mlfoundations-dev/multiple_samples_ground_truth_numina_aime

这个数据集包含了问题、解决方案以及不同模型生成的回应等字段。具体包括问题字段（problem）、解决方案字段（solution）、70B蒸馏模型回应字段（r1_distill_70b_response）、原始行索引字段（__original_row_idx）、多数回应字段（_majority_responses）、验证过的70B蒸馏模型回应字段（verified_r1_distill_70b_response）和会话字段（conversations，包含发送者from和消息value）。数据集分为训练集（train），共有1522个例子，总大小为156505013字节。

This dataset includes fields such as problems, solutions, and responses generated by different models. Specifically, it contains fields for the problem, solution, 70B distillation model response (r1_distill_70b_response), original row index (__original_row_idx), majority response (_majority_responses), verified 70B distillation model response (verified_r1_distill_70b_response), and conversations (including the sender from and message value). The dataset is split into a training set (train) with a total of 1522 examples and a size of 156505013 bytes.