一作ACL 2026Findings· Findings of ACL
P yra M ath B ench: Evaluating and Improving Mathematical Capability in Large Language Models
Zetian Ouyang · Linlin Wang · Gerard de Melo · Liang He
摘要
Despite the pivotal role of numerical reasoning as the cornerstone of mathematical capabilities in large language models (LLMs) across applications, few benchmarks evaluate LLMs by integrating numerical processing and mathematical reasoning, hindering the interpretability of failures in math task...