disi-unibo-nlp/MathGames

MathGames是一个包含2,183个高质量数学问题的数据集，这些问题包括纯文本和多媒体（文本+图像）形式，旨在评估大型语言模型在开放式的数学和逻辑推理任务上的表现。该数据集来源于国际数学和逻辑游戏锦标赛，涵盖了从小学到大学的各个年龄段，并根据所需的数学和逻辑技能对问题进行了分类。数据集的结构包括问题的唯一标识符、年份、比赛阶段、是否包含图像、比赛类别、主题领域、问题文本、答案以及可选的图像。

MathGames is a dataset containing 2,183 high-quality mathematical problems, which include both text-only and multimodal (text + image) formats, designed to evaluate the performance of large language models on open-ended mathematical and logical reasoning tasks. The dataset is sourced from the International Mathematical and Logical Games Championships, covering a range of age groups from elementary to university level, and categorizing problems based on the required mathematical and logical skills. The structure of the dataset includes a unique identifier for each problem, the year of the competition, the competition phase, indication of whether an image is included, the competition categories, the subject domain, the problem text, the solution, and an optional image.