arXiv cs.CV· Zekai Zhang, Yanran Zhang, Shengming Yin, Yan Shu, Zhening Liu, Bingxuan Zhao, Chenyang Zhu, Donghao Zhou, Jiahao Li, Jiahao Wang, Jie Zhang, Kaiyuan Gao, Kun Yan, Lihan Jiang, Ningyuan Tang, Qihang Ai, Rui Song, Shuo Cao, Xiao Xu, Xiaoyue Chen, Yuxiang Chen, Zhendong Wang, Zhouyang Li, Zihao Liu, Zikai Zhou, Bo Zheng, Huishuai Zhang, Dongyan Zhao, Chenfei Wu·· 2 小时前AI 评分44
Qwen-Image-PE：为文生图提示词增强学习视觉设计
Qwen-Image-PE: Learning Visual Design for Text-to-Image Prompt Enhancement
AI 导读
Qwen-Image-PE 通过两阶段训练学习视觉设计，把欠具体的用户请求转成连贯的视觉规格，用于增强文生图提示词。在 Qwen-Image-Bench 上搭配 Qwen-Image-2.1，Qwen-Image-PE-397B 较原始请求提升 7.81 分，Qwen-Image-PE-9B 超过所有评测的 PE 基线。
来源：arXiv cs.CV · arxiv.org