@misc{indiciaeb6f4f7a00900,
  title = {Approximation of Log-Partition Function in Policy Mirror Descent Induces Implicit Regularization for LLM Post-Training},
  author = {Zhenghao Xu and Qin Lu and Changlong Yu and Tuo Zhao},
  year = {2026},
  url = {https://arxiv.org/abs/2602.05933},
  note = {Source identifier: 2602.05933}
}