ANALOBENCH
收藏arXiv2024-02-20 更新2024-06-21 收录
下载链接：
https://github.com/JHU-CLSP/Analogical-Reasoning
下载链接
链接失效反馈官方服务：
资源简介：
ANALOBENCH是由约翰斯·霍普金斯大学研究人员开发的数据集，包含340对经过精心挑选和编辑的类比故事。该数据集旨在通过类比推理任务评估语言模型处理长上下文和抽象概念的能力。数据集中的故事涉及多种抽象关系，通过人类注释者的多轮审阅确保质量。ANALOBENCH的应用领域包括提升语言模型在复杂和长篇故事中的类比推理能力，以及在科学、法律等领域的创新应用。
ANALOBENCH is a dataset developed by researchers at Johns Hopkins University, containing 340 carefully selected and edited analogical stories. This dataset aims to evaluate the ability of language models to handle long contexts and abstract concepts via analogical reasoning tasks. The stories in the dataset cover a wide range of abstract relationships, and their quality is ensured through multiple rounds of reviews by human annotators. Application areas of ANALOBENCH include enhancing the analogical reasoning capabilities of language models in complex and long-form stories, as well as innovative applications across fields such as science and law.
提供机构：
约翰斯·霍普金斯大学
创建时间：
2024-02-20
