@misc{indiciaeb27f516c2b51,
  title = {Can We Break LLMs Out of Self-Loops? Fine-Grained Reasoning Control with Activation Steering},
  author = {Sheldon Yu and Tong Yu and Xunyi Jiang and Rohan Surana and Gagan Mundada and Sungchul Kim and Lina Yao and Julian McAuley and Junda Wu},
  year = {2026},
  url = {https://arxiv.org/abs/2607.18100},
  note = {Source identifier: 2607.18100}
}