Hinted Algorithmic Number Theory (HANT)
收藏资源简介:
HANT数据集由30个算术数论算法任务和相应的30个计算问题组成,每个问题都附有9个不同的提示策略。数据集包括60个文本文档,每个文档包括问题陈述、9个教学提示和正确的算法或解决方案。数据集旨在评估大型语言模型在算术数论领域的算法和计算任务上的性能。数据集已被证明能够以至少95%的准确率解决每个问题,这表明在算术数论这一高度专业化的数学领域,LLM具有强大的性能。
The HANT dataset comprises 30 arithmetic number theory algorithmic tasks and 30 corresponding computational problems, with each problem accompanied by 9 distinct prompting strategies. The dataset contains 60 text documents, each encompassing the problem statement, 9 instructional prompts, and the correct algorithm or solution. This dataset is intended to evaluate the performance of large language models (LLMs) on algorithmic and computational tasks within the field of arithmetic number theory. It has been verified that each problem in this dataset can be solved with an accuracy of at least 95%, which demonstrates the robust performance of LLMs in this highly specialized mathematical domain of arithmetic number theory.




