CohenQu/POPE-hard-no_guide-v2-verl

该数据集是一个包含多个特征字段的集合，包括数据源、提示（内容、角色）、能力、奖励模型（地面真实、风格）、额外信息（索引、分割）以及难度等级和问题ID。数据集分为训练集和测试集，分别用于模型的训练和评估。

This dataset is a collection with multiple feature fields, including data source, prompt (content, role), ability, reward model (ground truth, style), extra information (index, split), as well as difficulty level and problem ID. The dataset is divided into training and test sets for model training and evaluation respectively.