| title | Reinforcement Learning Elicits Contextual Learning of Unseen Language Translation |
|---|---|
| authors | Hanxu Hu, Zdeněk Šnajdr, Pinzhen Chen, Jannis Vamvas, Rico Sennrich |
| date | 2026-06-04 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2606.06428 |
| affiliation | University of Zurich, Department of Computational Linguistics |
| upvotes | 25 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Prior work has shown that large language models (LLMs) can translate unseen or low-resource languages by undergoing continued training or even by encoding a grammar book in their context. However, both methods typically overfit specific languages, with limited zero-shot transfer at test time. To translate extremely low-resource languages at scale, we argue that LLMs must acquire the meta-skill of utilizing in-context linguistic knowledge rather than memorizing specific languages. In this paper, we propose a reinforcement learning (RL) approach to unseen language translation given rich linguistic context, using a surface-level translation metric (chrF) as the reward. Empirically, despite the lightweight reward, our RL-trained models effectively extract and apply relevant linguistic information from the provided context, leading to better translations on completely unseen languages than in-context learning or supervised fine-tuning. Our analyses suggest that outcome-based RL can extend beyond conventional reasoning tasks like math and coding to serve as a recipe for language learning from context. |
cite
@misc{hu2026reinforcement,
title={Reinforcement Learning Elicits Contextual Learning of Unseen Language Translation},
author={Hanxu Hu and Zdeněk Šnajdr and Pinzhen Chen and Jannis Vamvas and Rico Sennrich},
year={2026},
eprint={2606.06428},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2606.06428}
}