lukasthede/WikiBigEdit
收藏Hugging Face2025-03-13 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/lukasthede/WikiBigEdit
下载链接
链接失效反馈官方服务：
资源简介：
WikiBigEdit是一个大规模的基准数据集，旨在评估大型语言模型在终身知识编辑方面的性能。它包含了从2024年2月至7月间Wikidata的实时编辑中提取的超过50万个问题-答案对，用于测试LLMs在事实更新、泛化、局部性和多跳推理方面的能力。
WikiBigEdit is a large-scale benchmark designed to evaluate the performance of large language models (LLMs) in lifelong knowledge editing. It consists of over 500,000 question-answer pairs extracted from real-time edits in Wikidata between February and July 2024, aiming to test LLMs capabilities in fact updating, generalization, locality, and multi-hop reasoning.
提供机构：
lukasthede
