8Fai/Mathbench-German

Mathbench-German是一个德语数学评估数据集，专为在德语学校风格的数学问题解决上对通用和推理聚焦的语言模型进行基准测试而设计。数据集包含750个德语评估示例，组织为一个单一的测试集。每个示例包括用户提示、评分标准、难度标签、类别标签、参考答案和详细解答。数据集强调答案的正确性、简洁的推导和清晰的基准测试，覆盖了常见的中学和早期高中阶段的数学技能。

Mathbench-German is a compact German-language math evaluation set designed for benchmarking general-purpose and reasoning-focused language models on school-style mathematical problem solving in German. The dataset contains 750 evaluation examples written in German and organized as a single test split. Each example includes a user prompt, rubric items for scoring, a difficulty label, a category label, a reference answer, and a worked reference solution. It emphasizes answer correctness, concise derivations, and clean benchmarking ergonomics over broad conversational coverage, spanning a range of common secondary-school and early upper-secondary math skills.