TIDE: Making Task-Agnostic Backdoors Harder to Erase in Pre-trained Language Models

Yan Meng, Zhigang Lu 0001, Bing Li 0002, AnAn Du, Shuchao Pang. TIDE: Making Task-Agnostic Backdoors Harder to Erase in Pre-trained Language Models. In Siqi Ma 0001, Surya Nepal, editors, Information Security and Privacy - 31st Australasian Conference, ACISP 2026, Perth, WA, Australia, July 6-9, 2026, Proceedings, Part II. Volume 16793 of Lecture Notes in Computer Science, pages 265-284, Springer, 2026. [doi]

@inproceedings{MengLLDP26,
  title = {TIDE: Making Task-Agnostic Backdoors Harder to Erase in Pre-trained Language Models},
  author = {Yan Meng and Zhigang Lu 0001 and Bing Li 0002 and AnAn Du and Shuchao Pang},
  year = {2026},
  doi = {10.1007/978-981-92-3015-0_12},
  url = {https://doi.org/10.1007/978-981-92-3015-0_12},
  researchr = {https://researchr.org/publication/MengLLDP26},
  cites = {0},
  citedby = {0},
  pages = {265-284},
  booktitle = {Information Security and Privacy - 31st Australasian Conference, ACISP 2026, Perth, WA, Australia, July 6-9, 2026, Proceedings, Part II},
  editor = {Siqi Ma 0001 and Surya Nepal},
  volume = {16793},
  series = {Lecture Notes in Computer Science},
  publisher = {Springer},
  isbn = {978-981-92-3015-0},
}