NinaCalvi/ultra-rm-truthfulness-1000-RayGRM3B

该数据集包含多个复杂的特征字段，主要用于评估模型在指令遵循、诚实性、真实性等方面的表现。数据集包含3000个训练样本，每个样本包含多个评分和注释字段，如有用性、诚实性、指令遵循、真实性等。数据集还包含了自定义系统提示、细粒度评分、总体评分等字段，用于全面评估模型的性能。

This dataset contains multiple complex feature fields, primarily used to evaluate the performance of models in areas such as instruction following, honesty, and truthfulness. The dataset includes 3000 training samples, each containing multiple scoring and annotation fields, such as helpfulness, honesty, instruction following, and truthfulness. The dataset also includes custom system prompts, fine-grained scores, overall scores, and other fields for a comprehensive evaluation of model performance.