| title | Crosslingual On-Policy Self-Distillation for Multilingual Reasoning |
|---|---|
| authors | Yihong Liu, Raoyuan Zhao, Michael A. Hedderich, Hinrich Schütze |
| date | 2026-05-10 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2605.09548 |
| affiliation | CIS, LMU Munich |
| upvotes | 2 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Large language models (LLMs) have achieved remarkable progress in mathematical reasoning, but this ability is not equally accessible across languages. Especially low-resource languages exhibit much lower reasoning performance. To address this, we propose Crosslingual On-Policy Self-Distillation (COPSD), which transfers a model's own high-resource reasoning behavior to low-resource languages. COPSD uses the same model as student and teacher: the student sees only the low-resource problem, while the teacher receives privileged crosslingual context, including the problem translation and reference solution in English. Training minimizes full-distribution token-level divergence on the student's own rollouts, providing dense supervision while avoiding the sparsity and instability of outcome-only reinforcement learning (RL). Experiments on 17 low-resource African languages show that COPSD consistently improves low-resource mathematical reasoning across model sizes and substantially outperforms Group Relative Policy Optimization (GRPO). Further analyses show that COPSD improves answer-format adherence, strengthens test-time scaling, and generalizes to harder multilingual reasoning benchmarks, with especially large gains for lower-resource languages. We make our code and data available at: https://github.com/cisnlp/COPSD. |
cite
@misc{liu2026crosslingual,
title={Crosslingual On-Policy Self-Distillation for Multilingual Reasoning},
author={Yihong Liu and Raoyuan Zhao and Michael A. Hedderich and Hinrich Schütze},
year={2026},
eprint={2605.09548},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2605.09548}
}