{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T09:59:15Z","timestamp":1777888755469,"version":"3.51.4"},"reference-count":135,"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\/501100001809","name":"NSFC","doi-asserted-by":"publisher","award":["62201342"],"award-info":[{"award-number":["62201342"]}],"id":[{"id":"10.13039\/501100001809","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\/iccv51701.2025.01986","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"21384-21397","source":"Crossref","is-referenced-by-count":0,"title":["Skip-Vision: Efficient and Scalable Acceleration of Vision-Language Models Via Adaptive Token Skipping"],"prefix":"10.1109","author":[{"given":"Weili","family":"Zeng","sequence":"first","affiliation":[{"name":"AI Institute, Shanghai Jiao Tong University,MoE Key Lab of Artificial Intelligence"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ziyuan","family":"Huang","sequence":"additional","affiliation":[{"name":"Ant Group"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Kaixiang","family":"Ji","sequence":"additional","affiliation":[{"name":"Ant Group"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yichao","family":"Yan","sequence":"additional","affiliation":[{"name":"AI Institute, Shanghai Jiao Tong University,MoE Key Lab of Artificial Intelligence"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v33i01.33018076"},{"key":"ref2","article-title":"Gpt-4 technical report","author":"Achiam","year":"2023","journal-title":"arXiv preprint"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1723"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2015.279"},{"key":"ref5","article-title":"Screenai: A vision-language model for ui and infographics understanding","author":"Baechler","year":"2024","journal-title":"arXiv preprint"},{"issue":"2","key":"ref6","first-page":"3","article-title":"Qwen-vl: A versatile vision-language model for understanding, localization, text reading, and beyond","volume":"1","author":"Bai","year":"2023","journal-title":"arXiv preprint"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0237"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00439"},{"key":"ref9","article-title":"Token merging: Your vit but faster","author":"Bolya","year":"2022","journal-title":"arXiv preprint"},{"key":"ref10","article-title":"Language models are few-shot learners","author":"Brown","year":"2020","journal-title":"arXiv preprint"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.44"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01487"},{"key":"ref13","volume-title":"Textocr-gpt4v","author":"Carter","year":"2024"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01311"},{"key":"ref15","article-title":"Manning. Auroracap: Efficient, performant video detailed captioning and a new benchmark","author":"Chai","year":"2024","journal-title":"arXiv preprint"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2020.2984627"},{"key":"ref17","article-title":"Mapqa: A dataset for question answering on choropleth maps","author":"Chang","year":"2022","journal-title":"arXiv preprint"},{"key":"ref18","author":"Hardy Chen","year":"2024","journal-title":"Allava: Harnessing gpt4v-synthesized data for a lite vision-language model"},{"key":"ref19","article-title":"Shikra: Unleashing multimodal llm\u2019s referential dialogue magic","author":"Chen","year":"2023","journal-title":"arXiv preprint"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72643-9_22"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0850"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73004-7_2"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.emnlp-main.343"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1007\/s11432-024-4231-5"},{"key":"ref25","volume-title":"Vicuna: An open-source chatbot impressing gpt-4 with 90 * chatgpt quality","author":"Chiang","year":"2023"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/ICDAR.2019.00252"},{"key":"ref27","article-title":"Mobilevlm: A fast, reproducible and strong vision language assistant for mobile devices","author":"Chu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2142"},{"key":"ref29","article-title":"The llama 3 herd of models","author":"Dubey","year":"2024","journal-title":"arXiv preprint"},{"key":"ref30","article-title":"Mme: A comprehensive evaluation benchmark for multimodal large language models","author":"Fu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref31","article-title":"G-llava: Solving geometric problem with multi-modal large language model","author":"Gao","year":"2023","journal-title":"arXiv preprint"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.751"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.254"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00103"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.naacl-long.222"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/ICPR.2018.8546143"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-emnlp.175"},{"key":"ref38","article-title":"mplug-docow12: High-resolution compressing for ocrfree multi-page document understanding","author":"Hu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref39","article-title":"Matryoshka query transformer for large vision-language models","author":"Hu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref40","article-title":"Accelerating pre-training of multimodal 11 ms via chain-of-sight","author":"Huang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref41","first-page":"4651","article-title":"Perceiver: General perception with iterative attention","volume-title":"International conference on machine learning","author":"Jaegle","year":"2021"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1007\/s44267-025-00099-6"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.215"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00592"},{"key":"ref45","article-title":"Figureqa: An annotated figure dataset for visual reasoning","author":"Ebrahimi Kahou","year":"2017","journal-title":"arXiv preprint"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.acl-long.277"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72640-8_7"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1109\/ICDAR.2015.7333942"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46493-0_15"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.571"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.findings-acl.526"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1038\/sdata.2018.251"},{"key":"ref53","author":"Lauren\u00e7on","year":"2024","journal-title":"Building and better understanding visionlanguage models: insights and future directions."},{"key":"ref54","article-title":"Volcano: mitigating multimodal hallucination through self-feedback guided revision","author":"Lee","year":"2023","journal-title":"arXiv preprint"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1145\/3477495.3531753"},{"key":"ref56","author":"Li","year":"2024","journal-title":"Llava-next: Stronger 11 ms supercharge multimodal capabilities in the wild"},{"key":"ref57","article-title":"Llava-onevision: Easy visual task transfer","author":"Li","year":"2024","journal-title":"arXiv preprint"},{"key":"ref58","first-page":"19730","article-title":"Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models","volume-title":"International conference on machine learning","author":"Li","year":"2023"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.775"},{"key":"ref60","article-title":"Tokenpacker: Efficient visual projector for multimodal 11 m","author":"Li","year":"2024","journal-title":"arXiv preprint"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1109\/tpami.2025.3637265"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52729.2023.01437"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02520"},{"key":"ref64","article-title":"Sphinx-x: Scaling data and parameters for a family of multi-modal large language models","author":"Liu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00566"},{"key":"ref66","article-title":"Mmc: Advancing multimodal chart understanding with large-scale instruction tuning","author":"Liu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02484"},{"key":"ref68","volume-title":"Llava-next: Improved reasoning, ocr, and world knowledge","author":"Liu","year":"2024"},{"key":"ref69","article-title":"Visual instruction tuning","volume":"36","author":"Liu","year":"2024","journal-title":"Advances in neural information processing systems"},{"key":"ref70","article-title":"On the hidden mystery of ocr in large multimodal models","author":"Liu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref71","article-title":"On the hidden mystery of ocr in large multimodal models","author":"Liu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72658-3_13"},{"key":"ref73","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-41679-8_28"},{"key":"ref74","article-title":"Iconqa: A new benchmark for abstract diagram understanding and visual language reasoning","author":"Lu","year":"2021","journal-title":"arXiv preprint"},{"key":"ref75","article-title":"Learn to explain: Multimodal reasoning via thought chains for science question answering","volume-title":"In The 36th Conference on Neural Information Processing Systems (NeurIPS)","author":"Lu","year":"2022"},{"key":"ref76","article-title":"Mathvista: Evaluating mathematical reasoning of foundation models in visual contexts","author":"Lu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref77","author":"Lu","year":"2024","journal-title":"Ovis: Structural embedding alignment for multimodal large language model"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00331"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.findings-acl.177"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.1007\/s10032-021-00383-3"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.1109\/WACV48630.2021.00225"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1109\/WACV51458.2022.00264"},{"key":"ref83","article-title":"Mass-editing memory in a transformer","author":"Meng","year":"2022","journal-title":"arXiv preprint"},{"key":"ref84","doi-asserted-by":"publisher","DOI":"10.1109\/WACV45572.2020.9093523"},{"key":"ref85","doi-asserted-by":"publisher","DOI":"10.1109\/ICDAR.2019.00156"},{"key":"ref86","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":"ref87","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P18-2124"},{"key":"ref88","article-title":"Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context","author":"Reid","year":"2024","journal-title":"arXiv preprint"},{"key":"ref89","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-015-0816-y"},{"key":"ref90","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v33i01.33018876"},{"key":"ref91","article-title":"Llava-prumerge: Adaptive token reduction for efficient large multimodal models","author":"Shang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref92","article-title":"Visual cot: Unleashing chain-of-thought reasoning in multi-modal language models","author":"Shao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref93","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-emnlp.268"},{"key":"ref94","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58536-5_44"},{"key":"ref95","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00851"},{"key":"ref96","article-title":"Docscan: Unsupervised text classification via learning from neighbors","author":"Stammbach","year":"2021","journal-title":"arXiv preprint"},{"key":"ref97","author":"AI","year":"2024","journal-title":"Sujet-finance-qa-vision-100k: A large-scale dataset for financial document vqa"},{"key":"ref98","author":"Svetlichnaya","year":"2020","journal-title":"Deepform: Understand structured documents at scale"},{"key":"ref99","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v35i15.17635"},{"key":"ref100","article-title":"Cambrian1: A fully open, vision-centric exploration of multimodal 1 lms","author":"Tong","year":"2024","journal-title":"arXiv preprint"},{"key":"ref101","article-title":"Llama: Open and efficient foundation language models","author":"Touvron","year":"2023","journal-title":"arXiv preprint"},{"key":"ref102","article-title":"Coco-text: Dataset and benchmark for text detection and recognition in natural images","author":"Veit","year":"2016","journal-title":"arXiv preprint"},{"key":"ref103","article-title":"Lipschitz regularity of deep neural networks: analysis and efficient estimation","volume":"31","author":"Virmaux","year":"2018","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref104","article-title":"Qwen2-vl: Enhancing vision-language model\u2019s perception of the world at any resolution","author":"Wang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref105","author":"Wang","year":"2023","journal-title":"Cogvlm: Visual expert for pretrained language models"},{"key":"ref106","author":"Wang","year":"2024","journal-title":"Internvid: A large-scale video-text dataset for multimodal understanding and generation"},{"key":"ref107","doi-asserted-by":"publisher","DOI":"10.5040\/9781350601529.16472"},{"key":"ref108","article-title":"Efficient vision-language models by summarizing visual tokens into compact registers","author":"Wen","year":"2024","journal-title":"arXiv preprint"},{"key":"ref109","volume-title":"Renderedtext","author":"Wendler","year":"2023"},{"key":"ref110","article-title":"Tabfact: A large-scale dataset for table-based fact verification","volume-title":"International Conference on Learning Representations (ICLR)","author":"Chen","year":"2020"},{"key":"ref111","article-title":"Pyramiddrop: Accelerating your large vision-language models via pyramid visual redundancy reduction","author":"Xing","year":"2024","journal-title":"arXiv preprint"},{"key":"ref112","article-title":"Llava-uhd: an 1 mm perceiving any aspect ratio and high-resolution images","author":"Xu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref113","article-title":"Chartbench: A benchmark for complex visual reasoning in charts","author":"Xu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref114","article-title":"Qwen2 technical report","author":"Yang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref115","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.01843"},{"key":"ref116","article-title":"Deco: Decoupling token compression from semantic abstraction in multimodal large language models","author":"Yao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref117","article-title":"Minicpm-v: A gpt-4v level mllm on your phone","author":"Yao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref118","article-title":"mplug-ow13: Towards long image-sequence understanding in multi-modal large language models","author":"Ye","year":"2024","journal-title":"arXiv preprint"},{"key":"ref119","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.02777"},{"key":"ref120","article-title":"Rlaif-v: Aligning mllms through open-source ai feedback for super gpt-4v trustworthiness","author":"Yu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref121","article-title":"Mm -vet: Evaluating large multimodal models for integrated capabilities","author":"Yu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref122","article-title":"Crepe: Open-domain question answering with false presuppositions","author":"Velocity Yu","year":"2022","journal-title":"arXiv preprint"},{"key":"ref123","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51701.2025.00038"},{"key":"ref124","article-title":"Tinygpt-v: Efficient multimodal large language model via small backbones","author":"Yuan","year":"2023","journal-title":"arXiv preprint"},{"key":"ref125","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00913"},{"key":"ref126","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00546"},{"key":"ref127","article-title":"[cls] attention is all you need for trainingfree visual token pruning: Make vlm inference faster","author":"Zhang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref128","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73242-3_10"},{"key":"ref129","first-page":"5","article-title":"Uber-text: A large-scale dataset for optical character recognition from street-level imagery","volume-title":"SUNw: Scene Understanding Workshop CVPR","author":"Zhang","year":"2017"},{"key":"ref130","article-title":"Sparsevlm: Visual token sparsification for efficient vision-language model inference","author":"Zhang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref131","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1506"},{"key":"ref132","article-title":"Svit: Scaling up visual instruction tuning","author":"Zhao","year":"2023","journal-title":"arXiv preprint"},{"key":"ref133","article-title":"Tinyllava: A framework of small-scale large multimodal models","author":"Zhou","year":"2024","journal-title":"arXiv preprint"},{"key":"ref134","article-title":"Minigpt-4: Enhancing vision-language understanding with advanced large language models","author":"Zhu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref135","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.acl-long.254"}],"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\/11444969.pdf?arnumber=11444969","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T04:54:56Z","timestamp":1777611296000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11444969\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":135,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.01986","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}