arXiv preprint / cs.CV

When Scene Text Hijacks the Scene: Uncovering, Exploiting, and Mitigating Rendered-Text Semantic Leakage in Image Generation Models

作者：Feifei Li, Runjie Wang, Xiaohan Zhang, Zhenxing Qian, Mi Wen, Mi Zhang · 单位：Fudan University, † Fudan University, Shanghai University of Electric Power · 会议/期刊：arXiv preprint · 方向：cs.CV · 发布日期：2026-10-09

2026-10-09视觉感知
新标签打开 AlphaXiv
When Scene Text Hijacks the Scene: Uncovering, Exploiting, and Mitigating Rendered-Text Semantic Leakage in Image Generation Models figure
AlphaXiv 中文概览（可滚动查看）