ku-nlp/jp-llm-evaluator-training

日本语言模型评估器训练数据集是一个用于训练日本语言模型评估器的数据集，它专注于从多个角度评估日本语言模型，以满足多样化的评估需求。数据集包含1000个多样化的评分量表，每个评分量表下有20个不同的指示，这些指示涵盖了在实际中有意义的各种问题。对于每个指示，都有从1到5的评分和相应的反馈。反馈是对评分的详细解释。

Japanese LLM Evaluator Training Dataset is a dataset used for training evaluators for Japanese language models, focusing on evaluating Japanese language models from multiple perspectives to meet various evaluation requirements. The dataset includes 1000 diverse score rubrics, under each rubric there are 20 different instructions covering various practically meaningful questions. For each instruction, there are scores from 1 to 5 and corresponding feedback. The feedback provides detailed explanations for the evaluation scores.