RoleMRC
收藏资源简介:
RoleMRC是由腾讯优图实验室、英国伦敦国王学院、华威大学和Alan Turing Institute共同创建的一个细粒度角色扮演和指令遵循复合数据集。该数据集包括理想角色与人类之间的多轮对话、基于文章回答性和角色能力的角色扮演机器阅读理解、以及嵌套、多轮和优先级指令的更复杂场景。RoleMRC拥有10.2k的角色配置元池,37.9k的精心合成的角色扮演指令,以及1.4k的测试样本,是迄今为止最全面的角色扮演数据集。
RoleMRC is a fine-grained role-playing and instruction-following composite dataset jointly created by Tencent YouTu Lab, King's College London, the University of Warwick, and the Alan Turing Institute. This dataset includes multi-turn dialogues between ideal role personas and human participants, role-playing machine reading comprehension tasks that require answering questions based on provided articles and adhering to the role's defined capabilities, as well as more complex scenarios featuring nested, multi-turn and prioritized instructions. RoleMRC features a 10.2k-sized role configuration meta-pool, 37.9k meticulously synthesized role-playing instructions, and 1.4k test samples, making it the most comprehensive role-playing dataset available to date.




