mAceReason-Math: A Dataset of High-Quality Multilingual Math Problems Ready For RLVR

TL;DR

This paper introduces mAceReason-Math, a multilingual dataset of challenging math problems designed for reinforcement learning with verifiable rewards, aiming to improve multilingual capabilities of math reasoning models.

Contribution

The creation of a high-quality, multilingual math problem dataset tailored for RLVR, covering 14 languages with over 10,000 samples each, addressing a gap in existing resources.

Findings

Provides a diverse, challenging set of math problems in multiple languages.

Enhances training signals for multilingual math reasoning models.

Facilitates benchmarking and research in multilingual RLVR.

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) has been successfully applied to significantly boost the capabilities of pretrained large language models, especially in the math and logic problem domains. However, current research and available training datasets remain English-centric. While multilingual training data and benchmarks have been created in the past, they were not created with RLVR and current model capability in mind, and their level of difficulty is often too low to provide appropriate training signals for current models. To address this gap, we provide mAceReason-Math, a dataset of high-quality translations of challenging math problems sourced from a corpus specifically curated for RLVR (AceReason-Math). We further take specific care to clean and improve our translations, resulting in a coverage of 14 languages with more than 10,000 samples per language. We release the dataset to facilitate multilingual RLVR research and benchmarking in the research community.