{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,18]],"date-time":"2026-06-18T06:04:38Z","timestamp":1781762678902,"version":"3.54.5"},"reference-count":32,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,4,20]],"date-time":"2026-04-20T00:00:00Z","timestamp":1776643200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,4,20]],"date-time":"2026-04-20T00:00:00Z","timestamp":1776643200000},"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":[[2026,4,20]]},"DOI":"10.23919\/date69613.2026.11539457","type":"proceedings-article","created":{"date-parts":[[2026,6,4]],"date-time":"2026-06-04T19:53:10Z","timestamp":1780602790000},"page":"1-7","source":"Crossref","is-referenced-by-count":1,"title":["SSR: Sparse Segment Reduction for Ternary GEMM Acceleration"],"prefix":"10.23919","author":[{"given":"Adeline","family":"Pittet","sequence":"first","affiliation":[{"name":"ETH Zurich,Systems Group,Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shien","family":"Zhu","sequence":"additional","affiliation":[{"name":"ETH Zurich,Systems Group,Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Val\u00e9rie","family":"Verdan","sequence":"additional","affiliation":[{"name":"ETH Zurich,Systems Group,Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Gustavo","family":"Alonso","sequence":"additional","affiliation":[{"name":"ETH Zurich,Systems Group,Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Large language models on small resource-constrained systems: Performance characterization, analysis and trade-offs","author":"Seymour","year":"2024"},{"key":"ref2","first-page":"87","article-title":"Awq: Activation-aware weight quantization for on-device llm compression and acceleration","volume-title":"Proceedings of Machine Learning and Systems","volume":"6","author":"Lin"},{"key":"ref3","article-title":"Tinyagent: Quantization-aware model compression and adaptation for on-device LLM agent deployment","volume-title":"Workshop on Efficient Systems for Foundation Models II @ ICML2024","author":"Kong"},{"key":"ref4","article-title":"Binaryconnect: Training deep neural networks with binary weights during propagations","volume-title":"Advances in Neural Information Processing Systems","volume":"28","author":"Courbariaux","year":"2015"},{"key":"ref5","article-title":"Ternary weight networks","author":"Li","year":"2022"},{"key":"ref6","article-title":"Trained ternary quantization","volume-title":"International Conference on Learning Representations","author":"Zhu"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/CVPRW53098.2021.00519"},{"key":"ref8","article-title":"The era of 1-bit llms: All large language models are in 1.58 bits","author":"Ma","year":"2025"},{"key":"ref9","first-page":"221","article-title":"Sparsity-control ternary weight networks","volume-title":"Neural Networks","volume":"145","author":"Deng","year":"2022"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/CVPRW.2017.61"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-70096-0_41"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/ASPDAC.2018.8297304"},{"key":"ref13","article-title":"Trained ternary quantization","author":"Zhu","year":"2017"},{"key":"ref14","article-title":"Csr sparse matrix rfc","year":"2020"},{"issue":"2","key":"ref15","first-page":"87","article-title":"An optimal storage format for sparse matrices","volume-title":"Information Processing Letters","volume":"90","author":"Montagne","year":"2004"},{"key":"ref16","article-title":"Sparse matrix formats","year":"2024"},{"key":"ref17","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2025.acl-long.457","article-title":"Bitnet.cpp: Efficient edge inference for ternary llms","author":"Wang","year":"2025"},{"key":"ref18","article-title":"An efficient matrix multiplication algorithm for accelerating inference in binary and ternary neural networks","author":"Dehghankar","year":"2025"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/TVLSI.2017.2767624"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v33i01.33013854"},{"key":"ref22","article-title":"Learning recurrent binary\/ternary weights","author":"Ardakani","year":"2019"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/SiPS.2017.8110021"},{"key":"ref24","first-page":"101 969","article-title":"Learn to be efficient: Build structured sparsity in large language models","volume-title":"Advances in Neural Information Processing Systems","volume":"37","author":"Zheng"},{"key":"ref25","article-title":"Fast matrix multiplications for lookup table-quantized llms","author":"Guo","year":"2025"},{"key":"ref26","first-page":"104039","article-title":"Tilenet: Hardware accelerator for ternary convolutional neural networks","volume-title":"Microprocessors and Microsystems","volume":"83","author":"Eetha","year":"2021"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1145\/3287624.3287721"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1145\/3270764"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/JSSC.2017.2778702"},{"key":"ref30","doi-asserted-by":"crossref","DOI":"10.1109\/ICPR56361.2022.9956533","article-title":"Fast matrix multiplication for binary and ternary cnns on arm cpu","author":"Trusov","year":"2022"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1145\/3410463.3414654"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1109\/CANDARW64572.2024.00031"},{"key":"ref33","article-title":"The era of 1-bit llms: All large language models are in 1.58 bits","author":"Ma","year":"2024"}],"event":{"name":"2026 Design, Automation & Test in Europe Conference (DATE)","location":"Verona, Italy","start":{"date-parts":[[2026,4,20]]},"end":{"date-parts":[[2026,4,22]]}},"container-title":["2026 Design, Automation & Test in Europe Conference (DATE)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11539023\/11539024\/11539457.pdf?arnumber=11539457","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,5]],"date-time":"2026-06-05T05:08:01Z","timestamp":1780636081000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11539457\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4,20]]},"references-count":32,"URL":"https:\/\/doi.org\/10.23919\/date69613.2026.11539457","relation":{},"subject":[],"published":{"date-parts":[[2026,4,20]]}}}