rspiocbis/Llama3.2-Mamba-3B-distill-best_of_n-PRM-3221040

该数据集是基于GSM8k任务的一个变体，包含了问题和相应的答案，以及多种预测和评分结果。数据集有两个配置版本，一个用于训练，另一个包含评估指标。每个样本都包括问题文本、答案文本、生成的文本序列、分数、预测结果、预测使用的标记数量、聚合分数以及在不同预测策略下的预测结果。

This dataset is a variant of the GSM8k task, including questions and corresponding answers, along with various prediction and scoring results. There are two configuration versions of the dataset, one for training and the other containing evaluation metrics. Each sample includes the question text, answer text, generated text sequence, scores, prediction results, number of tokens used for prediction, aggregated scores, and prediction results under different prediction strategies.