@article{usingmechanisticinterpretabilitytocraft,
  title = {Using Mechanistic Interpretability to Craft Adversarial Attacks against Large Language Models},
  author = {Thomas Winninger and Boussad ADDAD and Katarzyna Kapusta},
  year = {2025},
  eprint = {2503.06269},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2503.06269v2},
}