ulab-ai/Ranking-bench

该数据集包含两种配置：直接配置（direct）和迭代配置（iterative），每种配置都有训练集和测试集。数据集中的特征包括问题（problem）、正确答案（gt）、候选答案序列（candidates）和任务名称（task_name）。直接配置的训练集包含7448个示例，测试集包含1676个示例；迭代配置的训练集包含63771个示例，测试集包含14946个示例。

The dataset consists of two configurations: direct and iterative, each with its own training and test sets. The features in the dataset include the problem, the ground truth answer (gt), a sequence of candidate answers (candidates), and the task name. The direct configuration has a training set of 7448 examples and a test set of 1676 examples, while the iterative configuration has a training set of 63771 examples and a test set of 14946 examples.