@misc{indiciae787b04386039,
  title = {Patronus: Identifying and Mitigating Transferable Backdoors in Pre-trained Language Models},
  author = {Tianhang Zhao and Haodong Zhao and Wei Du and Pengzhou Cheng and Junxian Li and Sufeng Duan and Haojin Zhu and Gongshen Liu},
  year = {2026},
  url = {https://arxiv.org/abs/2512.06899},
  note = {Source identifier: 2512.06899}
}