{"ID":6023451,"CreatedAt":"2026-07-08T01:00:23.257252134Z","UpdatedAt":"2026-07-10T08:15:11.905439937Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.05992","arxiv_id":"2607.05992","title":"PluraMath: Extending Mathematical Reasoning Evaluation Beyond High-Resource Languages","abstract":"Mathematical reasoning has become a central task for evaluating and tuning reasoning Large Language Models (LLMs), yet existing benchmarks remain heavily biased toward high-resource languages, with English and Chinese dominating both pre-training corpora and evaluation suites. The recently released PolyMath (Wang et al., 2025) dataset represents a significant step forward, yet its coverage is still limited to 18 only high-resource languages. To address this gap, we introduce PluraMath, an extension of PolyMath to 18 additional {underrepresented languages spanning 6 language families -- ranging from mid-resource to extreme low-resource settings. We constructed the dataset through a human-curated pipeline, where native speakers thoroughly validated pre-computed translations. Using PluraMath, we then benchmark 27 reasoning LLMs across four model scales -- small, mid-size, large, and closed-source ensembles -- probing the multilingual mathematical reasoning capabilities of state-of-the-art models under diverse linguistic conditions. Our fine-grained analysis confirms a persistent gap in mathematical reasoning performance between high-resource and underrepresented languages, with stronger results largely associated with better instruction-following ability. We fully open-source our dataset, data acquisition pipeline, and evaluation framework, with the goal of lowering the barrier to multilingual benchmark development for underrepresented communities.","short_abstract":"Mathematical reasoning has become a central task for evaluating and tuning reasoning Large Language Models (LLMs), yet existing benchmarks remain heavily biased toward high-resource languages, with English and Chinese dominating both pre-training corpora and evaluation suites. The recently released PolyMath (Wang et al...","url_abs":"https://arxiv.org/abs/2607.05992","url_pdf":"https://arxiv.org/pdf/2607.05992v1","authors":"[\"Daryna Dementieva\",\"Nikolay Babakov\",\"Kathy Hämmerl\",\"Ilseyar Alimova\",\"Jindřich Libovický\",\"Shu Okabe\",\"Miras Baisbay\",\"Lukas Edman\",\"Abrorkhon Inomkhujaev\",\"Antonia Karamolegkou\",\"Mateusz Lango\",\"Volkan Özer\",\"Nikola Selic\",\"Subhankar Swain\",\"Tsedeniya Kinfe Temesgen\",\"Galit Bary Weisberg\",\"Alexander Fraser\"]","published":"2026-07-07T08:25:29Z","proceeding":"cs.CL","tasks":"[\"cs.CL\",\"cs.AI\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false}