arXiv preprint / cs.CV

Visual-Invariance-Augmented Feature Optimal Alignment for Transferable Adversarial Attacks against Closed-Source MLLMs

作者：Xiaojun Jia, Simeng Qin, Yiming Li, Jie Liao, Sensen Gao, Ke Ma, Yang Liu, Xiaochun Cao · 会议/期刊：arXiv preprint · 方向：cs.CV · 发布日期：2026-10-07

2026-10-07视觉感知
新标签打开 AlphaXiv
Visual-Invariance-Augmented Feature Optimal Alignment for Transferable Adversarial Attacks against Closed-Source MLLMs figure
AlphaXiv 中文概览（可滚动查看）