mlfoundations-dev/instruction_filtering_scale_up_code_base_embedding_filter_mean_per_domain_4K

该数据集是一个包含对话和评估得分的集合，具体应用场景可能是对话系统的训练和评估。数据集中的字段包括指导种子、来源、嵌入表示、正负评分、得分差、推理过程、解决方案、原始行索引、推理轨迹和对话信息。训练集包含了4000个样本。

This dataset is a collection containing conversations and evaluation scores, possibly used for training and evaluating dialogue systems. The fields in the dataset include instruction seed, source, embeddings, positive and negative scores, score difference, reasoning process, solution, original row index, reasoning trace, and conversation information. The training set contains 4000 samples.