ExpLang: Improved Exploration and Exploitation in LLM Reasoning with On-Policy Thinking Language Selection

TL;DR

ExpLang introduces a multilingual post-training approach for large reasoning models, enhancing exploration and exploitation during reinforcement learning by dynamically selecting thinking languages, leading to improved performance over English-only training.

Contribution

The paper presents a novel on-policy language selection method during RL training that leverages multilingual thinking to improve reasoning performance in large language models.

Findings

Outperforms English-only training with the same budget

Achieves high thinking language compliance for seen and unseen languages

Extends RL exploration space with diversified language preferences

Abstract

Current large reasoning models (LRMs) have shown strong ability on challenging tasks after reinforcement learning (RL) based post-training. However, previous work mainly focuses on English reasoning in expectation of the strongest performance, despite the demonstrated potential advantage of multilingual thinking, as well as the requirement for native thinking traces by global users. In this paper, we propose ExpLang, a novel LLM post-training pipeline that enables on-policy thinking language selection to improve exploration and exploitation during RL with the use of multiple languages. The results show that our method steadily outperforms English-only training with the same training budget, while showing high thinking language compliance for both seen and unseen languages. Analysis shows that, by enabling on-policy thinking language selection as an action during RL, ExpLang effectively extends the RL exploration space with diversified language preference and improves the RL exploitation outcome with leveraged non-English advantage. The method is orthogonal to most RL algorithms and opens up a new perspective on using multilinguality to improve LRMs.