Screen2Words
收藏arXiv2025-09-30 收录
官方服务：
资源简介：
该数据集通过 Mechanical Turk 研究让人类评估生成的屏幕摘要的质量，同时验证了自动评价指标与人类判断之间的相关性。这项任务的目的是对模型响应进行人工评估。
This dataset leverages Mechanical Turk human annotation studies to have human raters evaluate the quality of generated screen summaries, while validating the correlation between automatic evaluation metrics and human judgments. The objective of this evaluation task is to perform manual assessments of model responses.
提供机构：
Mechanical Turk
