rspiocbis/Llama3.2-Mamba2-3B-distill-best_of_n-PRM-3233678

该数据集是基于GSM8k问题的，包含了问题、答案和多种预测结果的训练集。数据集经过特定的配置，如温度(T)、顶点选择概率(top_p)、序列长度(n)、随机种子(seed)和聚合策略(agg_strategy)。此外，还包含了评估指标，如朴素准确率、加权准确率和多数投票准确率。

This dataset is based on the GSM8k questions and includes a training set with questions, answers, and various prediction outcomes. The dataset is configured with specific settings such as temperature (T), top-p selection probability, sequence length (n), random seed (seed), and aggregation strategy (agg_strategy). Additionally, it contains evaluation metrics like naive accuracy, weighted accuracy, and majority vote accuracy.