{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,8]],"date-time":"2026-05-08T16:18:39Z","timestamp":1778257119141,"version":"3.51.4"},"reference-count":86,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","license":[{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/legalcode"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Access"],"published-print":{"date-parts":[[2025]]},"DOI":"10.1109\/access.2025.3568702","type":"journal-article","created":{"date-parts":[[2025,5,9]],"date-time":"2025-05-09T18:01:29Z","timestamp":1746813689000},"page":"81917-81932","source":"Crossref","is-referenced-by-count":3,"title":["Addressing Activation Outliers in LLMs: A Systematic Review of Post-Training Quantization Techniques"],"prefix":"10.1109","volume":"13","author":[{"ORCID":"https:\/\/orcid.org\/0009-0007-9187-3331","authenticated-orcid":false,"given":"Patrik","family":"Czak\u00f3","sequence":"first","affiliation":[{"name":"Doctoral School of Applied Informatics and Applied Mathematics, Óbuda University, Budapest, Hungary"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8845-8301","authenticated-orcid":false,"given":"G\u00e1bor","family":"Kert\u00e9sz","sequence":"additional","affiliation":[{"name":"John von Neumann Faculty of Informatics, Óbuda University, Budapest, Hungary"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7292-0717","authenticated-orcid":false,"given":"S\u00e1ndor","family":"Sz\u00e9n\u00e1si","sequence":"additional","affiliation":[{"name":"John von Neumann Faculty of Informatics, Óbuda University, Budapest, Hungary"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Scaling laws for neural language models","author":"Kaplan","year":"2020","journal-title":"arXiv:2001.08361"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.12700\/APH.20.5.2023.5.11"},{"key":"ref3","article-title":"A survey on model compression for large language models","author":"Zhu","year":"2023","journal-title":"arXiv:2308.07633"},{"key":"ref4","article-title":"Model compression and efficient inference for large language models: A survey","author":"Wang","year":"2024","journal-title":"arXiv:2402.09748"},{"key":"ref5","article-title":"Pruning vs quantization: Which is better?","volume-title":"NeurIPS","author":"Kuzmin"},{"key":"ref6","first-page":"41","article-title":"Accurate LoRAFinetuning quantization of LLMs via information retention","volume-title":"Proc. PMLR","volume":"235","author":"Qin"},{"key":"ref7","article-title":"QuIP: 2-bit quantization of large language models with guarantees","volume-title":"NeurIPS","volume":"36","author":"Chee"},{"key":"ref8","first-page":"48","article-title":"QuIP#: Even betterLLMquantization with Hadamard incoherence and lattice codebooks","volume-title":"PMLR","volume":"235","author":"Tseng"},{"key":"ref9","article-title":"SPQR: A sparse-quantized representation for nearlossless llm weight compression","volume-title":"ICLR","author":"Dettmers"},{"key":"ref10","first-page":"23901","article-title":"SqueezeLLM: Dense-and-sparse quantization","volume-title":"Proc. PMLR","volume":"235","author":"Kim"},{"key":"ref11","article-title":"EfficientQAT: Efficient quantization-aware training for large language models","author":"Chen","year":"2024","journal-title":"arXiv:2407.11062"},{"key":"ref12","first-page":"20023","article-title":"BiLLM: Pushing the limit of post-training quantization for LLMs","volume-title":"Proc. PMLR","volume":"235","author":"Huang"},{"key":"ref13","article-title":"ARB-LLM: Alternating refined binarizations for large language models","volume-title":"ICLR","author":"Li"},{"key":"ref14","article-title":"A survey of low-bit large language models: Basics, systems, and algorithms","author":"Gong","year":"2024","journal-title":"arXiv:2409.16694"},{"key":"ref15","article-title":"Art and science of quantizing large-scale models: A comprehensive overview","author":"Wang","year":"2024","journal-title":"arXiv:2409.11650"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijsu.2021.105906"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr.2018.00286"},{"key":"ref18","article-title":"A white paper on neural network quantization","author":"Nagel","year":"2021","journal-title":"arXiv:2106.08295"},{"key":"ref19","article-title":"LUT-GEMM: Quantized matrix multiplication based on LUTs for efficient inference in large-scale generative language models","volume-title":"ICLR","author":"Park"},{"key":"ref20","first-page":"196","article-title":"ATOM: Low-bit quantization for efficient and accurate llm serving","volume-title":"Proc. MLSys","author":"Zhao"},{"key":"ref21","article-title":"QServe: W4A8KV4 quantization and system co-design for efficient LLM serving","author":"Lin","year":"2024","journal-title":"arXiv:2405.04532"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.197"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1145\/1498765.1498785"},{"key":"ref24","article-title":"LLM.int8(): 8-bit matrix multiplication for transformers at scale","volume-title":"NeurIPS","volume":"35","author":"Dettmers"},{"key":"ref25","first-page":"38087","article-title":"SmoothQuant: Accurate and efficient post-training quantization for large language models","volume-title":"Proc. PMLR","volume":"202","author":"Xiao"},{"key":"ref26","article-title":"Mitigating quantization errors due to activation spikes in GLU-based LLMs","author":"Yang","year":"2024","journal-title":"arXiv:2405.14428"},{"key":"ref27","article-title":"PrefixQuant: Eliminating outliers by prefixed tokens for large language models quantization","author":"Chen","year":"2024","journal-title":"arXiv:2410.05265"},{"key":"ref28","article-title":"DuQuant: Distributing outliers via dual transformation makes stronger quantized LLMs","author":"Lin","year":"2024","journal-title":"arXiv:2406.01721"},{"key":"ref29","article-title":"FPTQ: Fine-grained post-training quantization for large language models","author":"Li","year":"2023","journal-title":"arXiv:2308.15987"},{"key":"ref30","article-title":"RPTQ: Reorder-based post-training quantization for large language models","author":"Yuan","year":"2023","journal-title":"arXiv:2304.01089"},{"key":"ref31","article-title":"Outlier suppression: Pushing the limit of low-bit transformer language models","volume-title":"NeurIPS","volume":"35","author":"Wei"},{"key":"ref32","article-title":"CBQ: Cross-block quantization for large language models","author":"Ding","year":"2023","journal-title":"arXiv:2312.07950"},{"key":"ref33","article-title":"I-LLM: Efficient integer-only inference for fully-quantized low-bit large language models","author":"Hu","year":"2024","journal-title":"arXiv:2405.17849"},{"key":"ref34","article-title":"No token left behind: Reliable KV cache compression via importance-aware mixed precision quantization","author":"Yang","year":"2024","journal-title":"arXiv:2402.18096"},{"key":"ref35","article-title":"QuaRot: Outlier-free 4-bit inference in rotated LLMs","author":"Ashkboos","year":"2024","journal-title":"arXiv:2404.00456"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/MWSCAS60917.2024.10658840"},{"key":"ref37","article-title":"QAQ: Quality adaptive quantization for LLM KV cache","author":"Dong","year":"2024","journal-title":"arXiv:2403.04643"},{"key":"ref38","article-title":"SKVQ: Sliding-window key and value cache quantization for large language models","author":"Duanmu","year":"2024","journal-title":"arXiv:2405.06219"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-naacl.173"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-naacl.173"},{"key":"ref41","article-title":"ZipCache: Accurate and efficient KV cache quantization with salient token identification","author":"He","year":"2024","journal-title":"arXiv:2405.14256"},{"key":"ref42","article-title":"KVQuant: Towards 10 million context length LLM inference with KV cache quantization","author":"Hooper","year":"2024","journal-title":"arXiv:2401.18079"},{"key":"ref43","article-title":"GEAR: An efficient KV cache compression recipe for near-lossless generative inference of LLM","author":"Kang","year":"2024","journal-title":"arXiv:2403.05527"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.910"},{"key":"ref45","article-title":"AMXFP4: Taming activation outliers with asymmetric microscaling floating-point for 4-bit LLM inference","author":"Lee","year":"2024","journal-title":"arXiv:2411.09909"},{"key":"ref46","article-title":"QLLM: Accurate and efficient low-bitwidth quantization for large language models","volume-title":"ICLR","author":"Liu"},{"key":"ref47","article-title":"COMET: Towards partical W4A4KV4 LLMs serving","author":"Liu","year":"2024","journal-title":"arXiv:2410.12168"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.133"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.39"},{"key":"ref50","article-title":"CrossQuant: A post-training quantization method with smaller quantization kernel for precise large language model compression","author":"Liu","year":"2024","journal-title":"arXiv:2410.07505"},{"key":"ref51","first-page":"32332","article-title":"KIVI: A tuningfree asymmetric 2bit quantization for KV cache","volume-title":"Proc. PMLR","volume":"235","author":"Liu"},{"key":"ref52","article-title":"SpinQuant: LLM quantization with learned rotations","author":"Liu","year":"2024","journal-title":"arXiv:2405.16406"},{"key":"ref53","article-title":"AffineQuant: Affine transformation quantization for large language models","volume-title":"ICLR","author":"Ma"},{"key":"ref54","article-title":"Pyramid vector quantization for LLMs","author":"van der Ouderaa","year":"2024","journal-title":"arXiv:2410.16926"},{"key":"ref55","article-title":"OmniQuant: Omnidirectionally calibrated quantization for large language models","volume-title":"ICLR","author":"Shao"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.134"},{"key":"ref57","article-title":"FlatQuant: Flatness matters for LLM quantization","author":"Sun","year":"2024","journal-title":"arXiv:2410.09426"},{"key":"ref58","article-title":"AlignedKV: Reducing memory access of KV-cache with precision-aligned quantization","author":"Tan","year":"2024","journal-title":"arXiv:2409.16546"},{"key":"ref59","article-title":"AsymKV: Enabling 1-bit quantization of KV cache with layer-wise asymmetric quantization configurations","author":"Tao","year":"2024","journal-title":"arXiv:2410.13212"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.102"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i17.29908"},{"key":"ref62","article-title":"ZeroQuant: Efficient and affordable post-training quantization for large-scale transformers","volume-title":"NeurIPS","volume":"35","author":"Yao"},{"key":"ref63","article-title":"Rotated runtime smooth: Training-free activation smoother for accurate INT4 inference","author":"Yi","year":"2024","journal-title":"arXiv:2409.20361"},{"key":"ref64","article-title":"WKVQuant: Quantizing weight and key\/value cache for large language models gains more","author":"Yue","year":"2024","journal-title":"arXiv:2402.12065"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v39i21.34385"},{"key":"ref66","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.617"},{"key":"ref67","article-title":"KV cache is 1 bit per channel: Efficient large language model inference with coupled quantization","author":"Zhang","year":"2024","journal-title":"arXiv:2405.03917"},{"key":"ref68","first-page":"7356","article-title":"FlattenQuant: Breaking through the inference computebound for large language models with per-tensor quantization","volume-title":"Proc. LREC-COLING","author":"Zhang"},{"key":"ref69","article-title":"QQQ: Quality quattuor-bit quantization for large language models","author":"Zhang","year":"2024","journal-title":"arXiv:2406.09904"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.122"},{"key":"ref71","first-page":"62978","article-title":"BiE: Bi-exponent block floating-point for large language models quantization","volume-title":"Proc. PMLR","volume":"235","author":"Zou"},{"key":"ref72","article-title":"BiSup: Bidirectional quantization error suppression for large language models","author":"Zou","year":"2024","journal-title":"arXiv:2405.15346"},{"key":"ref73","article-title":"LoRA: Low-rank adaptation of large language models","volume-title":"ICLR","author":"Hu"},{"key":"ref74","article-title":"Efficient Riemannian optimization on the Stiefel manifold via the Cayley transform","volume-title":"ICLR","author":"Li"},{"key":"ref75","article-title":"H2O: Heavyhitter Oracle for efficient generative inference of large language models","volume-title":"NeuIPS","author":"Zhang"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.acl-long.418"},{"key":"ref77","article-title":"OPTQ: Accurate quantization for generative pretrained transformers","volume-title":"ICLR","author":"Frantar"},{"key":"ref78","article-title":"Microscaling data formats for deep learning","author":"Rouhani","year":"2023","journal-title":"arXiv:2310.10537"},{"key":"ref79","article-title":"Pointer sentinel mixture models","volume-title":"ICLR","author":"Merity"},{"key":"ref80","article-title":"LLaMA: Open and efficient foundation language models","author":"Touvron","year":"2023","journal-title":"arXiv:2302.13971"},{"key":"ref81","article-title":"Llama 2: Open foundation and fine-tuned chat models","author":"Touvron","year":"2023","journal-title":"arXiv:2307.09288"},{"key":"ref82","article-title":"The llama 3 herd of models","author":"Grattafiori","year":"2024","journal-title":"arXiv:2407.21783"},{"key":"ref83","article-title":"Training verifiers to solve math word problems","author":"Cobbe","year":"2021","journal-title":"arXiv:2110.14168"},{"key":"ref84","article-title":"OPT: Open pre-trained transformer language models","author":"Zhang","year":"2022","journal-title":"arXiv:2205.01068"},{"key":"ref85","article-title":"Mistral 7B","author":"Jiang","year":"2023","journal-title":"arXiv:2310.06825"},{"key":"ref86","article-title":"Low-rank quantization-aware training for LLMs","author":"Bondarenko","year":"2024","journal-title":"arXiv:2406.06385"}],"container-title":["IEEE Access"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/6287639\/10820123\/10994764.pdf?arnumber=10994764","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,5,16]],"date-time":"2025-05-16T17:46:54Z","timestamp":1747417614000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10994764\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025]]},"references-count":86,"URL":"https:\/\/doi.org\/10.1109\/access.2025.3568702","relation":{},"ISSN":["2169-3536"],"issn-type":[{"value":"2169-3536","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025]]}}}