{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,9]],"date-time":"2026-07-09T02:39:43Z","timestamp":1783564783041,"version":"3.55.0"},"reference-count":115,"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\/100000001","name":"National Science Foundation","doi-asserted-by":"publisher","award":["IIS-2212046"],"award-info":[{"award-number":["IIS-2212046"]}],"id":[{"id":"10.13039\/100000001","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\/iccv51701.2025.00474","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"4983-4995","source":"Crossref","is-referenced-by-count":1,"title":["Importance-Based Token Merging for Efficient Image and Video Generation"],"prefix":"10.1109","author":[{"given":"Haoyu","family":"Wu","sequence":"first","affiliation":[{"name":"Stony Brook University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jingyi","family":"Xu","sequence":"additional","affiliation":[{"name":"Stony Brook University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hieu","family":"Le","sequence":"additional","affiliation":[{"name":"EPFL"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Dimitris","family":"Samaras","sequence":"additional","affiliation":[{"name":"Stony Brook University"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"ediff-i: Text-to-image diffusion models with an ensemble of expert denoisers","author":"Balaji","year":"2022","journal-title":"arXiv preprint"},{"key":"ref2","article-title":"Stable video diffusion: Scaling latent video diffusion models to large datasets","author":"Blattmann","year":"2023","journal-title":"arXiv preprint"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/CVPRW59228.2023.00484"},{"key":"ref4","article-title":"Token merging: Your vit but faster","volume":"abs\/2210.09461","author":"Bolya","year":"2022","journal-title":"ArXiv"},{"key":"ref5","volume-title":"Video generation models as world simulators","author":"Brooks","year":"2024"},{"key":"ref6","author":"Chen","year":"2023","journal-title":"Pixart-alpha: Fast training of diffusion transformer for photorealistic text-to-image synthesis"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.02636"},{"key":"ref8","article-title":"Delta-dit: A training-free acceleration method tailored for diffusion transformers","author":"Chen","year":"2024","journal-title":"arXiv preprint"},{"key":"ref9","first-page":"19974","article-title":"Chasing sparsity in vision transformers: An end-to-end exploration","volume":"34","author":"Chen","year":"2021","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01776"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v39i15.33782"},{"key":"ref12","first-page":"8780","article-title":"Diffusion models beat gans on image synthesis","volume":"34","author":"Dhariwal","year":"2021","journal-title":"Advances in neural information processing systems"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/ICRA46639.2022.9811809"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.52202\/075280-0731"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-20083-0_24"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00815"},{"key":"ref17","article-title":"Animatediff: Animate your personalized text-to-image diffusion models without specific tuning","author":"Guo","year":"2023","journal-title":"arXiv preprint"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00798"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1016\/j.autcon.2022.104614"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/ICCVW60793.2023.00085"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72998-0_12"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.52202\/075280-0580"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.emnlp-main.595"},{"key":"ref24","article-title":"Gans trained by a two time-scale update rule converge to a local nash equilibrium","volume":"30","author":"Heusel","year":"2017","journal-title":"Advances in neural information processing systems"},{"key":"ref25","article-title":"Classifier-free diffusion guidance","author":"Ho","year":"2022","journal-title":"arXiv preprint"},{"key":"ref26","first-page":"6840","article-title":"Denoising diffusion probabilistic models","volume":"33","author":"Ho","year":"2020","journal-title":"Advances in neural information processing systems"},{"key":"ref27","first-page":"13916","article-title":"Make-an-audio: Text-to-audio generation with prompt-enhanced diffusion models","volume-title":"International Conference on Machine Learning","author":"Huang","year":"2023"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02060"},{"key":"ref29","article-title":"Categorical reparameterization with gumbel-softmax","author":"Jang","year":"2016","journal-title":"arXiv preprint"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01300"},{"key":"ref31","article-title":"Ryoo, and Tian Xie. Adaptive caching for faster video generation with diffusion transformers","author":"Kahatapitiya","year":"2024","journal-title":"arXiv preprint"},{"key":"ref32","first-page":"91","article-title":"Objectcentric diffusion for efficient video editing","volume-title":"European Conference on Computer Vision","author":"Kahatapitiya","year":"2025"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.52202\/079017-1679"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1109\/MIPR62202.2024.00019"},{"key":"ref35","author":"Kim","year":"2023","journal-title":"On architectural compression of text-to-image diffusion models"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/WACV57701.2024.00141"},{"key":"ref37","article-title":"Diffwave: A versatile diffusion model for audio synthesis","author":"Kong","year":"2020","journal-title":"arXiv preprint"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-20083-0_37"},{"key":"ref39","article-title":"PKU-Yuan Lab and Tuzhan AI etc","volume-title":"Open-sora-plan","year":"2024"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.52202\/079017-1713"},{"key":"ref41","volume-title":"xformers: A modular and hackable transformer modelling library","author":"Lefaudeux","year":"2022"},{"key":"ref42","article-title":"3ddesigner: Towards photorealistic 3d object generation and editing with textguided diffusion models","author":"Li","year":"2022","journal-title":"arXiv preprint"},{"key":"ref43","article-title":"Sait: Sparse vision transformers through adaptive token pruning","author":"Li","year":"2022","journal-title":"arXiv preprint"},{"key":"ref44","article-title":"Faster diffusion: Rethinking the role of unet encoder in diffusion models","author":"Li","year":"2023","journal-title":"CoRR"},{"key":"ref45","article-title":"Tokenpacker: Efficient visual projector for multimodal 11 m","author":"Li","year":"2024","journal-title":"arXiv preprint"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01608"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00715"},{"key":"ref48","article-title":"Snapfusion: Text-to-image diffusion model on mobile devices within two seconds","volume":"36","author":"Li","year":"2024","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref49","article-title":"Looking backward: Streaming video-to-video translation with feature banks","author":"Liang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref50","article-title":"Not all patches are what you need: Expediting vision transformers via token reorganizations","author":"Liang","year":"2022","journal-title":"arXiv preprint"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-10602-1_48"},{"key":"ref52","article-title":"Pseudo numerical methods for diffusion models on manifolds","author":"Liu","year":"2022","journal-title":"arXiv preprint"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1109\/iccv51070.2023.00853"},{"key":"ref54","article-title":"Instaflow: One step is enough for high-quality diffusion-based text-to-image generation","volume-title":"The Twelfth International Conference on Learning Representations","author":"Liu","year":"2023"},{"key":"ref55","article-title":"Syncdreamer: Generating multiview-consistent images from a single-view image","author":"Liu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1109\/WACV57701.2024.00264"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00996"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00951"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.52202\/068431-0418"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.1007\/s11633-025-1562-4"},{"key":"ref61","author":"Lu","year":"2024","journal-title":"Toma: Token merging with attention for diffusion models"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00286"},{"key":"ref63","article-title":"Latent consistency models: Synthesizing highresolution images with few-step inference","author":"Luo","year":"2023","journal-title":"arXiv preprint"},{"key":"ref64","article-title":"Wong. Fastercache: Training-free video diffusion model acceleration with high quality","author":"Lv","year":"2024","journal-title":"arXiv preprint"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01492"},{"key":"ref66","doi-asserted-by":"publisher","DOI":"10.1109\/WACV56688.2023.00010"},{"key":"ref67","article-title":"Sdedit: Guided image synthesis and editing with stochastic differential equations","author":"Meng","year":"2021","journal-title":"arXiv preprint"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01374"},{"key":"ref69","article-title":"Glide: Towards photorealistic image generation and editing with text-guided diffusion models","author":"Nichol","year":"2021","journal-title":"arXiv preprint"},{"key":"ref70","article-title":"Point-e: A system for generating 3d point clouds from complex prompts","author":"Nichol","year":"2022","journal-title":"arXiv preprint"},{"key":"ref71","volume-title":"Interpretability-aware redundancy reduction for vision transformers","author":"Pan","year":"2023"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00387"},{"key":"ref73","article-title":"W\u00fcrstchen: An efficient architecture for large-scale text-to-image diffusion models","author":"Pernias","year":"2023","journal-title":"arXiv preprint"},{"key":"ref74","article-title":"Dreamfusion: Text-to-3d using 2d diffusion","author":"Poole","year":"2022","journal-title":"arXiv preprint"},{"key":"ref75","first-page":"8748","article-title":"Learning transferable visual models from natural language supervision","volume-title":"International conference on machine learning","author":"Radford","year":"2021"},{"key":"ref76","article-title":"Hierarchical text-conditional image generation with clip latents","volume":"abs\/2204.06125","author":"Ramesh","year":"2022","journal-title":"ArXiv"},{"key":"ref77","first-page":"13937","article-title":"Dynamicvit: Efficient vision transformers with dynamic token sparsification","volume":"34","author":"Rao","year":"2021","journal-title":"Advances in neural information processing systems"},{"key":"ref78","article-title":"Learning to merge tokens in vision transformers","author":"Renggli","year":"2022","journal-title":"arXiv preprint"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref80","article-title":"Photorealistic text-to-image diffusion models with deep language understanding","author":"Saharia","year":"2022","journal-title":"arXiv preprint"},{"key":"ref81","article-title":"Progressive distillation for fast sampling of diffusion models","author":"Salimans","year":"2022","journal-title":"arXiv preprint"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73016-0_6"},{"key":"ref83","article-title":"Zero123++: a single image to consistent multi-view diffusion base model","author":"Shi","year":"2023","journal-title":"arXiv preprint"},{"key":"ref84","article-title":"Make-a-video: Text-to-video generation without text-video data","author":"Singer","year":"2022","journal-title":"arXiv preprint"},{"key":"ref85","article-title":"Todo: Token downsampling for efficient generation of highresolution images","author":"Smith","year":"2024","journal-title":"arXiv preprint"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73464-9_20"},{"key":"ref87","article-title":"Temporal dynamic quantization for diffusion models","volume":"36","author":"So","year":"2024","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref88","first-page":"2256","article-title":"Deep unsupervised learning using nonequilibrium thermodynamics","volume-title":"International conference on machine learning","author":"Sohl-Dickstein","year":"2015"},{"key":"ref89","article-title":"Less is more: A simple yet effective token reduction method for efficient multimodal 11 ms","author":"Song","year":"2024","journal-title":"arXiv preprint"},{"key":"ref90","article-title":"Denoising diffusion implicit models","author":"Song","year":"2020","journal-title":"arXiv preprint"},{"key":"ref91","article-title":"Consistency models","author":"Song","year":"2023","journal-title":"arXiv preprint"},{"key":"ref92","article-title":"Accelerating transformers with spectrum-preserving token merging","author":"Tran","year":"2024","journal-title":"arXiv preprint"},{"key":"ref93","first-page":"16070","article-title":"Zerotprune: Zero-shot token pruning through leveraging of the attention graph in pre-trained transformers","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Wang","year":"2024"},{"key":"ref94","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01522"},{"key":"ref95","article-title":"Quest: Low-bit diffusion model quantization via efficient selective finetuning","author":"Wang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref96","article-title":"Sparsedm: Toward sparse efficient diffusion models","author":"Wang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref97","article-title":"Efficient visual transformer by learnable token merging","author":"Wang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref98","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00733"},{"key":"ref99","doi-asserted-by":"publisher","DOI":"10.1109\/TIP.2003.819861"},{"key":"ref100","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00208"},{"key":"ref101","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00594"},{"key":"ref102","article-title":"Ppt: Token pruning and pooling for efficient vision transformers","author":"Wu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref103","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01760"},{"key":"ref104","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v36i3.20202"},{"key":"ref105","article-title":"Raphael: Text-to-image generation via large mixture of diffusion paths","volume":"36","author":"Xue","year":"2024","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref106","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.02160"},{"key":"ref107","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01054"},{"key":"ref108","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01082"},{"key":"ref109","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00068"},{"key":"ref110","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.02065"},{"key":"ref111","article-title":"Dynamic diffusion transformer","author":"Zhao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref112","article-title":"Real-time video generation with pyramid attention broadcast","author":"Zhao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref113","article-title":"Mobilediffusion: Subsecond text-to-image generation on mobile devices","author":"Zhao","year":"2023","journal-title":"arXiv preprint"},{"key":"ref114","author":"Zheng","year":"2024","journal-title":"Open-sora: Democratizing efficient video production for all"},{"key":"ref115","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-20083-0_26"}],"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\/11445843.pdf?arnumber=11445843","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:05:03Z","timestamp":1777611903000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11445843\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":115,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.00474","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}