DodecaDialogue

该数据集是一个包含12种子数据集的多任务对话基准，旨在让对话代理具备提问、回答问题、利用知识资源以及感知图像的能力。此外，该数据集还鼓励发展多样化的对话技能，任务类型为多任务对话。

This dataset is a multi-task dialogue benchmark consisting of 12 sub-datasets, which aims to equip dialogue agents with the abilities to ask questions, answer questions, leverage knowledge resources, and perceive images. Additionally, this dataset is designed to foster the development of diverse dialogue skills, with its task type being multi-task dialogue.