PipelineRL: Faster On-policy Reinforcement Learning for Long Sequence Generation

@misc{piché2025pipelinerlfasteronpolicy,
      title={PipelineRL: Faster On-policy Reinforcement Learning for Long Sequence Generatio},
      author={Alexandre Piché and Ehsan Kamaloo and Rafael Pardinas and Dzmitry Bahdanau},
      year={2025},
      eprint={2509.19128},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2509.19128},
}