@misc{indiciaeec8cc27fab62,
  title = {TROLL: Trust Regions improve Reinforcement Learning for Large Language Models},
  author = {Philipp Becker and Niklas Freymuth and Serge Thilges and Fabian Otto and Gerhard Neumann},
  year = {2026},
  url = {https://arxiv.org/abs/2510.03817},
  note = {Source identifier: 2510.03817}
}