When Thinking Backfires: Mechanistic Insights Into Reasoning-Induced Misalignment

@misc{yan2025when,
  title = {When Thinking Backfires: Mechanistic Insights Into Reasoning-Induced Misalignment},
  author = {Hanqi Yan and Hainiu Xu and Yulan He},
  year = {2025},
  eprint = {2509.00544},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2509.00544},
}