Enregistré dans:
Détails bibliographiques
Auteurs principaux: Xing, Jinbo, Jiang, Zeyinzi, Tuo, Yuxiang, Mao, Chaojie, Gai, Xiaotang, Chen, Xi, Zhang, Jingfeng, Pan, Yulin, Han, Zhen, Xiao, Jie, Yan, Keyu, Xie, Chenwei, Zhong, Chongyang, Zhu, Kai, Shen, Tong, Huang, Lianghua, Liu, Yu, Yang, Yujiu
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:https://arxiv.org/abs/2603.25706
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866915898046021632
author Xing, Jinbo
Jiang, Zeyinzi
Tuo, Yuxiang
Mao, Chaojie
Gai, Xiaotang
Chen, Xi
Zhang, Jingfeng
Pan, Yulin
Han, Zhen
Xiao, Jie
Yan, Keyu
Xie, Chenwei
Zhong, Chongyang
Zhu, Kai
Shen, Tong
Huang, Lianghua
Liu, Yu
Yang, Yujiu
author_facet Xing, Jinbo
Jiang, Zeyinzi
Tuo, Yuxiang
Mao, Chaojie
Gai, Xiaotang
Chen, Xi
Zhang, Jingfeng
Pan, Yulin
Han, Zhen
Xiao, Jie
Yan, Keyu
Xie, Chenwei
Zhong, Chongyang
Zhu, Kai
Shen, Tong
Huang, Lianghua
Liu, Yu
Yang, Yujiu
contents Recent unified models have made unprecedented progress in both understanding and generation. However, while most of them accept multi-modal inputs, they typically produce only single-modality outputs. This challenge of producing interleaved content is mainly due to training data scarcity and the difficulty of modeling long-range cross-modal context. To address this issue, we decompose interleaved generation into textual planning and visual consistency modeling, and introduce a framework consisting of a planner and a visualizer. The planner produces dense textual descriptions for visual content, while the visualizer synthesizes images accordingly. Under this guidance, we construct large-scale textual-proxy interleaved data (where visual content is represented in text) to train the planner, and curate reference-guided image data to train the visualizer. These designs give rise to Wan-Weaver, which exhibits emergent interleaved generation ability with long-range textual coherence and visual consistency. Meanwhile, the integration of diverse understanding and generation data into planner training enables Wan-Weaver to achieve robust task reasoning and generation proficiency. To assess the model's capability in interleaved generation, we further construct a benchmark that spans a wide range of use cases across multiple dimensions. Extensive experiments demonstrate that, even without access to any real interleaved data, Wan-Weaver achieves superior performance over existing methods.
format Preprint
id arxiv_https___arxiv_org_abs_2603_25706
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Wan-Weaver: Interleaved Multi-modal Generation via Decoupled Training
Xing, Jinbo
Jiang, Zeyinzi
Tuo, Yuxiang
Mao, Chaojie
Gai, Xiaotang
Chen, Xi
Zhang, Jingfeng
Pan, Yulin
Han, Zhen
Xiao, Jie
Yan, Keyu
Xie, Chenwei
Zhong, Chongyang
Zhu, Kai
Shen, Tong
Huang, Lianghua
Liu, Yu
Yang, Yujiu
Computer Vision and Pattern Recognition
Recent unified models have made unprecedented progress in both understanding and generation. However, while most of them accept multi-modal inputs, they typically produce only single-modality outputs. This challenge of producing interleaved content is mainly due to training data scarcity and the difficulty of modeling long-range cross-modal context. To address this issue, we decompose interleaved generation into textual planning and visual consistency modeling, and introduce a framework consisting of a planner and a visualizer. The planner produces dense textual descriptions for visual content, while the visualizer synthesizes images accordingly. Under this guidance, we construct large-scale textual-proxy interleaved data (where visual content is represented in text) to train the planner, and curate reference-guided image data to train the visualizer. These designs give rise to Wan-Weaver, which exhibits emergent interleaved generation ability with long-range textual coherence and visual consistency. Meanwhile, the integration of diverse understanding and generation data into planner training enables Wan-Weaver to achieve robust task reasoning and generation proficiency. To assess the model's capability in interleaved generation, we further construct a benchmark that spans a wide range of use cases across multiple dimensions. Extensive experiments demonstrate that, even without access to any real interleaved data, Wan-Weaver achieves superior performance over existing methods.
title Wan-Weaver: Interleaved Multi-modal Generation via Decoupled Training
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.25706