Meta-RecipeQA
收藏arXiv2021-10-23 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2110.11899v1
下载链接
链接失效反馈官方服务：
资源简介：
Meta-RecipeQA是一个针对多模态机器阅读理解（M3C）的数据集，由斯坦福国际研究所创建。该数据集包含8000条数据，主要用于评估模型在不同模态下理解和回答问题的能力。数据集的创建过程中，通过三个控制旋钮（Control-Knobs）来调整数据集的难度，从而生成不同难度的测试集。Meta-RecipeQA的应用领域主要集中在评估模型的泛化能力，特别是在处理多模态数据时的理解和推理能力。
Meta-RecipeQA is a dataset for multimodal machine reading comprehension (M3C), created by Stanford Research Institute International. It contains 8,000 instances, and is mainly used to evaluate models' abilities to understand and answer questions across different modalities. During the dataset construction, three control-knobs are utilized to adjust the dataset's difficulty, thereby generating test sets with varying difficulty levels. The main application scenario of Meta-RecipeQA focuses on evaluating a model's generalization capabilities, particularly its understanding and reasoning abilities when processing multimodal data.
提供机构：
斯坦福国际研究所
创建时间：
2021-10-23
