HiUni Unifies Visual Understanding and Generation in a Shared Backbone
The paper HiUni: Harmonizing Understanding and Generation in Shared Visual Space presents a unified multimodal model that reuses the same language-aligned ViT and dense Transformer backbone for visual understanding and generation. It adds a Pixel Diffusion Head for pixel-space generation and a lightweight HiGate pathway for fine-grained spatial details near the output stage. The authors report an average understanding score similar to the pretrained VLM and competitive image generation and editing performance, based on paper experiments.
HiUni offers a simpler architecture for unified multimodal modeling by avoiding separate visual representations or backbone computation, while retaining task-specific processing at the output stage.
Computer Science > Computer Vision and Pattern Recognition
arXiv:2611.26107 (cs)
[Submitted on 19 Nov 2026]
Title:HiUni: Harmonizing Understanding and Generation in Shared Visual Space
Authors:Jiahao Hu, Jingdong Du, Yichen Zhang, Yi Song, Zhenyu Zhang, Jiaxin Niu, Ruihang Li, Xiangzhao Hao, Yiqian Zhang, Yong Huang, Zefeng Zhang, Zheyuan Gu, Shuohuan Wang
View a PDF of the paper titled HiUni: Harmonizing Understanding and Generation in Shared Visual Space, by Jiahao Hu and 12 other authors