nn-tech/Alloy-Bench

Alloy-Bench是一个俄语的多选题基准数据集，专门用于评估冶金和采矿工程领域的大型语言模型。该数据集遵循MMLU风格的考试设计，侧重于专业知识而非一般常识。数据集包含1,120个问题，以Parquet格式存储，每个问题包括问题文本、选项列表、正确答案和知识领域。数据集的验证分割公开可用，便于开放评估和比较。

Alloy-Bench is a Russian-language multiple-choice benchmark for evaluating large language models in the domain of metallurgy and mining engineering. It is designed in the spirit of MMLU-style exams and focuses on professional knowledge rather than general trivia. This is the validation split - publicly available for open evaluation and comparison. The dataset contains 1,120 questions, stored in Parquet format, with each question including the question text, list of options, correct answer, and knowledge area.