TY  - RPRT
TI  - Visually Legible but Adversarial: Watermark-Constrained Attacks on End-to-End VLMs for Document Understanding
AU  - Qingzhe Han
AU  - Fan Dang
AU  - Shengqi Chen
AU  - Xinyi Sun
AU  - Haoyu Xiong
AU  - Lingkun Li
AU  - Di Zhang
PY  - 2026
UR  - https://arxiv.org/abs/2610.15728
ID  - 2610.15728
ER  -