TY  - RPRT
TI  - GuardAlign: Test-time Safety Alignment in Multimodal Large Language Models
AU  - Xingyu Zhu
AU  - Beier Zhu
AU  - Junfeng Fang
AU  - Shuo Wang
AU  - Yin Zhang
AU  - Xiang Wang
AU  - Xiangnan He
PY  - 2026
UR  - https://arxiv.org/abs/2602.24027
ID  - 2602.24027
ER  -