TY  - RPRT
TI  - When Embedding-Based Defenses Fail: Rethinking Safety in LLM-Based Multi-Agent Systems
AU  - Lingxi Zhang
AU  - Guangtao Zheng
AU  - Hanjie Chen
PY  - 2026
UR  - https://arxiv.org/abs/2605.01133
ID  - 2605.01133
ER  -