{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:11:29Z","timestamp":1777889489851,"version":"3.51.4"},"reference-count":91,"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\/iccv51701.2025.02214","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"23880-23891","source":"Crossref","is-referenced-by-count":0,"title":["Make Your Training Flexible: Towards Deployment-Efficient Video Models"],"prefix":"10.1109","author":[{"given":"Chenting","family":"Wang","sequence":"first","affiliation":[{"name":"Shanghai Jiao Tong University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Kunchang","family":"Li","sequence":"additional","affiliation":[{"name":"Shanghai AI Laboratory"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Tianxiang","family":"Jiang","sequence":"additional","affiliation":[{"name":"Shanghai AI Laboratory"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xiangyu","family":"Zeng","sequence":"additional","affiliation":[{"name":"Shanghai AI Laboratory"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yi","family":"Wang","sequence":"additional","affiliation":[{"name":"Shanghai AI Laboratory"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Limin","family":"Wang","sequence":"additional","affiliation":[{"name":"Shanghai Jiao Tong University"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2017.618"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00676"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00175"},{"key":"ref4","article-title":"Beit: Bert pre-training of image transformers","author":"Bao","year":"2021","journal-title":"ICLR"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2102.05095"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01393"},{"key":"ref7","article-title":"Once-for-all: Train one network and specialize it for efficient deployment","author":"Cai","year":"2019","journal-title":"ICLR"},{"key":"ref8","volume":"abs\/2405.17430","author":"Cai","year":"2024","journal-title":"Matryoshka multimodal models"},{"key":"ref9","volume":"abs\/1808.01340","author":"Carreira","year":"2018","journal-title":"A short note about kinetics600"},{"key":"ref10","volume":"abs\/1907.06987","author":"Carreira","year":"2019","journal-title":"A short note on the kinetics-700 human action dataset"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00356"},{"key":"ref12","article-title":"Collecting highly parallel data for paraphrase evaluation","author":"Chen","year":"2011","journal-title":"ACL"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.52202\/075280-3185"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52729.2023.01034"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01776"},{"key":"ref16","volume":"abs\/1810.04805","author":"Devlin","year":"2018","journal-title":"Bert: Pre-training of deep bidirectional transformers for language understanding"},{"key":"ref17","article-title":"Unified language model pre-training for natural language understanding and generation","volume":"32","author":"Dong","year":"2019","journal-title":"NeurIPS"},{"key":"ref18","article-title":"An image is worth 16x16 words: Transformers for image recognition at scale","author":"Dosovitskiy","year":"2021","journal-title":"ICLR"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00675"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00630"},{"key":"ref21","article-title":"Masked autoencoders as spatiotemporal learners","author":"Feichtenhofer","year":"2022","journal-title":"NeurIPS"},{"key":"ref22","volume":"abs\/2111.12681","author":"Fu","year":"2021","journal-title":"Violet: End-to-end video-language transformers with masked visual-token modeling"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2017.622"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.acl-long.524"},{"key":"ref25","article-title":"Turbo training with token dropout","author":"Han","year":"2022","journal-title":"In BMVC"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01282"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01553"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7298698"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/icassp49660.2025.10888259"},{"key":"ref30","volume":"abs\/1704.04861","author":"Howard","year":"2017","journal-title":"Mobilenets: Efficient convolutional\\\\ neural networks for mobile vision applications"},{"key":"ref31","article-title":"Everest: Efficient masked video autoencoder by removing redundant spatiotemporal tokens","author":"Hwang","year":"2022","journal-title":"ICML"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-19833-5_6"},{"key":"ref33","article-title":"Bridging episodes and semantics: A novel framework for long-form video understanding","volume-title":"In ECCV workshops","author":"Gueter Josmy","year":"2024"},{"key":"ref34","volume":"abs\/1705.06950","author":"Kay","year":"2017","journal-title":"The kinetics human action video dataset"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-016-0981-7"},{"key":"ref36","article-title":"Matryoshka representation learning","author":"Kusupati","year":"2022","journal-title":"NIPS"},{"key":"ref37","volume":"abs\/2206.03428","author":"Lei","year":"2022","journal-title":"Revealing single frame bias for video-and-language learning"},{"key":"ref38","article-title":"Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models","author":"Li","year":"2022","journal-title":"ICML"},{"key":"ref39","volume":"abs\/2211.09552","author":"Li","year":"2022","journal-title":"Uniformerv 2: Spatiotemporal learning by arming image vits with video uniformer"},{"key":"ref40","article-title":"Uniformer: Unified transformer for efficient spatial-temporal representation learning","author":"Li","year":"2022","journal-title":"In ICLR"},{"key":"ref41","volume":"abs\/2305.06355","author":"Li","year":"2023","journal-title":"Videochat: Chat-centric video understanding"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01826"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73347-5_14"},{"key":"ref44","volume":"abs\/2112.01526","author":"Li","year":"2021","journal-title":"Im-\\\\ proved multiscale vision transformers for classification and detection"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00718"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-10602-1_48"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01348"},{"key":"ref48","volume":"abs\/2310.03744","author":"Liu","year":"2023","journal-title":"Improved baselines with visual instruction tuning"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1516"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.00320"},{"key":"ref51","volume":"abs\/2409.12961","author":"Liu","year":"2024","journal-title":"Oryx mllm: On-demand spatialtemporal understanding at arbitrary resolution"},{"key":"ref52","article-title":"Vilbert: Pretraining task-agnostic visiolinguistic representations for vision-and-language tasks","author":"Lu","year":"2019","journal-title":"NeurIPS"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2022.07.028"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.679"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.01463"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2019.2901464"},{"key":"ref57","article-title":"Im2text: Describing images using 1 million captioned photographs","author":"Ordonez","year":"2011","journal-title":"In NeurIPS"},{"key":"ref58","volume":"abs\/2208.06366","author":"Peng","year":"2022","journal-title":"Beit v2: Masked image modeling with vector-quantized visual tokenizers"},{"key":"ref59","article-title":"Mar: Masked autoencoders for efficient action recognition","author":"Qing","year":"2022","journal-title":"In ACMMM"},{"key":"ref60","article-title":"Learning transferable visual\\\\ models from natural language supervision","author":"Radford","year":"2021","journal-title":"ICML"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-016-0987-1"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P18-1238"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2023.127063"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00130"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.02176"},{"key":"ref66","article-title":"VideoMAE: Masked autoencoders are data-efficient learners for self-supervised video pre-training","author":"Tong","year":"2022","journal-title":"NeurIPS"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01511"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00243"},{"key":"ref69","volume":"abs\/2209.07526","author":"Wang","year":"2022","journal-title":"Omnivl: One foundation model for image-language and video-language tasks"},{"key":"ref70","volume":"abs\/2407.00634","author":"Wang","year":"2024","journal-title":"Tarsier: Recipes for training and evaluating large video description models"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46484-8_2"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01398"},{"key":"ref73","volume":"abs\/2409.12191","author":"Wang","year":"2024","journal-title":"Qwen2-vl: Enhancing vision-language model\u2019s perception of the world at any resolution"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01432"},{"key":"ref75","volume":"abs\/2212.03191","author":"Wang","year":"2022","journal-title":"Internvideo: General video foundation models via generative and discriminative learning"},{"key":"ref76","author":"Wang","year":"2023","journal-title":"Internvid: A large-scale video-text dataset for multimodal understanding and generation"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73013-9_23"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01103"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-20056-4_20"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00023"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.571"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.00333"},{"key":"ref83","volume":"abs\/2212.04979","author":"Yan","year":"2022","journal-title":"Video-text modeling with zero-shot transfer from contrastive captioners"},{"key":"ref84","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58571-6_41"},{"key":"ref85","article-title":"Coca: Contrastive captioners are image-text foundation models","author":"Yu","year":"2022","journal-title":"Transactions on Machine Learning Research"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01100"},{"key":"ref87","volume":"abs\/2112.07175","author":"Zhang","year":"2021","journal-title":"Co-training transformer with videos and images improves action recognition"},{"key":"ref88","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01012"},{"key":"ref89","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00876"},{"key":"ref90","volume":"abs\/2111.07832","author":"Zhou","year":"2021","journal-title":"ibot: Image bert pre-training with online tokenizer"},{"key":"ref91","article-title":"Languagebind: Extending video-language pretraining to n modality by language-based semantic alignment","author":"Zhu","year":"2024","journal-title":"In 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\/11443394.pdf?arnumber=11443394","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:11:42Z","timestamp":1777612302000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11443394\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":91,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.02214","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}