@inproceedings{fan-etal-2026-lang,
    title = "{LANG}: Reinforcement Learning for Multilingual Reasoning with Language-Adaptive Hint Guidance",
    author = "Fan, Yuchun  and
      Li, Bei  and
      Li, Peiguang  and
      Wang, Yilin  and
      Mu, Yongyu  and
      Yang, Jian  and
      Chen, Xin  and
      Weng, Rongxiang  and
      Wang, Jingang  and
      Cai, Xunliang  and
      Zhu, JingBo  and
      Xiao, Tong",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.2029/",
    pages = "43838--43866",
    ISBN = "979-8-89176-390-6",
    abstract = "Reinforcement learning has proven effective for enhancing multi-step reasoning in Large Language Models (LLMs), yet its benefits have not fully translated to multilingual contexts. Existing methods struggle with a fundamental trade-off: prioritizing input-language consistency severely hampers reasoning quality, while prioritizing reasoning often leads to unintended language drift toward English. We address this challenge with LANG, a novel framework that leverages language-conditioned hints to guide exploration in non-English reasoning tasks. Our method incorporates two key mechanisms to prevent dependency on these hints: a progressive decay schedule that gradually withdraws scaffolding, and a language-adaptive switch that tailors learning horizons to specific language difficulties. Empirical results on challenging multilingual mathematical benchmarks reveal that LANG substantially enhances reasoning performance without compromising language consistency. Moreover, we show that our framework generalizes beyond mathematics, fostering more consistent language alignment across model layers."
}