{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T03:31:35Z","timestamp":1777865495442,"version":"3.51.4"},"reference-count":45,"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\/501100012166","name":"National Key Research and Development Program of China","doi-asserted-by":"publisher","award":["2022YFC3602601,2022ZD0160102"],"award-info":[{"award-number":["2022YFC3602601,2022ZD0160102"]}],"id":[{"id":"10.13039\/501100012166","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\/iccv51701.2025.01904","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"20477-20486","source":"Crossref","is-referenced-by-count":0,"title":["ZipVL: Accelerating Vision-Language Models Through Dynamic Token Sparsity"],"prefix":"10.1109","author":[{"given":"Yefei","family":"He","sequence":"first","affiliation":[{"name":"Zhejiang University,China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Feng","family":"Chen","sequence":"additional","affiliation":[{"name":"The University of Adelaide,Australia"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jing","family":"Liu","sequence":"additional","affiliation":[{"name":"Monash University,ZIP Lab,Australia"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Wenqi","family":"Shao","sequence":"additional","affiliation":[{"name":"Shanghai AI Laboratory,China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hong","family":"Zhou","sequence":"additional","affiliation":[{"name":"Zhejiang University,China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Kaipeng","family":"Zhang","sequence":"additional","affiliation":[{"name":"Shanghai AI Laboratory,China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Bohan","family":"Zhuang","sequence":"additional","affiliation":[{"name":"Zhejiang University,China"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Gpt-4 technical report","author":"Achiam","year":"2023","journal-title":"arXiv preprint"},{"key":"ref2","article-title":"Hired: Attention-guided token dropping for efficient inference of high-resolution vision-language models in resourceconstrained environments","author":"Arif","year":"2024","journal-title":"arXiv preprint"},{"key":"ref3","article-title":"Longformer: The long-document transformer","author":"Beltagy","year":"2020","journal-title":"arXiv preprint"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73004-7_2"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1145\/3572848.3577500"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/DAC56929.2023.10247993"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1189"},{"key":"ref8","article-title":"Mme: A comprehensive evaluation benchmark for multimodal large language models","author":"Fu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref9","article-title":"Video-mme: The first-ever comprehensive evaluation benchmark of multi-modal 11 ms in video analysis","author":"Fu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref10","article-title":"Model tells you what to discard: Adaptive kv cache compression for 11 ms","volume-title":"ICLR","author":"Ge","year":"2024"},{"key":"ref11","article-title":"Making llama see and draw with seed tokenizer","volume-title":"ICLR","author":"Ge","year":"2024"},{"key":"ref12","article-title":"Seed-x: Multimodal models with unified multi-granularity comprehension and generation","author":"Ge","year":"2024","journal-title":"arXiv preprint"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.670"},{"key":"ref14","volume-title":"Flexattention: The flexibility of pytorch with the performance of flashattention","author":"He","year":"2024"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2181"},{"key":"ref16","article-title":"Kvquant: Towards 10 million context length llm inference with kv cache quantization","author":"Hooper","year":"2024","journal-title":"arXiv preprint"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00686"},{"key":"ref18","article-title":"Minference: Accelerating pre-filling for long-context 11 ms via dynamic sparse attention","volume-title":"Workshop on Efficient Systems for Foundation Models II@ ICML","author":"Jiang","year":"2024"},{"key":"ref19","article-title":"Gear: An efficient kv cache compression recipefor near-lossless generative inference of 11 m","author":"Kang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref20","volume-title":"xformers: A modular and hackable transformer modelling library","author":"Lefaudeux","year":"2022"},{"key":"ref21","volume-title":"Lmms-eval: Accelerating the development of large multimoal models","author":"Li","year":"2024"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.342"},{"key":"ref23","volume-title":"Llava-next: Improved reasoning, ocr, and world knowledge","author":"Liu","year":"2024"},{"key":"ref24","article-title":"Visual instruction tuning","author":"Liu","year":"2024","journal-title":"NeurIPS"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72643-9_4"},{"key":"ref26","article-title":"Kivi: A tuning-free asymmetric 2bit quantization for kv cache","author":"Liu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.679"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.findings-acl.177"},{"key":"ref29","volume-title":"Introducing meta llama 3: The most capable openly available llm to date","year":"2024"},{"key":"ref30","article-title":"Fast attention over long sequences with dynamic sparse flash attention","author":"Pagliardini","year":"2023","journal-title":"NeurIPS"},{"key":"ref31","article-title":"On the efficacy of eviction policy for key-value constrained generative language model inference","author":"Ren","year":"2024","journal-title":"arXiv preprint"},{"key":"ref32","article-title":"Sparq attention: Bandwidth-efficient 11 m inference","volume-title":"ICML","author":"Ribar","year":"2024"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00851"},{"key":"ref34","article-title":"Chameleon: Mixed-modal early-fusion foundation models","year":"2024","journal-title":"arXiv preprint"},{"key":"ref35","article-title":"Gemini: a family of highly capable multimodal models","author":"Team","year":"2023","journal-title":"arXiv preprint"},{"key":"ref36","volume-title":"Llama 3.1: An in-depth analysis of the next-generation large language model","author":"Vavekanand","year":"2024"},{"key":"ref37","article-title":"Look-m: Lookonce optimization in kv cache for efficient multimodal longcontext inference","author":"Wan","year":"2024","journal-title":"arXiv preprint"},{"key":"ref38","article-title":"Model tells you where to merge: Adaptive kv cache merging for 11 ms on long-context tasks","author":"Wang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref39","article-title":"Efficient streaming language models with attention sinks","volume-title":"ICLR","author":"Xiao","year":"2024"},{"key":"ref40","article-title":"Pyramidinfer: Pyramid kv cache compression for high-throughput 11 m inference","author":"Yang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref41","article-title":"No token left behind: Reliable kv cache compression via importance-aware mixed precision quantization","author":"Yang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref42","article-title":"Big bird: Transformers for longer sequences","author":"Zaheer","year":"2020","journal-title":"NeurIPS"},{"key":"ref43","article-title":"Long context transfer from language to vision","author":"Zhang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref44","article-title":"H2o: Heavy-hitter oracle for efficient generative inference of large language models","author":"Zhang","year":"2023","journal-title":"NeurIPS"},{"key":"ref45","article-title":"Near-lossless acceleration of long context 11 m inference with adaptive structured sparse attention","author":"Zhu","year":"2024","journal-title":"arXiv preprint"}],"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\/11446161.pdf?arnumber=11446161","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,30]],"date-time":"2026-04-30T06:40:23Z","timestamp":1777531223000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11446161\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":45,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.01904","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}