CFC
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/visipedia/caltech-fish-counting
下载链接
链接失效反馈官方服务：
资源简介：
该数据集由来自CommonGen数据集的含缺失槽的上下文句子组成，并为这些缺失槽值加入了众包注释。它通过允许多个正确答案，捕捉了常识的概率特性，并采用聚类方法对回应进行分类。该数据集规模包括101对（上下文，缺失槽）共10,100条注释，其任务是通过开放式生成进行常识推理和评估。
This dataset consists of context sentences with missing slots sourced from the CommonGen dataset, paired with crowdsourced annotations for these missing slot values. It captures the probabilistic nature of commonsense by permitting multiple valid answers, and adopts clustering methods to categorize model responses. The dataset contains a total of 10,100 annotations across 101 (context, missing slot) pairs, and its task targets commonsense reasoning and evaluation via open-ended generation.
提供机构：
Amazon Mechanical Turk
