mlfoundations-dev/instruction_filtering_scale_up_code_base_embedding_filter_mean_16K

该数据集包含多个字段，如指令种子(instruction_seed)、来源(source)、嵌入向量(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)、最终推理轨迹(final_reasoning_trace)和对话(conversations)。数据集分为训练集，包含16000个示例，总大小为1303324364字节。这些信息表明数据集可能用于某种推理或对话系统的训练。

The dataset includes 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. It is split into a training set with 16,000 examples and a total size of 1,303,324,364 bytes. This suggests that the dataset may be used for training some kind of reasoning or dialogue system.