@misc{indiciaef3193f183bef,
  title = {Chain-of-thought obfuscation learned from output supervision can generalise to unseen tasks},
  author = {Nathaniel Mitrani Hadida and Sassan Bhanji and Cameron Tice and Puria Radmard},
  year = {2026},
  url = {https://arxiv.org/abs/2601.23086},
  note = {Source identifier: 2601.23086}
}