NuclearQA
收藏资源简介:
NuclearQA是由太平洋西北国家实验室创建的一个专为评估核领域语言模型的人工基准。该数据集包含100个由专家精心设计的问题,覆盖物理、材料科学、化学等多个相关领域,旨在测试语言模型在核科学领域的理解和推理能力。数据集的创建过程中,特别注重问题的多样性和难度平衡,确保能够全面评估模型的能力。NuclearQA的应用领域主要集中在核科学研究和技术评估,旨在解决现有语言模型在专业科学领域理解不足的问题。
NuclearQA is a human-curated benchmark developed by the Pacific Northwest National Laboratory, designed specifically for evaluating language models in the nuclear domain. This dataset contains 100 expertly crafted questions spanning multiple relevant disciplines including physics, materials science and chemistry, with the objective of testing the understanding and reasoning abilities of language models in the field of nuclear science. During the construction of this dataset, special emphasis was placed on both the diversity of questions and the balance of difficulty levels, to ensure a holistic and comprehensive assessment of model capabilities. The primary application scenarios of NuclearQA are nuclear science research and technical evaluation, and it is developed to address the insufficient comprehension of specialized scientific fields exhibited by existing language models.




