用户报告场景(URS)数据集
收藏资源简介:
用户报告场景(URS)数据集是由清华大学和蒙特利尔大学合作创建的,旨在从用户中心的角度评估大型语言模型(LLMs)的性能。该数据集包含1846个真实世界的用户交互案例,涉及7种不同的用户意图,如事实问答、专业问题解决等。数据集通过全球712名参与者的贡献,覆盖了多种文化和语言背景,确保了数据的多样性和实用性。URS数据集不仅用于评估LLMs在满足用户需求方面的有效性,还强调了在主观场景中LLMs的性能,为研究和开发提供了宝贵的资源。
The User Report Scenario (URS) dataset was collaboratively created by Tsinghua University and the University of Montreal, aiming to evaluate the performance of Large Language Models (LLMs) from a user-centric perspective. This dataset includes 1,846 real-world user interaction cases, covering seven distinct user intentions such as factual question answering and professional problem-solving, among others. Developed with contributions from 712 global participants, the dataset spans diverse cultural and linguistic backgrounds, ensuring its diversity and practical utility. The URS dataset not only serves to assess the effectiveness of LLMs in meeting user needs, but also emphasizes the performance of LLMs in subjective scenarios, providing a valuable resource for research and development.




