--- 格式版本: 2 标题: "MoE-ViE: Mixture of Experts Vision Encoder for Efficient Image and Video Understanding" 原文链接: "https://arxiv.org/abs/2608.17402" 发布日期: "2026-08-19" 发布时间校准状态: "found" 发布时间需复核: "否" 发布时间来源: "llm:local:strict_original_body" 发布时间证据: "Published Time: Wed, 19 Aug 2026 00:32:00 GMT" 发布时间校准原因: "正文中明确标注发布时间为2026-08-19，优先于提交日期。" 发布时间校准置信度: "1" 发布时间候选数量: 12 发布时间严格候选数量: 3 发布时间原页读取状态: "source template page reused from URL open" 发布时间未找到原因: "" 发布时间校准时间: "2026-08-19T21:19:09+08:00" 发布时间仲裁状态: "confirmed" 发布时间仲裁尝试次数: 3 发布时间仲裁耗时毫秒: 25641 发现时间: "2026-08-19T21:16:38+08:00" 入库时间: "2026-08-19T13:19:35.515Z" 来源平台: "arXiv 学术论文搜索" 搜索渠道: "source_template" 搜索词: "https://arxiv.org/search/?query=Scale-up&searchtype=all" 匹配关键词: - "Scale-up" - "performance" - "latency" - "AI" 相关厂家: - "Google" 相关专家: [] 内容类型: "网页" 抓取工具: "Jina Reader" 清洗工具: "Jina Reader Markdown + Defuddle/Readability 正文提取" 原始附件: [] AI优质: "否" AI打分: 20 AI分档: "非优质" AI质检状态: "不通过" AI打分理由: "该资料为计算机视觉论文，讨论MoE视觉编码器，与超节点/AI Rack/机柜级AI基础设施完全无关，仅来源权威，无相关技术或商业信息。" AI质检模型: "ali-deepseek-v4-flash" AI质检时间: "2026-08-19T21:19:43+08:00" AI主题相关性: 0 AI来源权威性: 10 AI新颖性: 5 AI技术细节: 0 AI商业部署信号: 0 AI完整性: 5 AI摘要: "该论文提出MoE-ViE，一种用于图像与视频理解的混合专家视觉编码器，通过细粒度MoE拓扑、无辅助损失平衡和专用内核显著提升效率。其最大模型在零样本性能上匹配1.7倍大小的SOTA编码器，且延迟仅为后者的76%；" AI摘要模型: "ali-deepseek-v4-flash" AI摘要时间: "2026-09-07T03:19:50.350Z" 采集批次: "2026年8月19日19点33分50秒" 采集批次ID: "20260819-193350-682" 去重键: "https://arxiv.org/abs/2608.17402" --- Title: MoE-ViE: Mixture of Experts Vision Encoder for Efficient Image and Video Understanding URL Source: https://arxiv.org/abs/2608.17402 Published Time: Wed, 19 Aug 2026 00:32:00 GMT Markdown Content: [Skip to main content](https://arxiv.org/abs/2608.17402#content)[](https://arxiv.org/IgnoreMe)[![Image 1: archive](https://arxiv.org/static/base/1.0.1/images/arxiv-logo-primary-light.svg)](https://arxiv.org/)[Search](https://arxiv.org/search)[Submit](https://arxiv.org/user/create)[Donate](https://info.arxiv.org/about/donate.html)[Log in](https://arxiv.org/login) Search arXiv Press Enter to search · [Advanced search](https://arxiv.org/search/advanced) # Computer Science > Computer Vision and Pattern Recognition **arXiv:2608.17402** (cs) [Submitted on 18 Aug 2026] # Title:MoE-ViE: Mixture of Experts Vision Encoder for Efficient Image and Video Understanding Authors:[Bonan Zhang](https://arxiv.org/search/cs?searchtype=author&query=Zhang,+B), [Shiyu Dong](https://arxiv.org/search/cs?searchtype=author&query=Dong,+S), [Quan Hung Tran](https://arxiv.org/search/cs?searchtype=author&query=Tran,+Q+H), [Katharina Gschwind](https://arxiv.org/search/cs?searchtype=author&query=Gschwind,+K), [Shuqi Yang](https://arxiv.org/search/cs?searchtype=author&query=Yang,+S), [Sijia Chen](https://arxiv.org/search/cs?searchtype=author&query=Chen,+S), [Adel Ahmadyan](https://arxiv.org/search/cs?searchtype=author&query=Ahmadyan,+A), [Seungwhan Moon](https://arxiv.org/search/cs?searchtype=author&query=Moon,+S), [Lu Zhang](https://arxiv.org/search/cs?searchtype=author&query=Zhang,+L), [Ahmed Kirmani](https://arxiv.org/search/cs?searchtype=author&query=Kirmani,+A), [Babak Damavandi](https://arxiv.org/search/cs?searchtype=author&query=Damavandi,+B), [Anuj Kumar](https://arxiv.org/search/cs?searchtype=author&query=Kumar,+A) View a PDF of the paper titled MoE-ViE: Mixture of Experts Vision Encoder for Efficient Image and Video Understanding, by Bonan Zhang and 11 other authors [View PDF](https://arxiv.org/pdf/2608.17402)[HTML (experimental)](https://arxiv.org/html/2608.17402v1) > Abstract:Vision encoders are a critical component of vision-language models, and scaling their capacity effectively improves performance. However, dense scaling increases compute cost and inference latency. Mixture-of-Experts (MoE) architectures offer a compelling alternative, having enabled efficient scaling in LLMs, yet the MoE design space for CLIP-style vision encoders remains underexplored at State-of-the-Art (SOTA) levels. In this work, we systematically study MoE designs for vision encoder scaling and find that fine-grained MoE topologies yield substantial gains over both dense and standard MoE counterparts. We further propose an auxiliary-loss-free balancing variant for better expert utilization, and design a specialized MoE kernel to mitigate inference latency overhead. To enhance video capabilities while preserving image knowledge, we introduce frame-level distillation paired with a novel freezing mechanism. We pretrain a series of Mixture-of-Experts Vision Encoders (MoE-ViE) across a range of sizes, all consistently outperforming their dense counterparts. Our largest model matches the zero-shot performance of a SOTA encoder 1.7x its size at 76% of its latency. When aligned with an LLM, MoE-ViE surpasses all compared encoders on image and video benchmarks, including those with up to 5x more activated parameters. Code is available at [this https URL](https://github.com/facebookresearch/moe_vie). Comments:Accepted to ECCV 2026 Subjects:Computer Vision and Pattern Recognition (cs.CV) Cite as:[arXiv:2608.17402](https://arxiv.org/abs/2608.17402) [cs.CV] (or [arXiv:2608.17402v1](https://arxiv.org/abs/2608.17402v1) [cs.CV] for this version) [https://doi.org/10.48550/arXiv.2608.17402](https://doi.org/10.48550/arXiv.2608.17402) Focus to learn more arXiv-issued DOI via DataCite (pending registration) ## Submission history From: Bonan Zhang [[view email](https://arxiv.org/show-email/4b9c5e25/2608.17402)] **[v1]** Tue, 18 Aug 2026 05:58:40 UTC (472 KB) [](https://arxiv.org/abs/2608.17402)Full-text links: ## Access Paper: View a PDF of the paper titled MoE-ViE: Mixture of Experts Vision Encoder for Efficient Image and Video Understanding, by Bonan Zhang and 11 other authors * [View PDF](https://arxiv.org/pdf/2608.17402) * [HTML (experimental)](https://arxiv.org/html/2608.17402v1) * [TeX Source](https://arxiv.org/src/2608.17402) [view license](http://arxiv.org/licenses/nonexclusive-distrib/1.0/ "Rights to this article") ### Current browse context: cs.CV [](https://arxiv.org/prevnext?id=2608.17402&function=next&context=cs.CV "next in cs.CV (accesskey n)") [new](https://arxiv.org/list/cs.CV/new) | [recent](https://arxiv.org/list/cs.CV/recent) | [2026-08](https://arxiv.org/list/cs.CV/2026-08) Change to browse by: [cs](https://arxiv.org/abs/2608.17402?context=cs) ### References & Citations * [NASA ADS](https://ui.adsabs.harvard.edu/abs/arXiv:2608.17402) * [Google Scholar](https://scholar.google.com/scholar_lookup?arxiv_id=2608.17402) * [Semantic Scholar](https://api.semanticscholar.org/arXiv:2608.17402) export BibTeX citation Loading... ## BibTeX formatted citation × Data provided by: [](https://arxiv.org/abs/2608.17402) ### Bookmark [![Image 2: BibSonomy](https://arxiv.org/static/browse/0.3.4/images/icons/social/bibsonomy.png)](http://www.bibsonomy.org/BibtexHandler?requTask=upload&url=https://arxiv.org/abs/2608.17402&description=MoE-ViE:%20Mixture%20of%20Experts%20Vision%20Encoder%20for%20Efficient%20Image%20and%20Video%20Understanding "Bookmark on BibSonomy")[![Image 3: Reddit](https://arxiv.org/static/browse/0.3.4/images/icons/social/reddit.png)](https://reddit.com/submit?url=https://arxiv.org/abs/2608.17402&title=MoE-ViE:%20Mixture%20of%20Experts%20Vision%20Encoder%20for%20Efficient%20Image%20and%20Video%20Understanding "Bookmark on Reddit") Bibliographic Tools # Bibliographic and Citation Tools - [x] Bibliographic Explorer Toggle Bibliographic Explorer _([What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))_ - [x] Connected Papers Toggle Connected Papers _([What is Connected Papers?](https://www.connectedpapers.com/about))_ - [x] Litmaps Toggle Litmaps _([What is Litmaps?](https://www.litmaps.co/))_ - [x] scite.ai Toggle scite Smart Citations _([What are Smart Citations?](https://www.scite.ai/))_ Code, Data, Media # Code, Data and Media Associated with this Article - [x] alphaXiv Toggle alphaXiv _([What is alphaXiv?](https://alphaxiv.org/))_ - [x] Links to Code Toggle CatalyzeX Code Finder for Papers _([What is CatalyzeX?](https://www.catalyzex.com/))_ - [x] DagsHub Toggle DagsHub _([What is DagsHub?](https://dagshub.com/))_ - [x] GotitPub Toggle Gotit.pub _([What is GotitPub?](http://gotit.pub/faq))_ - [x] Huggingface Toggle Hugging Face _([What is Huggingface?](https://huggingface.co/huggingface))_ - [x] ScienceCast Toggle ScienceCast _([What is ScienceCast?](https://sciencecast.org/welcome))_ Demos # Demos - [x] Replicate Toggle Replicate _([What is Replicate?](https://replicate.com/docs/arxiv/about))_ - [x] Spaces Toggle Hugging Face Spaces _([What is Spaces?](https://huggingface.co/docs/hub/spaces))_ - [x] Spaces Toggle TXYZ.AI _([What is TXYZ.AI?](https://txyz.ai/))_ Related Papers # Recommenders and Search Tools - [x] Link to Influence Flower Influence Flower _([What are Influence Flowers?](https://influencemap.cmlab.dev/))_ - [x] Core recommender toggle CORE Recommender _([What is CORE?](https://core.ac.uk/services/recommender))_ * [Author](https://arxiv.org/abs/2608.17402) * [Venue](https://arxiv.org/abs/2608.17402) * [Institution](https://arxiv.org/abs/2608.17402) * [Topic](https://arxiv.org/abs/2608.17402) About arXivLabs # arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? [**Learn more about arXivLabs**](https://info.arxiv.org/labs/index.html). [Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.17402) | [Disable MathJax](javascript:setMathjaxCookie()) ([What is MathJax?](https://info.arxiv.org/help/mathjax.html)) We gratefully acknowledge support from our **major funders**, [**member institutions**](https://info.arxiv.org/about/ourmembers.html), , and all contributors. [About](https://info.arxiv.org/about)·[Help](https://info.arxiv.org/help)·[Contact](https://info.arxiv.org/help/contact.html)·[Subscribe](https://info.arxiv.org/help/subscribe)·[Copyright](https://info.arxiv.org/help/license/index.html)·[Privacy](https://info.arxiv.org/help/policies/privacy_policy.html)·[Accessibility](https://info.arxiv.org/help/web_accessibility.html)·[Operational Status (opens in new tab)](https://status.arxiv.org/) Major funding support from [![Image 4: Simons Foundation](https://arxiv.org/static/base/1.0.1/images/funders/simons-foundation.png)](https://www.simonsfoundation.org/)[![Image 5: Simons Foundation International](https://arxiv.org/static/base/1.0.1/images/funders/simons-foundation-international.png)](https://www.sfi.org.bm/)[![Image 6: Schmidt Sciences](https://arxiv.org/static/base/1.0.1/images/funders/schmidt-sciences.png)](https://www.schmidtsciences.org/)