REC-Data

该数据集是一个多任务混合型数据集，覆盖了广泛的大型语言模型（LLM）能力范围，旨在微调通用目的的大型语言模型自动评估器。该数据集包含了多种评估类型，如成对评估、点对点评估、开放式评估和引用任务，并经过基于规则的质检流程，以确保合成数据的有效性。规模上，该数据集大约包含14万个数据点。任务内容涉及对生成文本在不同维度上的评估，如忠实度、指令遵循、连贯性和完整性。

This dataset is a multi-task hybrid dataset that covers a wide range of large language model (LLM) capabilities, and is designed for fine-tuning general-purpose automatic LLM evaluators. It includes multiple evaluation types such as pairwise evaluation, point-to-point evaluation, open-ended evaluation and citation tasks, and has undergone a rule-based quality inspection process to ensure the validity of the synthetic data. In terms of scale, this dataset contains approximately 140,000 data points. The task content involves evaluating generated texts across different dimensions, such as faithfulness, instruction following, coherence and completeness.