@misc{indiciae4d64dd85666f,
  title = {BEAT: Visual Backdoor Attacks on VLM-based Embodied Agents via Contrastive Trigger Learning},
  author = {Qiusi Zhan and Hyeonjeong Ha and Rui Yang and Sirui Xu and Hanyang Chen and Liang-Yan Gui and Yu-Xiong Wang and Huan Zhang and Heng Ji and Daniel Kang},
  year = {2026},
  url = {https://arxiv.org/abs/2510.27623},
  note = {Source identifier: 2510.27623}
}