@misc{indiciae3a285619115e,
  title = {Decoupling the Benefits of Subword Tokenization for Language Model Training via Byte-level Simulation},
  author = {Théo Gigant and Bowen Peng and Jeffrey Quesnelle},
  year = {2026},
  url = {https://arxiv.org/abs/2604.27263},
  note = {Source identifier: 2604.27263}
}