[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-138467-105":59,"doc-detail-138467-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","doracycle-domain-oriented-adaptation-of-unified-generative-model-in-multimodal-cycles-cvpr-2025-paper","DoraCycle - Domain-Oriented Adaptation of Unified Generative Model in Multimodal Cycles - CVPR 2025 paper","","Adapting generative models to specific domains offers a practical path to meet specialized requirements, yet complex domains often need substantial paired data to learn the target distributions. DoraCycle addresses this by training with unpaired images and texts, leveraging the bidirectional vision-language mappings already learned by a unified generative model. The method builds two multimodal cycles, text-to-image-to-text and image-to-text-to-image, optimizing cross-entropy at cycle endpoints to enable self-evolution without annotated pairs. Experiments show effective stylization adaptation using only unpaired data, and improved adaptation for new identity knowledge with a small paired set plus larger-scale unpaired data.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/doracycle-domain-oriented-adaptation-of-unified-generative-model-in-multimodal-cycles-cvpr-2025-paper/138467/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/doracycle-domain-oriented-adaptation-of-unified-generative-model-in-multimodal-cycles-cvpr-2025-paper/138467.png","ImageObject",300,407,{"name":92,"@type":93},"Maeve","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-01","2026-08-23",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",12,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"为什么 DoraCycle 能在缺少配对数据的情况下进行领域适配？","Question",{"text":113,"@type":114},"DoraCycle利用统一生成模型预训练阶段学到的双向视觉-语言映射，通过多模态循环在同一模态端点计算交叉熵损失，从而不依赖标注的图文配对监督。","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"DoraCycle 的核心训练由哪些循环构成？",{"text":118,"@type":114},"它集成了两种多模态循环：text-to-image-to-text（T cycle）和 image-to-text-to-image（I cycle）。",{"name":120,"@type":111,"acceptedAnswer":121},"在不同任务类型上，DoraCycle 对配对数据的需求有何差异？",{"text":122,"@type":114},"对不依赖成对知识的任务（如风格化），可仅用非配对数据完成有效适配；对需要新配对知识（如特定身份）的任务，结合少量配对样本与更大规模的非配对数据即可获得良好效果。","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},138467,1787481583,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":105,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":130,"read_time":41,"preview_page_count":4},5909877438554,"https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272","This CVPR paper is the Open Access version, provided by the Computer Vision Foundation.  \nExcept for this watermark, it is identical to the accepted version; the final published version of the proceedings is available on IEEE Xplore.  \nDoraCycle: Domain-Oriented Adaptation of Unified Generative Model  \nin Multimodal Cycles  \nRui Zhao, Weijia Mao, Mike Zheng Shou*  \nShow Lab, National University of Singapore  \nAbstract  \nAdapting generative models to specific domains presentsan effective solution for satisfying specialized requirements. However, adapting to some complex domains remains challenging, especially when these domains require substantial paired data to capture the targeted distributions. Since unpaired data from a single modality, such as vision or language, is more readily available, we utilize the bidirectional mappings between vision and language learned by the unified generative model to enable training on unpaired data for domain adaptation. Specifically, we propose DoraCycle, which integrates two multimodal cycles: text-toimage-to-text and image-to-text-to-image. The model is optimized through cross-entropy loss computed at the cycle endpoints, where both endpoints share the same modality. This facilitates self-evolution of the model without reliance on annotated text-image pairs. Experimental results demonstrate that for tasks independent of paired knowledge, such as stylization, DoraCycle can effectively adapt the unified model using only unpaired data. For tasks involving new paired knowledge, such as specific identities, a combination of a small set of paired image-text examples and larger-scale unpaired data is sufficient for effective domain-oriented adaptation. The code will be released at [https://github.com/showlab/DoraCycle](https://github.com/showlab/DoraCycle).  \n1. Introduction  \nThe adaptation of pre-trained generative models to specific domains is an important aspect of advancing personalized content creation, from stylized media outputs to customized identity generation [47, 62] . However, effectively adapting generative models to complex domains remains challenging, particularly when these domains require extensive amounts of paired data to accurately capture the desired distributions. For instance, learning the visual styles and character identities across a unique movie, which involves understanding multiple characters, their relationships, and di-  \n*Corresponding Author.  \n(a) Training on paired data (b) Training in Multimodal Cycles  \nFigure 1 . Training paradigms for unified generative models. (a) Traditional training involves using paired image-text data and optimizing the unified model with paired losses for both image-to-text (I2T) and text-to-image (T2I) generation tasks. (b) In contrast, the proposed multimodal cycle training framework leverages unpaired images and texts. By using cycle-consistency losses, the unified model learns to maintain consistency between input and output across modalities, enabling adaptation without the need for extensive paired datasets.  \nverse settings, is a highly complex task that demands vast amounts of paired frames and captions data. Collecting such paired data, especially for multimodal tasks involving vision and language, is often laborious and impractical, limiting the potential to adapt generative models at scale.  \nHigh-quality image-text paired data is relatively rare and scarce, but unpaired images and texts are readily available in our daily lives, such as video websites, image platforms, and content from novel websites. Therefore, we aim to explore whether it is possible to adapt generative models to target domains based on unpaired data. To achieve this goal, it is crucial for the model to have an internal capability to align the two modalities to some extent. Fortunately, recent advancements in unified generative models have encouraged us to pursue this direction.  \nRecent advanced unified generative models [14, 18, 58, 67, 77] have shown great potentia","cbCaiduna2UQAr9A","https://ap.wps.com/l/cbCaiduna2UQAr9A","pdf",1213335,"English","# Introduction\n## Motivation and challenge of paired data\n## Bidirectional mapping in unified generative models\n## Proposed DoraCycle framework with multimodal cycles\n## Training procedure and cycle-consistent learning","[{\"question\":\"为什么 DoraCycle 能在缺少配对数据的情况下进行领域适配？\",\"answer\":\"DoraCycle利用统一生成模型预训练阶段学到的双向视觉-语言映射，通过多模态循环在同一模态端点计算交叉熵损失，从而不依赖标注的图文配对监督。\"},{\"question\":\"DoraCycle 的核心训练由哪些循环构成？\",\"answer\":\"它集成了两种多模态循环：text-to-image-to-text（T cycle）和 image-to-text-to-image（I cycle）。\"},{\"question\":\"在不同任务类型上，DoraCycle 对配对数据的需求有何差异？\",\"answer\":\"对不依赖成对知识的任务（如风格化），可仅用非配对数据完成有效适配；对需要新配对知识（如特定身份）的任务，结合少量配对样本与更大规模的非配对数据即可获得良好效果。\"}]","DoraCycle - Domain-Oriented Adaptation of Unified Generative Model in Multimodal Cycles - CVPR 2025 paper | PDF"]