{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:19:48Z","timestamp":1777889988233,"version":"3.51.4"},"reference-count":66,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100010877","name":"Shenzhen Science and Technology Innovation Commission","doi-asserted-by":"publisher","award":["SGDX20220530111405040"],"award-info":[{"award-number":["SGDX20220530111405040"]}],"id":[{"id":"10.13039\/501100010877","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,19]]},"DOI":"10.1109\/iccv51701.2025.01545","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"16641-16651","source":"Crossref","is-referenced-by-count":0,"title":["\u201cPrincipal Components\u201d Enable a New Language of Images"],"prefix":"10.1109","author":[{"given":"Xin","family":"Wen","sequence":"first","affiliation":[{"name":"University of Hong Kong"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Bingchen","family":"Zhao","sequence":"additional","affiliation":[{"name":"University of Edinburgh"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ismail","family":"Elezi","sequence":"additional","affiliation":[{"name":"Huawei London Research Centre"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jiankang","family":"Deng","sequence":"additional","affiliation":[{"name":"Imperial College London"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xiaojuan","family":"Qi","sequence":"additional","affiliation":[{"name":"University of Hong Kong"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","article-title":"FlexTok: Resampling images into 1d token sequences of flexible length","author":"Bachmann","year":"2025","journal-title":"ICML"},{"key":"ref2","author":"Bai","year":"2025","journal-title":"Qwen2.5VL technical report"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1126\/science.144.3617.424"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01103"},{"key":"ref5","article-title":"Muse: Text-to-image generation via masked generative Transformers","author":"Chang","year":"2023","journal-title":"ICML"},{"key":"ref6","author":"Chen","year":"2025","journal-title":"Masked autoencoders are effective tokenizers for diffusion models"},{"key":"ref7","author":"Chen","year":"2025","journal-title":"Diffusion autoencoders are scalable image tokenizers"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2009.5206848"},{"key":"ref9","article-title":"Diffusion models beat GANs on image synthesis","volume-title":"NeurIPS","author":"Dhariwal","year":"2021"},{"key":"ref10","author":"Dieleman","year":"2024","journal-title":"Diffusion is spectral autoregression"},{"key":"ref11","author":"Dieleman","year":"2025","journal-title":"Generative modelling in latent space"},{"key":"ref12","article-title":"An image is worth 16x16 words: Transformers for image recognition at scale","author":"Dosovitskiy","year":"2021","journal-title":"ICLR"},{"key":"ref13","article-title":"Adaptive length image tokenization via recurrent allocation","author":"Duggal","year":"2025","journal-title":"ICLR"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.01268"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1167\/7.1.10"},{"key":"ref16","author":"Ge","year":"2023","journal-title":"Planting a seed of vision in large language model"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.01270"},{"key":"ref18","author":"Hansen-Estruch","year":"2025","journal-title":"Learnings from scaling visual tokenizers for reconstruction and generation"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01553"},{"key":"ref20","article-title":"GANs trained by a two time-scale update rule converge to a local Nash equilibrium","author":"Heusel","year":"2017","journal-title":"NeurIPS"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1126\/science.1127647"},{"key":"ref22","author":"Ho","year":"2022","journal-title":"Classifier-free diffusion guidance"},{"key":"ref23","article-title":"Denoising diffusion probabilistic models","author":"Ho","year":"2020","journal-title":"NeurIPS"},{"key":"ref24","author":"Ho","year":"2023","journal-title":"Information-ordered bottlenecks for adaptive semantic compression"},{"key":"ref25","article-title":"Auto-encoding variational bayes","author":"Kingma","year":"2014","journal-title":"ICLR"},{"key":"ref26","article-title":"Matryoshka representation learning","author":"Kusupati","year":"2022","journal-title":"NeurIPS"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01123"},{"key":"ref28","article-title":"Return of unconditional generation: A self-supervised representation generation method","author":"Li","year":"2024","journal-title":"NeurIPS"},{"key":"ref29","article-title":"Autoregressive image generation without vector quantization","author":"Li","year":"2024","journal-title":"NeurIPS"},{"key":"ref30","article-title":"Imagefolder: Autoregressive image generation with folded tokens","author":"Li","year":"2025","journal-title":"ICLR"},{"key":"ref31","article-title":"Decoupled weight decay regularization","author":"Loshchilov","year":"2019","journal-title":"ICLR"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72980-5_2"},{"key":"ref33","article-title":"One-d-piece: Image tokenizer meets quality-controllable compression","author":"Miwa","year":"2025","journal-title":"ICML Tokenization Workshop"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1016\/0010-0285(77)90012-3"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1006\/cogp.1999.0728"},{"key":"ref36","article-title":"DINOv2: Learning robust visual features without supervision","author":"Oquab","year":"2024","journal-title":"TMLR"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00387"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01036"},{"key":"ref39","author":"Radford","year":"2018","journal-title":"Improving language understanding by generative pre-training"},{"key":"ref40","article-title":"Learning transferable visual models from natural language supervision","author":"Radford","year":"2021","journal-title":"ICML"},{"key":"ref41","author":"Ramanujan","year":"2024","journal-title":"When worse is better: Navigating the compression-generation tradeoff in visual tokenization"},{"key":"ref42","author":"Ramesh","year":"2022","journal-title":"Hierarchical text-conditional image generation with CLIP latents"},{"key":"ref43","article-title":"Learning ordered representations with nested dropout","author":"Rippel","year":"2014","journal-title":"ICML"},{"key":"ref44","author":"Rissanen","year":"2022","journal-title":"Generative modelling with inverse heat dissipation"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref46","article-title":"Improved techniques for training GANs","author":"Salimans","year":"2016","journal-title":"NeurIPS"},{"key":"ref47","article-title":"GLU variants improve Transformer","author":"Shazeer","journal-title":"arXiv:2002.05202"},{"key":"ref48","author":"Shlens","year":"2014","journal-title":"A tutorial on principal component analysis"},{"key":"ref49","article-title":"Denoising diffusion implicit models","author":"Song","year":"2021","journal-title":"ICLR"},{"key":"ref50","author":"Sun","year":"2024","journal-title":"Autoregressive model beats diffusion: Llama for scalable image generation"},{"key":"ref51","article-title":"Visual autoregressive modeling: Scalable image generation via next-scale prediction","author":"Tian","year":"2024","journal-title":"NeurIPS"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.1991.139758"},{"key":"ref53","article-title":"Pixel recurrent neural networks","author":"van den Oord","journal-title":"ICML 2016"},{"key":"ref54","article-title":"Conditional image generation with PixelCNN decoders","author":"van den Oord","year":"2016","journal-title":"NeurIPS"},{"key":"ref55","article-title":"Neural discrete representation learning","author":"van den Oord","journal-title":"NeurIPS 2017. 1, 2"},{"key":"ref56","article-title":"Chao. Learning deep Transformer models for machine translation","author":"Wang","year":"2019","journal-title":"ACL"},{"key":"ref57","article-title":"MaskBit: Embedding-free image generation via bit tokens","author":"Weber","year":"2024","journal-title":"TMLR"},{"key":"ref58","article-title":"Reconstruction vs","author":"Yao","year":"2025","journal-title":"generation: Taming optimization dilemma in latent diffusion models"},{"key":"ref59","article-title":"Vector-quantized image modeling with improved VQGAN","author":"Yu","year":"2022","journal-title":"ICLR"},{"key":"ref60","article-title":"Gundavarapu, Luca Versari, Kihyuk Sohn, David Minnen, Yong Cheng, Vighnesh Birodkar, Agrim Gupta, Xiuye Gu, Alexander G. Hauptmann, Boqing Gong, Ming-Hsuan Yang, Irfan Essa, David A. Ross, and Lu Jiang. Language model beats diffusion - tokenizer is key to visual generation","author":"Yu","year":"2024","journal-title":"ICLR"},{"key":"ref61","author":"Yu","year":"2024","journal-title":"Randomized autoregressive visual generation"},{"key":"ref62","article-title":"An image is worth 32 tokens for reconstruction and generation","author":"Yu","year":"2024","journal-title":"NeurIPS"},{"key":"ref63","article-title":"Representation alignment for generation: Training diffusion transformers is easier than you think","author":"Yu","year":"2025","journal-title":"ICLR"},{"key":"ref64","article-title":"Root mean square layer normalization","author":"Zhang","year":"2019","journal-title":"NeurIPS"},{"key":"ref65","author":"Zhao","year":"2024","journal-title":"\u03b5-VAE: Denoising as visual decoding"},{"key":"ref66","article-title":"Scaling the codebook size of VQ-GAN to 100,000 with a utilization rate of 99%","author":"Zhu","year":"2024","journal-title":"NeurIPS"}],"event":{"name":"2025 IEEE\/CVF International Conference on Computer Vision (ICCV)","location":"Honolulu, HI, USA","start":{"date-parts":[[2025,10,19]]},"end":{"date-parts":[[2025,10,25]]}},"container-title":["2025 IEEE\/CVF International Conference on Computer Vision (ICCV)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11443115\/11443287\/11445415.pdf?arnumber=11445415","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:23:46Z","timestamp":1777613026000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11445415\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":66,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.01545","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}