@article{theanatomyofalignmentdecomposingpreferen,
  title = {The Anatomy of Alignment: Decomposing Preference Optimization by Steering Sparse Features},
  author = {Jeremias Ferrao and Matthijs van der Lende and Ilija Lichkovski and Clement Neo},
  year = {2025},
  eprint = {2509.12934},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2509.12934},
}