Search papers, labs, and topics across Lattice.
The paper introduces mAceReason-Math, a multilingual dataset of challenging math problems translated from the AceReason-Math corpus, designed for Reinforcement Learning with Verifiable Rewards (RLVR). The dataset covers 14 languages with over 10,000 samples per language, addressing the English-centric bias in existing RLVR training data. By providing high-quality translations and challenging problems, the dataset aims to improve the capabilities of large language models in multilingual math and logic reasoning.
Multilingual math reasoning just got a serious upgrade: mAceReason-Math offers a meticulously translated and cleaned dataset of challenging problems across 14 languages, purpose-built for RLVR training.
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 mul- tilingual 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.