rlhfchicken/EVALppo_introex_pos_len_neg_minp0.1

该数据集包含用户评论（review）、查询（query）和响应（responses），以及与这些评论和响应相关的多个特征，如正面性（positiveness）、负面性（negativeness）、token长度（token_length）、归一化token长度（normalized_token_length）、简洁度（conciseness）等。还包括分组ID（group_id）、分组得分（group_score）、选择的响应（chosen）、拒绝的响应（rejected）以及ppo模型生成的响应及其特性。数据集分为训练集，共有25000个样本。

The dataset includes user reviews (review), queries (query), and responses (responses), along with multiple features related to these comments and responses, such as positivity (positiveness), negativity (negativeness), token length (token_length), normalized token length (normalized_token_length), conciseness, etc. It also includes group ID (group_id), group score (group_score), chosen response (chosen), rejected response (rejected), and the characteristics of responses generated by the ppo model. The dataset is split into a training set with a total of 25,000 samples.