LM-Polygraph/coqa

这是一个为LM-Polygraph基准测试预处理的coqa数据集版本。数据集包含多个子集，每个子集都有训练和测试分割，每个分割包含输入和输出的字符串列。数据集是从https://huggingface.co/datasets/coqa收集并处理的，具有与源数据集相同的偏见、风险和限制。

This is a preprocessed version of the coqa dataset, specifically designed for benchmarks in LM-Polygraph. The dataset includes multiple configurations, each with train and test splits, and corresponding input and output fields. The creation of this dataset aims to separate dataset creation code from benchmarking code. The source data is from https://huggingface.co/datasets/coqa and is processed using the build_dataset.py script in the repository. The dataset contains the same biases, risks, and limitations as its source dataset.