mlfoundations-dev/instruction_filtering_scale_up_math_base_embedding_filter_mean_per_domain_4K

该数据集包含多个字段，如指令种子(instruction_seed)，来源(source)，嵌入(embeddings)，正面评分均值(mean_positive_score)，负面评分均值(mean_negative_score)，评分差(difference_score)，推理(reasoning)，DeepSeek解决方案(deepseek_solution)，原始行索引(__original_row_idx)，最终推理跟踪(final_reasoning_trace)，以及会话(conversations)。数据集被分为训练集(train)，包含4000个示例，文件大小为291,273,377字节。

The dataset includes multiple fields such as instruction_seed, source, embeddings, mean_positive_score, mean_negative_score, difference_score, reasoning, deepseek_solution, __original_row_idx, final_reasoning_trace, and conversations. The dataset is split into a training set (train) containing 4000 examples, with a file size of 291,273,377 bytes.