math-reruns/HuggingFaceTB_SmolLM2-MagPieProMT

这是一个包含问题、答案和相关评分的数据集，适用于问答系统评估。数据集包含训练集(train)，共有5000个示例。每个示例包括问题、黄金标准答案、目标、预测、子集以及三种不同的答案提取方法（lighteval、harness、qwen）提取的答案和对应的评分。

This is a dataset containing questions, answers, and related scores, suitable for问答 system evaluation. The dataset includes a training set (train) with a total of 5,000 examples. Each example includes a question, a golden standard answer, a target, a prediction, a subset, and answers extracted by three different methods (lighteval, harness, qwen) along with their corresponding scores.