mathreward/8b_llama3_2epoch_selfcorr_horizon2_tmp1

该数据集包含5000个训练样本，每个样本具有以下特征：索引（idx）、真实值（gt）、级别（level）、类型（type）、我的解决方案（my_solu）和预测值（pred，为字符串序列）。数据集的总大小为32788528字节，下载大小为12550066字节。数据文件位于train分割路径下。

This dataset contains 5000 training samples, each with the following features: index (idx), ground truth (gt), level (level), type (type), my solution (my_solu), and predictions (pred, as a sequence of strings). The total size of the dataset is 32788528 bytes, with a download size of 12550066 bytes. The data files are located under the train split path.