[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85757-en":3,"doc-seo-85757-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85757,5909877438554,"Maeve","https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272",6,"Technology","Next Dense Stride Prediction for Multimodal Autoregressive Visual Modeling","DenseAR introduces a new generative paradigm that reframes autoregressive image generation as coarse-to-fine next-dense-stride prediction on a compact single-scale tokenizer. By traversing one latent grid with progressively denser strides, the method captures transitions from global structure to fine detail. DenseAR avoids raster-order slow inference via parallel token prediction and sidesteps multi-scale token-stack costs. A unified autoregressive model extends across modalities and imaging tasks. Validation includes medical multi-contrast brain MRI and ImageNet, improving FID/IS and unifying translation, generation, and tumor segmentation.","Next-Dense-Stride Prediction for Multimodal Autoregressive Visual Modeling  \nChicago Y. Park 1 , * Jialin Mao2 Xiaojian Xu2  \nTaha Kass-Hout2 Ulugbek S. Kamilov 1 Cao Xiao2  \n1University of Wisconsin–Madison 2 GE HealthCare  \n{chicago.park, [kamilov}@wisc.edu](kamilov}@wisc.edu)  \n{jialin.mao, xiaojian.xu, taha.kass-hout, [cao.xiao}@gehealthcare.com](cao.xiao}@gehealthcare.com)  \narXiv :2607 .09892v1 [ ee ss .IV] 10 Jul 2026  \nAbstract  \nWe introduce DenseAR, a new generative paradigm that reformulates autoregressive image generation as coarse-to-fine next-dense-stride prediction using a compact single-scale tokenizer. Our key insight is that traversing a single-scale latent grid with progressively denser strides naturally captures the transition from global structure to fine detail. This addresses two limitations of existing autoregressive models at once: the slow inference of raster-order autoregression, which DenseAR avoids by predicting multiple tokensin parallel, and the heavy cost of multi-scale approaches, which need long, multi-resolution token sequences to achieve coarse-to-fine prediction. Building on our efficient framework and the flexibility of autoregressive modeling, we further extend DenseAR to a unified model that handles multiple modalities and imaging tasks within a single backbone. We validate DenseAR on both medical and natural images. On multi-contrast brain MRI, a single DenseAR model unifies cross-modal translation, modality-conditioned generation, and tumor segmentation, while remaining competitive with task-specific methods. On ImageNet, DenseAR improves class-conditional generation quality (FID and IS) over both a single-grid baseline without stride ordering anda multi-scale tokenizer-based baseline. Code is available [https://github.com/uw-cig/DenseAR](https://github.com/uw-cig/DenseAR).  \n1 Introduction  \nVisual generative modeling has advanced rapidly, driven in large part by diffusion models, which produce detailed, coherent images through iterative denoising [1–4] . Following the success of Large Language Models (LLMs), an alternative paradigm treats images as discrete sequences: a visual tokenizer compresses an image into vector-quantized (VQ) tokens, and an autoregressive (AR) model predicts these tokens one by one, like words in a sentence [5–7] . Within this paradigm, how visual tokens are ordered for generation has emerged as a central design axis that trades quality against efficiency: two orders dominate visual AR—raster order, which generates tokens row by row, and next-scale order, used by visual autoregressive (VAR) modeling [6],  \n*This work was done during an internship at GE HealthCare.  \nFigure 1: Intuition for how stride density controls coarse-tofine generation: sparse strides capture global structure, while denser strides add fine detail, across medical and natural images. The bottom row visualizes the sampling density at each stride level. This figure is illustrative only: we stride in pixel space here for intuition, whereas DenseAR strides on the latent token grid, never on raw pixels.  \nwhich generates a sequence of token maps from low to high resolution. By generating global structure before fine detail—mirroring how images are naturally composed—this coarseto-fine order substantially improves generation quality over raster-order AR [6] .  \nEach order has a complementary strength and weakness. Raster order [5, 8] operates on a single, compact latent grid, but generates tokens strictly one after another; this sequential decoding makes inference slow. Next-scale order [6, 9] instead achieves coarse-to-fine generation, but at a cost: it represents each image as a stack of token maps at several resolutions. The model processes all multi-scale tokens at once, so the sequence becomes far longer than a single grid, sharply raising memory and compute. This worsens in con-  \nditional generation like image translation and editing, where each source image adds its own multi-scale stack","cbCaihVFZ6swoCV6","https://ap.wps.com/l/cbCaihVFZ6swoCV6","pdf",25069051,1,20,"English","en",105,"# Abstract\n# Introduction\n## Coarse-to-fine vs raster-order trade-offs\n## DenseAR approach and key insight\n# Contributions","[{\"question\":\"What is DenseAR’s main idea for autoregressive visual modeling?\",\"answer\":\"DenseAR predicts tokens using next-dense-stride traversal on a compact single-scale latent grid, enabling coarse-to-fine generation by progressively increasing stride density.\"},{\"question\":\"How does DenseAR address the inefficiency of raster-order autoregression?\",\"answer\":\"DenseAR avoids slow raster-order decoding by predicting multiple tokens in parallel within each stride, reducing the number of generation steps.\"},{\"question\":\"How does DenseAR unify multiple medical imaging tasks?\",\"answer\":\"A single DenseAR backbone handles multi-contrast synthesis, cross-modal translation, and tumor segmentation, using one short grid rather than multi-scale token stacks.\"}]",1784206047,50,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"next-dense-stride-prediction-for-multimodal-autoregressive-visual-modeling","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/technology/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/next-dense-stride-prediction-for-multimodal-autoregressive-visual-modeling/85757/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What is DenseAR’s main idea for autoregressive visual modeling?","Question",{"text":75,"@type":76},"DenseAR predicts tokens using next-dense-stride traversal on a compact single-scale latent grid, enabling coarse-to-fine generation by progressively increasing stride density.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does DenseAR address the inefficiency of raster-order autoregression?",{"text":80,"@type":76},"DenseAR avoids slow raster-order decoding by predicting multiple tokens in parallel within each stride, reducing the number of generation steps.",{"name":82,"@type":73,"acceptedAnswer":83},"How does DenseAR unify multiple medical imaging tasks?",{"text":84,"@type":76},"A single DenseAR backbone handles multi-contrast synthesis, cross-modal translation, and tumor segmentation, using one short grid rather than multi-scale token stacks.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,112,117,122,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":28,"slug":111},"technology",{"id":113,"doc_module":4,"doc_module_name":45,"category_name":114,"show_sort_weight":115,"slug":116},7,"Healthcare",40,"healthcare",{"id":118,"doc_module":4,"doc_module_name":45,"category_name":119,"show_sort_weight":120,"slug":121},8,"Research & Report",30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":21,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":21,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":106,"slug":136},19,"General","general"]