mlfoundations-dev/multiple_samples_majority_consensus_tiger_math

该数据集包含了对话指令和响应的相关信息，具体字段包括指令种子(instruction_seed)、响应种子(response_seed)、实际响应(response)、原始行索引(__original_row_idx)、对话信息(conversations)、70B蒸馏模型响应(r1_distill_70b_response)、70B蒸馏模型提取的答案(r1_distill_70b_extracted_answer)、多数响应(_majority_responses)和验证过的70B蒸馏模型响应(verified_r1_distill_70b_response)。数据集分为训练集(train)，共有7个示例，总大小为650266字节。

The dataset contains information related to dialogue instructions and responses, including fields such as instruction_seed, response_seed, actual response (response), original row index (__original_row_idx), conversation information (conversations), 70B distillation model response (r1_distill_70b_response), extracted answer from the 70B distillation model (r1_distill_70b_extracted_answer), majority responses (_majority_responses), and verified 70B distillation model response (verified_r1_distill_70b_response). The dataset is split into a training set (train) with a total of 7 examples and a size of 650266 bytes.