ClarQ-LLM
收藏资源简介:
ClarQ-LLM是由伦敦玛丽女王大学和广西师范大学共同创建的双语(英汉)任务导向对话评估框架,旨在评估对话代理在任务导向对话中提出澄清问题的能力。该数据集包含31种不同任务类型,每种类型有10个独特的对话场景,总计310个任务实例。数据集的创建过程涉及四名博士和研究生共计700人小时的努力,确保了任务的多样性和复杂性。ClarQ-LLM主要用于评估对话代理在任务完成过程中通过对话收集必要信息的能力,特别是在处理不确定性时的表现。
ClarQ-LLM is a bilingual (English-Chinese) task-oriented dialogue evaluation framework jointly developed by Queen Mary University of London and Guangxi Normal University. It is designed to assess the capability of dialogue agents to pose clarifying questions during task-oriented dialogues. The dataset encompasses 31 distinct task types, with 10 unique dialogue scenarios per type, resulting in a total of 310 task instances. A total of 700 person-hours of work from four doctoral and graduate researchers was invested in the dataset's development, which ensures the diversity and complexity of the included tasks. Primarily, ClarQ-LLM is utilized to evaluate dialogue agents' ability to gather necessary information via dialogue throughout task completion, particularly their performance when handling uncertain situations.




