{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,27]],"date-time":"2026-06-27T19:46:29Z","timestamp":1782589589292,"version":"3.54.5"},"reference-count":82,"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\/iccv51701.2025.00009","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"12-23","source":"Crossref","is-referenced-by-count":5,"title":["Scaling Laws for Native Multimodal Models"],"prefix":"10.1109","author":[{"given":"Mustafa","family":"Shukor","sequence":"first","affiliation":[{"name":"Sorbonne University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Enrico","family":"Fini","sequence":"additional","affiliation":[{"name":"Apple"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Victor","family":"Guilherme Turrisi da Costa","sequence":"additional","affiliation":[{"name":"Apple"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Matthieu","family":"Cord","sequence":"additional","affiliation":[{"name":"Sorbonne University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Joshua","family":"Susskind","sequence":"additional","affiliation":[{"name":"Apple"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Alaaeldin","family":"El-Nouby","sequence":"additional","affiliation":[{"name":"Apple"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","author":"Abdin","year":"2024","journal-title":"Phi-3 technical report: A highly capable language model locally on your phone"},{"key":"ref2","author":"Abnar","year":"2025","journal-title":"Parameters vs flops: Scaling laws for optimal sparsity for mixture-of-experts language models"},{"key":"ref3","author":"Achiam","year":"2023","journal-title":"Gpt-4 technical report"},{"key":"ref4","author":"Aghajanyan","year":"2022","journal-title":"Cm3: A causal masked multimodal model of the internet"},{"key":"ref5","first-page":"265","article-title":"Scaling laws for generative mixed-modal language models","volume-title":"International Conference on Machine Learning","author":"Aghajanyan","year":"2023"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1723"},{"key":"ref7","author":"Bao","year":"2021","journal-title":"Vlmo: Unified vision-language pretraining with mixture-of-modality-experts"},{"key":"ref8","author":"Bavishi","year":"2023","journal-title":"Introducing our multimodal models"},{"key":"ref9","author":"Beyer","year":"2024","journal-title":"Paligemma: A versatile 3b vlm for transfer"},{"key":"ref10","first-page":"1877","article-title":"Language models are few-shot learners","volume":"33","author":"Brown","year":"2020","journal-title":"Advances in neural information processing systems"},{"key":"ref11","volume-title":"Coyo700m: Image-text pair dataset","author":"Byeon","year":"2022"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i2.27872"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52733.2024.02283"},{"key":"ref14","article-title":"Training compute-optimal protein language models","author":"Cheng","year":"2024","journal-title":"bioRxiv"},{"key":"ref15","first-page":"4057","article-title":"Unified scaling laws for routed language models","volume-title":"International conference on machine learning","author":"Clark","year":"2022"},{"key":"ref16","author":"Dai","year":"2024","journal-title":"Nvlm: Open frontier-class multimodal 11 ms"},{"key":"ref17","first-page":"7480","article-title":"Scaling vision transformers to 22 billion parameters","volume-title":"International Conference on Machine Learning","author":"Dehghani","year":"2023"},{"key":"ref18","author":"Diao","year":"2024","journal-title":"Unveiling encoder-free vision-language models"},{"key":"ref19","author":"Dubey","year":"2024","journal-title":"The llama 3 herd of models"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.01268"},{"key":"ref21","author":"Fang","year":"2023","journal-title":"Data filtering networks"},{"issue":"120","key":"ref22","first-page":"1","article-title":"Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity","volume":"23","author":"Fedus","year":"2022","journal-title":"Journal of Machine Learning Research"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.00901"},{"key":"ref24","first-page":"288","article-title":"Megablocks: Efficient sparse training with mixture-of-experts","volume-title":"Proceedings of Machine Learning and Systems","volume":"5","author":"Gale","year":"2023"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2427"},{"key":"ref26","first-page":"30016","article-title":"Training compute-optimal large language models","volume-title":"Proceedings of the 36th International Conference on Neural Information Processing Systems","author":"Hoffmann","year":"2022"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.52202\/075280-3155"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4612-4380-9_35"},{"key":"ref29","author":"Hurst","year":"2024","journal-title":"Gpt-4o system card"},{"key":"ref30","author":"Jiang","year":"2024","journal-title":"Mixtral of experts"},{"key":"ref31","author":"Kaplan","year":"2020","journal-title":"Scaling laws for neural language models"},{"key":"ref32","first-page":"17283","article-title":"Grounding language models to images for multimodal inputs and outputs","volume-title":"International Conference on Machine Learning","author":"Yu Koh","year":"2023"},{"key":"ref33","author":"Krajewski","year":"2024","journal-title":"Scaling laws for fine-grained mixture of experts"},{"key":"ref34","article-title":"Obelics: An open web-scale filtered dataset of interleaved image-text documents","volume":"36","author":"Lauren\u00e7on","year":"2024","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2789"},{"key":"ref36","author":"Lepikhin","year":"2020","journal-title":"Gshard: Scaling giant models with conditional computation and automatic sharding"},{"key":"ref37","first-page":"6265","article-title":"Base layers: Simplifying training of large, sparse models","volume-title":"International Conference on Machine Learning","author":"Lewis","year":"2021"},{"key":"ref38","author":"Li","year":"2024","journal-title":"Aria: An open multimodal native mixture-ofexperts model"},{"key":"ref39","author":"Li","year":"2024","journal-title":"Datacomp-1m: In search of the next generation of training sets for language models"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1109\/tmm.2026.3654458"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02520"},{"key":"ref42","author":"Liu","year":"2024","journal-title":"Deepseek-v3 technical report"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02484"},{"key":"ref44","author":"Loshchilov","year":"2017","journal-title":"Decoupled weight decay regularization"},{"key":"ref45","article-title":"Unified-io: A unified model for vision, language, and multi-modal tasks","volume-title":"The Eleventh International Conference on Learning Representations","author":"Lu","year":"2022"},{"key":"ref46","first-page":"304","article-title":"Mm1: methods, analysis and insights from multimodal 11 m pre-training","volume-title":"European Conference on Computer Vision","author":"McKinzie","year":"2025"},{"key":"ref47","article-title":"Linearly mapping from image to text space","volume-title":"The Eleventh International Conference on Learning Representations","author":"Merullo","year":"2023"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2544"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-industry.98"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.52202\/068431-0695"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.2307\/2006193"},{"key":"ref52","author":"Oquab","year":"2023","journal-title":"Dinov2: Learning robust visual features without supervision"},{"key":"ref53","author":"Pearce","year":"2024","journal-title":"Reconciling kaplan and chinchilla scaling laws"},{"key":"ref54","author":"Pearce","year":"2024","journal-title":"Scaling laws for pre-training agents and world models"},{"key":"ref55","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":"ref56","author":"Rajasegaran","year":"2025","journal-title":"An empirical study of autoregressive pretraining from videos"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00513"},{"key":"ref58","author":"Shazeer","year":"2020","journal-title":"Glu variants improve transformer"},{"key":"ref59","author":"Shazeer","year":"2017","journal-title":"Outrageously large neural networks: The sparsely-gated mixture-of-experts layer"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.findings-emnlp.758"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1109\/iccv51070.2023.02016"},{"key":"ref62","article-title":"Unival: Unified model for image, video, audio and language tasks","author":"Shukor","year":"2023","journal-title":"Transactions on Machine Learning Research Journal"},{"key":"ref63","author":"Sun","year":"2024","journal-title":"Hunyuan-large: An open-source moe model with 52 billion activated parameters by tencent"},{"key":"ref64","year":"2024","journal-title":"Chameleon: Mixed-modal early-fusion foundation models"},{"key":"ref65","author":"Team","year":"2023","journal-title":"Gemini: a family of highly capable multimodal models"},{"key":"ref66","first-page":"200","article-title":"Multimodal few-shot learning with frozen language models","volume":"34","author":"Tsimpoukelli","year":"2021","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-92089-9_23"},{"key":"ref68","article-title":"Neural discrete representation learning","author":"van den Oord","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref69","article-title":"Attention is all you need","author":"Vaswani","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref70","first-page":"23318","article-title":"Ofa: Unifying architectures, tasks, and modalities through a simple sequence-to-sequence learning framework","volume-title":"International conference on machine learning","author":"Wang","year":"2022"},{"key":"ref71","author":"Wang","year":"2024","journal-title":"Qwen2-vl: Enhancing vision-language model\u2019s perception of the world at any resolution"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.319"},{"key":"ref73","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52729.2023.01838"},{"key":"ref74","author":"Wang","year":"2024","journal-title":"Emu3: Next-token prediction is all you need"},{"key":"ref75","author":"Wei","year":"2024","journal-title":"Skywork-moe: A deep dive into training techniques for mixture-of-experts language models"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1109\/iccvw69036.2025.00644"},{"key":"ref77","article-title":"Coca: Contrastive captioners are image-text foundation models","author":"Yu","year":"2022","journal-title":"Transactions on Machine Learning Research"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01100"},{"key":"ref79","author":"Zhang","year":"2024","journal-title":"Mm1. 5: Methods, analysis & insights from multimodal 11 m fine-tuning"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.14778\/3611540.3611569"},{"key":"ref81","article-title":"MiniGPT-4: Enhancing vision-language understanding with advanced large language models","volume-title":"The Twelfth International Conference on Learning Representations","author":"Zhu","year":"2024"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1109\/ipdpsw55747.2022.00171"}],"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\/11446043.pdf?arnumber=11446043","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T04:58:29Z","timestamp":1777611509000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11446043\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":82,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.00009","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}