Reinforcement Learning with Semantic Rewards Enables Low-Resource Language Expansion without Alignment Tax

@inproceedings{su2026reinforcement,
  title = {Reinforcement Learning with Semantic Rewards Enables Low-Resource Language Expansion without Alignment Tax},
  author = {Zeli Su and Ziyin Zhang and Zhou Liu and Xuexian Song and Zhankai Xu and Longfei Zheng and Xiaolu Zhang and Rong Fu and Guixian Xu and Wentao Zhang},
  year = {2026},
  booktitle = {Annual Meeting of the Association for Computational Linguistics},
  eprint = {2605.14366},
  archivePrefix = {arXiv},
  doi = {10.18653/v1/2026.findings-acl.880},
  url = {https://arxiv.org/abs/2605.14366},
}