ciol-research/multilevel-legal-reasoning

法律推理数据集，包含多层次的人工和模型标注解释。该数据集包括150个多项选择题，涵盖简单、中等和困难三个难度级别，涉及人权、国际贸易、环境法和人道主义法律等多个领域。每个问题都有经过人工验证的真相推理和三种角色视角（一般、法官、律师）的模型生成推理痕迹。

Legal Reasoning Dataset with Multilevel Human and Model-Annotated Explanations. This dataset includes 150 multiple-choice questions covering three difficulty levels, easy, medium, and hard, across various domains such as human rights, international trade, environmental law, and humanitarian law. Each question features human-verified ground truth reasoning and three model-generated reasoning traces from the perspectives of general, judge, and lawyer.