sodabori/20251106_BoNComp_256_1_agg_last_20251114

该数据集包含数学问题的描述、解决方案、答案、科目、难度级别等信息，以及针对每个问题的多个生成解决方案及其相关指标，如得分、梯度范数、token数量、输入长度、对数概率、相似度矩阵等。数据集还包含了不同参数设置下的聚合得分，用于评估模型性能。

This dataset includes descriptions, solutions, answers, subjects, difficulty levels, and other information for mathematical problems, as well as multiple generated solutions for each problem along with their respective metrics such as scores, gradient norms, token counts, input lengths, log probabilities, and similarity matrices. The dataset also contains aggregated scores under various parameter settings to evaluate model performance.