"2023"},{"key":"ref30","article-title":"An image is worth 16x16 words: Transformers for image recognition at scale","author":"Dosovitskiy","year":"2020"},{"key":"ref31","first-page":"5547","article-title":"Glam: Efficient scaling of language models with mixture-of-experts","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Du"},{"key":"ref32","article-title":"SiDA: Sparsity-inspired data-aware serving for efficient and scalable large mixture-of-experts models","author":"Du","year":"2023"},{"key":"ref33","article-title":"GPTs are GPTs: An early look at the labor market impact potential of large language models","author":"Eloundou","year":"2023"},{"key":"ref34","article-title":"A review of sparse expert models in deep learning","author":"Fedus","year":"2022"},{"issue":"1","key":"ref35","first-page":"5232","article-title":"Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity","volume":"23","author":"Fedus","year":"2022","journal-title":"J. Mach. Learn. Res."},{"key":"ref36","article-title":"QMOE: Practical sub-1-bit compression of trillion-parameter models","author":"Frantar","year":"2023"},{"key":"ref37","article-title":"SparseGPT: Massive language models can be accurately pruned in one-shot","author":"Frantar","year":"2023"},{"key":"ref38","article-title":"GPTQ: Accurate post-training quantization for generative pre-trained transformers","author":"Frantar","year":"2022"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pcbi.1000211"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1109\/SMARTCOMP.2017.7946996"},{"key":"ref41","article-title":"Specializing smaller language models towards multi-step reasoning","author":"Fu","year":"2023"},{"key":"ref42","article-title":"In-context autoencoder for context compression in a large language model","author":"Ge","year":"2023"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1145\/3575693.3575698"},{"key":"ref44","article-title":"Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding","author":"Han","year":"2015"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1145\/3081333.3081360"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1162\/neco.1991.3.1.79"},{"key":"ref47","article-title":"Mixtral of experts","author":"Jiang","year":"2024"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1145\/3297858.3304038"},{"key":"ref49","article-title":"$\\lbrace${FAST $\\rbrace$}: Quick application launch on $\\lbrace${ Solid-State$\\rbrace$} drives","volume-title":"Proc. 9th USENIX Conf. File Storage Technol.","author":"Joo","year":"2011"},{"key":"ref50","article-title":"SqueezeLLM: Dense-and-sparse quantization","author":"Kim","year":"2023"},{"key":"ref51","article-title":"Mixture of quantized experts (MoQE): Complementary effect of low-bit quantization and robustness","author":"Kim","year":"2023"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1109\/tc.2025.3575905"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1109\/IPSN.2016.7460664"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1145\/3372224.3419194"},{"key":"ref55","article-title":"GShard: Scaling giant models with conditional computation and automatic sharding","author":"Lepikhin","year":"2020"},{"key":"ref56","first-page":"6265","article-title":"Base layers: Simplifying training of large, sparse models","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Lewis"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.acl-long.150"},{"key":"ref58","article-title":"Losparse: Structured compression of large language models based on low-rank and sparse approximation","author":"Li","year":"2023"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1145\/3495243.3517020"},{"key":"ref60","article-title":"Recall-oriented understudy for gisting evaluation (rouge)","volume":"20","author":"Lin","year":"2005","journal-title":"Retrieved August"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1145\/3714983.3714987"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1145\/3210240.3210337"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.26"},{"key":"ref64","first-page":"22137","article-title":"Deja VU: Contextual sparsity for efficient llms at inference time","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Liu"},{"key":"ref65","article-title":"LLM-pruner: On the structural pruning of large language models","author":"Ma","year":"2023"},{"key":"ref66","doi-asserted-by":"publisher","DOI":"10.1145\/3081333.3081359"},{"key":"ref67","article-title":"Enabling large neural networks on tiny microcontrollers with swapping","author":"Miao","year":"2021"},{"key":"ref68","article-title":"Specinfer: Accelerating generative llm serving with speculative inference and token tree verification","author":"Miao","year":"2023"},{"key":"ref69","article-title":"Skeleton-of-thought: Large language models can do parallel decoding","author":"Ning","year":"2023"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.1145\/3373376.3378534"},{"key":"ref71","article-title":"GPT-4 technical report","year":"2023"},{"key":"ref72","first-page":"27730","article-title":"Training language models to follow instructions with human feedback","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Ouyang"},{"key":"ref73","article-title":"Improving language understanding by generative pre-training","author":"Radford","year":"2018"},{"issue":"8","key":"ref74","article-title":"Language models are unsupervised multitask learners","volume":"1","author":"Radford","year":"2019","journal-title":"OpenAI Blog"},{"key":"ref75","first-page":"17555","article-title":"Hash layers for large sparse models","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Roller"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1109\/ICCAD57390.2023.10323651"},{"key":"ref77","article-title":"Outrageously large neural networks: The sparsely-gated mixture-of-experts layer","author":"Shazeer","year":"2017"},{"key":"ref78","article-title":"Accelerating LLM inference with staged speculative decoding","author":"Spector","year":"2023"},{"key":"ref79","article-title":"A simple and effective pruning approach for large language models","author":"Sun","year":"2023"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.acl-industry.15"},{"key":"ref81","article-title":"LLAMA: Open and efficient foundation language models","author":"Touvron","year":"2023"},{"key":"ref82","article-title":"LLAMA 2: Open foundation and fine-tuned chat models","author":"Touvron","year":"2023"},{"key":"ref83","article-title":"LLMzip: Lossless text compression using large language models","author":"Valmeekam","year":"2023"},{"key":"ref84","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref85","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.acl-long.304"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.1145\/3498361.3538928"},{"key":"ref87","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.eacl-long.57"},{"key":"ref88","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"},{"key":"ref89","doi-asserted-by":"publisher","DOI":"10.1109\/TMC.2019.2893250"},{"key":"ref90","doi-asserted-by":"publisher","DOI":"10.1145\/3241539.3241563"},{"key":"ref91","article-title":"TensorGPT: Efficient compression of the embedding layer in LLMs based on the tensor-train decomposition","author":"Xu","year":"2023"},{"key":"ref92","article-title":"MLLM: Fast and lightweight multimodal LLM inference engine for mobile and edge devices","author":"Yi","year":"2023"},{"key":"ref93","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.acl-long.236"},{"key":"ref94","article-title":"RPTQ: Reorder-based post-training quantization for large language models","author":"Yuan","year":"2023"},{"key":"ref95","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO50266.2020.00071"},{"key":"ref96","first-page":"7103","article-title":"Mixture-of-experts with expert choice routing","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhou"}],"container-title":["IEEE Transactions on Mobile Computing"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/7755\/11068893\/10906629.pdf?arnumber=10906629","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,8]],"date-time":"2025-07-08T17:40:55Z","timestamp":1751996455000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10906629\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,8]]},"references-count":95,"journal-issue":{"issue":"8"},"URL":"https:\/\/doi.org\/10.1109\/tmc.2025.3546466","relation":{},"ISSN":["1536-1233","1558-0660","2161-9875"],"issn-type":[{"value":"1536-1233","type":"print"},{"value":"1558-0660","type":"electronic"},{"value":"2161-9875","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,8]]}}}