CICEROv2

CICEROv2是由密歇根大学和新加坡科技与设计大学联合创建的数据集，包含8,351个实例，来源于2,379个对话。该数据集旨在通过多视角上下文常识推理任务，提供多个合理但独立的推理答案，涵盖原因、后续事件、动机和情感反应四个维度。创建过程中，通过人工标注确保答案的多样性和准确性。CICEROv2的应用领域包括自然语言处理中的对话理解和常识推理，旨在提高模型对复杂对话情境的理解能力。

CICEROv2 is a dataset jointly created by the University of Michigan and the Singapore University of Technology and Design. It contains 8,351 instances sourced from 2,379 conversations. This dataset aims to provide multiple plausible yet independent reasoning answers via multi-perspective contextual commonsense reasoning tasks, covering four dimensions: causes, subsequent events, motivations, and emotional reactions. During the dataset creation process, manual annotation was adopted to ensure the diversity and accuracy of the answers. The application fields of CICEROv2 include dialogue understanding and commonsense reasoning in natural language processing, with the goal of enhancing models' ability to comprehend complex conversational contexts.