mlfoundations-dev/instruction_filtering_scale_up_math_base_embedding_filter_mean_2K

该数据集包含多个字段，如指令种子、来源、嵌入向量、正面分数平均值、负面分数平均值、分数差、推理过程、DeepSeek解决方案、原始行索引和最终推理轨迹等。数据集被划分为训练集，共有2000个示例，大小为159,978,860字节。

The dataset includes multiple fields such as instruction seed, source, embedding vectors, mean positive score, mean negative score, difference score, reasoning process, DeepSeek solution, original row index, and final reasoning trace. The dataset is split into a training set with 2000 examples, totaling 159,978,860 bytes in size.