IneqMath
收藏资源简介:
IneqMath是一个由专家精心策划的不等式问题数据集,包含200个由国际数学奥林匹克奖牌得主设计的测试问题和1252个来自高级教材的训练问题。每个训练问题都附有最多四个逐步解决方案,并且76.8%的训练问题被标注了与解决方案相关的83个定理。数据集采用非正式的语言,通过边界估计和关系预测两个子任务来评估不等式证明,旨在推动大型语言模型在不等式证明方面的能力。
IneqMath is a meticulously curated dataset of inequality problems developed by experts. It includes 200 test problems designed by International Mathematical Olympiad medalists and 1252 training problems sourced from advanced textbooks. Each training problem is accompanied by up to four step-by-step solutions, and 76.8% of the training problems are annotated with 83 theorems relevant to their solutions. The dataset uses informal language, and evaluates inequality proof capabilities through two subtasks: bound estimation and relationship prediction, aiming to advance the capacity of large language models (LLMs) in inequality proving.




