wikieval
收藏arXiv2025-09-30 收录
下载链接：
https://huggingface.co/datasets/explodinggradients/wikieval
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为WikiEval，包含了2022年创建的50个维基百科页面中提出的问题。它对比了GPT-3.5在有和无上下文情况下给出的答案，以评估其忠实度。此外，两名人类标注者判定哪个答案更忠实，他们之间的人类一致性达到了95%。该任务的目的是评估答案的忠实性。
This dataset, named WikiEval, consists of questions sourced from 50 Wikipedia pages created in 2022. It compares the answers generated by GPT-3.5 with and without contextual information to evaluate the faithfulness of model responses. Additionally, two human annotators judged which answer was more faithful, with an inter-annotator agreement reaching 95%. The goal of this task is to assess the faithfulness of the answers.
提供机构：
Es et al.
