gupta-tanish/Ultrafeedback-SimPO-seed-191

该数据集包含了用于训练和测试的文本数据，每个数据点包括一个提示(prompt)、提示ID(prompt_id)、选定的文本内容(chosen)、被拒绝的文本内容(rejected)、对话消息(messages)，以及选定文本和被拒绝文本的评分(score_chosen和score_rejected)。数据集分为训练和测试两部分，训练部分又分为偏好训练(train_prefs)、SFT训练(train_sft)、生成训练(train_gen)，测试部分分为偏好测试(test_prefs)、SFT测试(test_sft)和生成测试(test_gen)。

The dataset consists of text data for training and testing, with each data point including a prompt, prompt ID, chosen text content, rejected text content, conversation messages, and scores for the chosen and rejected text. The dataset is divided into training and testing sets, with the training set further divided into preference training (train_prefs), SFT training (train_sft), and generation training (train_gen), and the testing set divided into preference testing (test_prefs), SFT testing (test_sft), and generation testing (test_gen).