{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:02:03Z","timestamp":1777888923157,"version":"3.51.4"},"reference-count":78,"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"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,19]]},"DOI":"10.1109\/iccv51701.2025.01902","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"20456-20466","source":"Crossref","is-referenced-by-count":0,"title":["Representation Shift: Unifying Token Compression with Flashattention"],"prefix":"10.1109","author":[{"given":"Joonmyung","family":"Choi","sequence":"first","affiliation":[{"name":"Korea University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sanghyeok","family":"Lee","sequence":"additional","affiliation":[{"name":"Korea University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Byungoh","family":"Ko","sequence":"additional","affiliation":[{"name":"Korea University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Eunseo","family":"Kim","sequence":"additional","affiliation":[{"name":"Korea University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jihyung","family":"Kil","sequence":"additional","affiliation":[{"name":"Adobe Research"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hyunwoo J.","family":"Kim","sequence":"additional","affiliation":[{"name":"KAIST"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","author":"Alif","year":"2025","journal-title":"Yolov12: A breakdown of the key architectural features"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2017.618"},{"key":"ref3","author":"Beltagy","year":"2004","journal-title":"Longformer: The long-document transformer"},{"key":"ref4","article-title":"Token merging: Your vit but faster","author":"Bolya","year":"2023","journal-title":"ICLR"},{"key":"ref5","article-title":"The pagerank citation ranking: bringing order to the web","author":"Brin","year":"1998","journal-title":"ASIS"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7298698"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01587"},{"key":"ref8","article-title":"End-toend object detection with transformers","author":"Carion","year":"2020","journal-title":"ECCV. Springer"},{"key":"ref9","article-title":"Scatterbrain: Unifying sparse and lowrank attention","author":"Chen","year":"2021","journal-title":"NeurIPS"},{"key":"ref10","article-title":"Collecting highly parallel data for paraphrase evaluation","volume-title":"Proceedings of the 49th annual meeting of the association for computational linguistics: human language technologies","author":"Chen","year":"2011"},{"key":"ref11","article-title":"Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks","author":"Chen","year":"2024","journal-title":"CVPR"},{"key":"ref12","article-title":"Perpixel classification is not all you need for semantic segmentation","author":"Cheng","year":"2021","journal-title":"NeurIPS"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01776"},{"key":"ref14","article-title":"Rethinking attention with performers","author":"Choromanski","year":"2021","journal-title":"ICLR"},{"key":"ref15","article-title":"Twins: Revisiting the design of spatial attention in vision transformers","author":"Chu","year":"2021","journal-title":"NeurIPS"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1189"},{"key":"ref17","article-title":"Vision transformers need registers","author":"Darcet","year":"2024","journal-title":"ICLR"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2009.5206848"},{"key":"ref19","article-title":"An image is worth 16 \\times 16 words: Transformers for image recognition at scale","author":"Dosovitskiy","year":"2021","journal-title":"ICLR"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.01204"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref22","article-title":"Transformers are rnns: Fast autoregressive transformers with linear attention","author":"Katharopoulos","year":"2020","journal-title":"ICML"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02660"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1145\/3534678.3539260"},{"key":"ref25","article-title":"Reformer: The efficient transformer","author":"Kitaev","year":"2020","journal-title":"ICLR"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.00496"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01925"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v39i4.32474"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01490"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.01390"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.acl-long.29"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01826"},{"key":"ref33","article-title":"Not all patches are what you need: Expediting vision transformers via token reorganizations","author":"Liang","year":"2022","journal-title":"ICLR"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01386"},{"key":"ref35","article-title":"Efficient training of visual transformers with small datasets","author":"Liu","year":"2021","journal-title":"NeurIPS"},{"key":"ref36","article-title":"Vmamba: Visual state space model","author":"Liu","year":"2024","journal-title":"NeurIPS"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00986"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01167"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00996"},{"key":"ref40","article-title":"Mobilevit: lightweight, general-purpose, and mobile-friendly vision transformer","author":"Mehta","year":"2022","journal-title":"ICLR"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01199"},{"key":"ref42","article-title":"Dinov2: Learning robust visual features without supervision","author":"Oquab","year":"2023","journal-title":"TMLR"},{"key":"ref43","article-title":"IA-RED2: Interpretability-aware redundancy reduction for vision transformers","author":"Pan","year":"2021","journal-title":"NeurIPS"},{"key":"ref44","article-title":"Deepvideo-r1: Video reinforcement fine-tuning via difficulty-aware regressive grpo","author":"Park","year":"2025","journal-title":"arXiv preprint"},{"key":"ref45","article-title":"cosformer: Rethinking softmax in attention","author":"Qin","year":"2022","journal-title":"ICLR"},{"key":"ref46","article-title":"Dynamicvit: Efficient vision transformers with dynamic token sparsification","author":"Rao","year":"2021","journal-title":"NeurIPS"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-016-0987-1"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00353"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00717"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73220-1_19"},{"key":"ref51","article-title":"Videomae: Masked autoencoders are data-efficient learners for self-supervised video pre-training","author":"Tong","year":"2022","journal-title":"NeurIPS"},{"key":"ref52","article-title":"Training data-efficient image transformers & distillation through attention","author":"Touvron","year":"2021","journal-title":"ICML"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00010"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-20053-3_27"},{"key":"ref55","article-title":"Attention is all you need","author":"Vaswani","year":"2017","journal-title":"NeurIPS"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52733.2024.01521"},{"key":"ref57","author":"Wang","year":"2006","journal-title":"Linformer: Self-attention with linear complexity"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00061"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1007\/s41095-022-0274-8"},{"key":"ref60","article-title":"Videocomposer: Compositional video synthesis with motion controllability","volume-title":"NeurIPS","author":"Wang","year":"2023"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00863"},{"key":"ref62","article-title":"Internvideo: General video foundation models via generative and discriminative learning","author":"Wang","year":"2022","journal-title":"arXiv:2212"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v36i3.20158"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73013-9_23"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01548"},{"key":"ref66","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v35i16.17664"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.1145\/3123266.3123427"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.571"},{"key":"ref69","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01054"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00060"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.1109\/CVPRW63382.2024.00369"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00550"},{"key":"ref73","article-title":"Big bird: Transformers for longer sequences","author":"Zaheer","year":"2020","journal-title":"NeurIPS"},{"key":"ref74","article-title":"Exploring token pruning in vision state space models","author":"Zhan","year":"2024","journal-title":"NeurIPS"},{"key":"ref75","article-title":"Dino: Detr with improved denoising anchor boxes for end-to-end object detection","author":"Zhang","year":"2023","journal-title":"ICLR"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00681"},{"key":"ref77","article-title":"Vision mamba: Efficient visual representation learning with bidirectional state space model","author":"Zhu","year":"2024","journal-title":"ICML"},{"key":"ref78","article-title":"Deformable detr: Deformable transformers for end-to-end object detection","author":"Zhu","year":"2021","journal-title":"ICLR"}],"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\/11445453.pdf?arnumber=11445453","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:02:34Z","timestamp":1777611754000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11445453\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":78,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.01902","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}