@misc{indiciae5803b68d1d67,
  title = {On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models},
  author = {Leon Sick and Lukas Hoyer and Yannick Strümpler and Christian Reisswig and Timo Ropinski and Federico Tombari},
  year = {2026},
  url = {https://arxiv.org/abs/2610.32770},
  note = {Source identifier: 2610.32770}
}