mlfoundations-dev/seed_math_multiple_samples_scale_up_random_real_run_16K

这是一个包含对话数据的训练集，数据集包含了指令种子(instruction_seed)、数据来源(source)、模型响应(r1_distill_70b_response)等信息。每个样本都有一个唯一的索引(__original_row_idx)，并且提供了多数投票的响应(_majority_responses)和经过验证的模型响应(verified_r1_distill_70b_response)。此外，每个样本还包含了对话信息，对话信息由对话发起者(from)和对话内容(value)组成。数据集划分为训练集(train)，共有15785个样本，大小为1,033,385,653字节。This is a training dataset containing conversation data, which includes information such as instruction seeds (instruction_seed), data sources (source), model responses (r1_distill_70b_response), etc. Each sample has a unique index (__original_row_idx) and provides majority-voted responses (_majority_responses) and verified model responses (verified_r1_distill_70b_response). In addition, each sample contains conversation information, which consists of the conversation initiator (from) and the conversation content (value). The dataset is split into a training set (train) with a total of 15,785 samples and a size of 1,033,385,653 bytes.