MuTual
收藏资源简介:
MuTual是一个专为多轮对话推理设计的数据集,由浙江大学和微软亚洲研究院共同创建。该数据集包含8860个手动标注的对话,基于中国学生的英语听力理解考试。与之前的非任务导向对话系统基准相比,MuTual更具挑战性,因为它要求模型能够处理各种推理问题。数据集的应用领域主要集中在提升对话模型的推理能力,特别是在多轮对话中正确选择响应的能力。
MuTual is a dataset dedicated to multi-turn dialogue reasoning, jointly created by Zhejiang University and Microsoft Research Asia. It comprises 8,860 manually annotated dialogues sourced from English listening comprehension examinations for Chinese students. Compared with prior non-task-oriented dialogue system benchmarks, MuTual poses greater challenges, as it demands models to address a wide range of reasoning tasks. The primary application scope of this dataset lies in enhancing the reasoning capabilities of dialogue models, especially their capacity to correctly select appropriate responses in multi-turn dialogues.




