{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,11]],"date-time":"2026-06-11T20:56:58Z","timestamp":1781211418390,"version":"3.54.1"},"reference-count":37,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,4,26]],"date-time":"2026-04-26T00:00:00Z","timestamp":1777161600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,4,26]],"date-time":"2026-04-26T00:00:00Z","timestamp":1777161600000},"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":[[2026,4,26]]},"DOI":"10.1109\/ispass69572.2026.00053","type":"proceedings-article","created":{"date-parts":[[2026,5,26]],"date-time":"2026-05-26T19:39:19Z","timestamp":1779824359000},"page":"484-494","source":"Crossref","is-referenced-by-count":0,"title":["HeadSkip: Characterizing and Accelerating \u2018Small\u2019 Language Models on Edge CPUs"],"prefix":"10.1109","author":[{"given":"Ian","family":"Sartor","sequence":"first","affiliation":[{"name":"University of Toronto,The Edward S. Rogers Sr. Department of Electrical and Computer Engineering,Toronto,ON"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Karthik","family":"Ganesan","sequence":"additional","affiliation":[{"name":"University of Toronto,The Edward S. Rogers Sr. Department of Electrical and Computer Engineering,Toronto,ON"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Anh","family":"Nguyen","sequence":"additional","affiliation":[{"name":"Fujitsu Limited,Toronto,ON"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Andreas","family":"Moshovos","sequence":"additional","affiliation":[{"name":"University of Toronto,The Edward S. Rogers Sr. Department of Electrical and Computer Engineering,Toronto,ON"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","volume-title":"Apple intelligence","year":"2023"},{"key":"ref2","volume-title":"Introducing apple\u2019s on-device and server foundation models","year":"2024"},{"key":"ref3","volume-title":"Apple intelligence gets even more powerful with new capabilities across apple devices","year":"2025"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.insights-1.17"},{"key":"ref5","volume-title":"Training verifiers to solve math word problems","author":"Cobbe","year":"2021"},{"key":"ref6","article-title":"Skipdecode: Autoregressive skip decoding with batching and caching for efficient 11 m inference","author":"Del Corro","year":"2023","journal-title":"arXiv preprint arXiv:2307.02628"},{"key":"ref7","article-title":"Gguf","author":"Gerganov","year":"2023","journal-title":"GitHub repository"},{"key":"ref8","article-title":"llama.cpp: Llm inference in c\/c++","year":"2023","journal-title":"GitHub repository"},{"key":"ref9","volume-title":"Introducing gemini: our largest and most capable ai model","year":"2025"},{"key":"ref10","article-title":"Mamba: Linear-time sequence modeling with selective state spaces","volume-title":"First conference on language modeling","author":"Gu"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0040"},{"key":"ref12","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 the 2018 Conference on Empirical Methods in Natural Language Processing","author":"Narayan"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.3758\/s13423-014-0585-6"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/COMST.2025.3527641"},{"key":"ref15","volume-title":"Qwen3-next: Towards ultimate training inference efficiency","year":"2025"},{"key":"ref16","volume-title":"Squad: 100,000+ questions for machine comprehension of text","author":"Rajpurkar","year":"2016"},{"key":"ref17","volume-title":"Llms with largest context windows","author":"scape","year":"2025"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1269"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1007\/978-981-96-9248-4_4"},{"key":"ref20","article-title":"Gemini: a family of highly capable multimodal models","author":"Team","year":"2023","journal-title":"arXiv preprint arXiv:2312.11805"},{"key":"ref21","article-title":"Gemma 3 technical report","volume-title":"arXiv preprint arXiv:2503.19786","author":"Team","year":"2025"},{"key":"ref22","article-title":"Gemma 2: Improving open language models at a practical size","author":"Team","year":"2024","journal-title":"arXiv preprint arXiv:2408.00118"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/ICPR.2016.7900006"},{"key":"ref24","article-title":"Llama 2: Open foundation and fine-tuned chat models","author":"Touvron","year":"2023","journal-title":"arXiv preprint arXiv:2307.09288"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.findings-emnlp.981"},{"key":"ref26","article-title":"A survey of small language models","author":"Van Nguyen","year":"2024","journal-title":"arXiv preprint arXiv:2410.20011"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref28","volume-title":"Simple english wikipedia.","year":"2025"},{"key":"ref29","volume-title":"Decoder-only transformers: The workhorse of generative 11 ms","author":"Wolfe","year":"2024"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/tmc.2026.3674827"},{"key":"ref31","article-title":"On-device language models: A comprehensive review","author":"Xu","year":"2024","journal-title":"arXiv preprint arXiv:2409.00088"},{"key":"ref32","article-title":"Qwen3 technical report","volume-title":"arXiv preprint arXiv:2505.09388","author":"Yang","year":"2025"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.acl-demo.4"},{"key":"ref34","article-title":"Towards superior quantization accuracy: A layer-sensitive approach","author":"Zhang","year":"2025","journal-title":"arXiv preprint arXiv:2503.06518"},{"key":"ref35","article-title":"Vocabtailor: Dynamic vocabulary selection for downstream tasks in small language models","author":"Zhang","year":"2025","journal-title":"arXiv preprint arXiv:2508.15229"},{"key":"ref36","volume-title":"Challenging gpu dominance: When cpus outperform for on-device llm inference","author":"Zhang","year":"2025"},{"key":"ref37","volume-title":"Bertscore: Evaluating text generation with bert","author":"Zhang","year":"2020"}],"event":{"name":"2026 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS)","location":"Seoul, Korea, Republic of","start":{"date-parts":[[2026,4,26]]},"end":{"date-parts":[[2026,4,28]]}},"container-title":["2026 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11527204\/11527232\/11527274.pdf?arnumber=11527274","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,11]],"date-time":"2026-06-11T19:57:52Z","timestamp":1781207872000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11527274\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4,26]]},"references-count":37,"URL":"https:\/\/doi.org\/10.1109\/ispass69572.2026.00053","relation":{},"subject":[],"published":{"date-parts":[[2026,4,26]]}}}