MFQ-30-LLM
收藏arXiv2025-09-30 收录
官方服务：
资源简介：
该数据集是一个旨在评估大型语言模型道德推理能力的基准数据集，采用二元的“同意/不同意”格式，针对单个道德陈述进行评估。该数据集涵盖了包括LLaMA-2和GPT-4在内的多种大型语言模型，重点关注它们在理解和回应道德陈述方面的表现。这项任务被称为道德推理评估。
This dataset is a benchmark designed to evaluate the moral reasoning capabilities of large language models (LLMs). It adopts a binary "agree/disagree" format to assess individual moral statements. This dataset includes multiple LLMs such as LLaMA-2 and GPT-4, focusing on their performance in understanding and responding to moral statements. This task is termed moral reasoning evaluation.
提供机构：
AGI Research
