{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,29]],"date-time":"2026-01-29T23:21:47Z","timestamp":1769728907303,"version":"3.49.0"},"reference-count":25,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,5,19]],"date-time":"2025-05-19T00:00:00Z","timestamp":1747612800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,5,19]],"date-time":"2025-05-19T00:00:00Z","timestamp":1747612800000},"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":[[2025,5,19]]},"DOI":"10.1109\/saci66288.2025.11030191","type":"proceedings-article","created":{"date-parts":[[2025,6,23]],"date-time":"2025-06-23T17:24:33Z","timestamp":1750699473000},"page":"000211-000216","source":"Crossref","is-referenced-by-count":1,"title":["Turning LLM Activations Quantization-Friendly"],"prefix":"10.1109","author":[{"given":"Patrik","family":"Czak\u00f3","sequence":"first","affiliation":[{"name":"Doctoral School of Applied, Informatics and Applied, Mathematics, Obuda University,Budapest,Hungary"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"G\u00e1bor","family":"Kert\u00e9sz","sequence":"additional","affiliation":[{"name":"Obuda University,John von Neumann Faculty of Informatics,Budapest,Hungary"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"S\u00e1ndor","family":"Sz\u00e9n\u00e1si","sequence":"additional","affiliation":[{"name":"Obuda University,John von Neumann Faculty of Informatics,Budapest,Hungary"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","volume-title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","author":"Touvron","year":"2023"},{"key":"ref2","volume-title":"The Llama 3 Herd of Models.","author":"Grattafiori","year":"2024"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.12700\/APH.20.5.2023.5.11"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.12700\/aph.22.1.2025.1.2"},{"key":"ref5","volume-title":"Model Compression and Efficient Inference for Large Language Models: A Survey","author":"Wang","year":"2024"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00704"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr.2018.00286"},{"key":"ref8","article-title":"LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale","volume-title":"Advances in Neural Information Processing Systems","volume":"35","author":"Dettmers","year":"2022"},{"key":"ref9","first-page":"38087","article-title":"SmoothQuant: Accurate and Efficient PostTraining Quantization for Large Language Models","volume-title":"Proceedings of Machine Learning Research","volume":"202","author":"Xiao","year":"2023"},{"key":"ref10","volume-title":"Mitigating Quantization Errors Due to Activation Spikes in GLU-Based LLMs","author":"Yang","year":"2024"},{"key":"ref11","volume-title":"DuQuant: Distributing Outliers via Dual Transformation Makes Stronger Quantized LLMs","author":"Lin","year":"2024"},{"key":"ref12","volume-title":"Atom: Low-bit Quantization for Efficient and Accurate LLM Serving","author":"Zhao","year":"2024"},{"key":"ref13","volume-title":"QuaRot: Outlier-Free 4-Bit Inference in Rotated LLMs","author":"Ashkboos","year":"2024"},{"key":"ref14","article-title":"OmniQuant: Omnidirectionally calibrated quantization for large language models","volume-title":"12th International Conference on Learning Representations, ICLR 2024","author":"Shao","year":"2024"},{"key":"ref15","volume-title":"SpinQuant: LLM quantization with learned rotations","author":"Liu","year":"2024"},{"key":"ref16","article-title":"SPQR: A sparse-quantized representation for near-lossless 11 m weight compression","volume-title":"12th International Conference on Learning Representations, ICLR 2024","author":"Dettmers","year":"2024"},{"key":"ref17","first-page":"23901","article-title":"SqueezeLLM: Dense-and-Sparse Quantization","volume-title":"Proceedings of Machine Learning Research","volume":"235","author":"Kim","year":"2024"},{"key":"ref18","first-page":"48630","article-title":"QuIP#: Even Better LLM Quantization with Hadamard Incoherence and Lattice Codebooks","volume-title":"Proceedings of Machine Learning Research","volume":"235","author":"Tseng","year":"2024"},{"key":"ref19","volume-title":"EfficientQAT: Efficient Quantization-Aware Training for Large Language Models","author":"Chen","year":"2024"},{"key":"ref20","article-title":"OPTQ: Accurate quantization for generative pretrained transformers","volume-title":"11th International Conference on Learning Representations, ICLR 2023","author":"Frantar","year":"2023"},{"key":"ref21","volume-title":"FlatQuant: Flatness Matters for LLM Quantization","author":"Sun","year":"2024"},{"key":"ref22","article-title":"Pointer Sentinel Mixture Models","volume-title":"International Conference on Learning Representations","author":"Merity"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.emnlp-demos.6"},{"key":"ref24","first-page":"8026","article-title":"PyTorch: An imperative style, high-performance deep learning library","volume-title":"Proceedings of the 33rd International Conference on Neural Information Processing Systems","author":"Paszke"},{"key":"ref25","volume-title":"Mistral 7B","author":"Jiang","year":"2023"}],"event":{"name":"2025 IEEE 19th International Symposium on Applied Computational Intelligence and Informatics (SACI)","location":"Timisoara, Romania","start":{"date-parts":[[2025,5,19]]},"end":{"date-parts":[[2025,5,24]]}},"container-title":["2025 IEEE 19th International Symposium on Applied Computational Intelligence and Informatics (SACI)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11030063\/11030073\/11030191.pdf?arnumber=11030191","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,24]],"date-time":"2025-06-24T06:18:05Z","timestamp":1750745885000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11030191\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,5,19]]},"references-count":25,"URL":"https:\/\/doi.org\/10.1109\/saci66288.2025.11030191","relation":{},"subject":[],"published":{"date-parts":[[2025,5,19]]}}}