TY  - RPRT
TI  - Separator Injection Attack: Uncovering Dialogue Biases in Large Language Models Caused by Role Separators
AU  - Xitao Li
AU  - Haijun Wang
AU  - Jiang Wu
AU  - Ting Liu
PY  - 2025
UR  - https://arxiv.org/abs/2504.05689
ID  - 2504.05689
ER  -