sodabori/20251106_BoNComp_32_2_agg_last_debug

该数据集包含与问题解决相关的特征，如问题、解决方案、答案和主题。此外，还包括与模型性能相关的特征，如得分、梯度范数和完成令牌。数据集旨在评估模型在解决问题方面的性能，包括不同级别的预测。数据集还提供了关于模型配置的详细信息，如配置名称和特征数据类型。The dataset includes features related to problem-solving, such as problem, solution, answer, and subject. It also includes features related to the models performance, such as score, grad_norm, and completion_tokens. The dataset is focused on evaluating the performance of a model in solving problems, as indicated by the presence of features like preds, preds@1, preds@2, etc., which represent predictions at different levels of granularity. The dataset also provides detailed information about the models configuration, such as the configuration name and feature data types.