trl-lib/lm-human-preferences-descriptiveness

LM-Human-Preferences-Descriptiveness数据集是一个专注于提高生成文本描述性的数据集。它包含了根据人类对描述详细程度和生动性偏好的标记为chosen（选择的）和rejected（拒绝的）的文本样本对。这个数据集可以帮助模型学习人类的描述性语言偏好，从而提高它们生成丰富、引人入胜叙述的能力。

The LM-Human-Preferences-Descriptiveness dataset is a dataset focused on enhancing the descriptiveness of generated text. It contains pairs of text samples labeled as chosen and rejected based on human preferences for detail and vividness in descriptions. This dataset helps models learn human preferences in descriptive language, improving their ability to generate rich and engaging narratives.