GRPO Beyond English: A Large-Scale Study of GRPO in Non-English and Multilingual Settings

A large-scale empirical study from arXiv (2608.13698v1) finds that training language models to reason in their native language via Group Relative Policy Optimization (GRPO) leaves only a small gap to English reasoning, with strong crosslingual transfer but also language-specific regressions. The study, covering multiple base models and training languages, concludes that RLVR beyond English can yield broad gains but requires broad evaluation to detect regressions.

arXiv:2608.13698v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards RLVR , often optimized with Group Relative Policy Optimization GRPO , has become a central recipe for improving the reasoning capabilities of pretrained language models but current studies remain heavily English-centric. We conduct a large-scale empirical study of multilingual and non-English GRPO across a wide range of base models, training languages, and different reasoning language rewards. We find that training to reason in the native language often leaves only a small gap to training for English reasoning. We further observe strong crosslingual transfer: training in one language often improves performance in many others. However, specific trends are highly model- and language-dependent. In some cases, training in a particular language induces severe regressions on out-of-domain capabilities in other languages. Our analysis shows that RLVR beyond English can provide broad crosslingual gains, but also requires broad evaluation to detect language-specific regressions.