rasdani/countdown

该数据集包含多个字段，如目标值(target)、数字序列(nums)、数据来源(data_source)、提示(prompt)、能力(ability)、奖励模型(reward_model)和额外信息(extra_info)。其中，奖励模型包含地面真实值(numbers和target)和风格(style)。数据集分为训练集(train)、验证集(val)和测试集(test)。

The dataset includes multiple fields such as target, number sequence (nums), data source (data_source), prompt, ability, reward model (reward_model), and extra information (extra_info). The reward model contains ground truth (numbers and target) and style. The dataset is split into training set (train), validation set (val), and test set (test).