这是论文“Evaluating Large Language Models Trained on Code”中描述的 HumanEval 问题解决数据集的评估工具。它用于测量从文档字符串合成程序的功能正确性。它由 164 个原始编程问题组成，评估语言理解、算法和简单的数学，还有一些 类似于简单的软件面试问题。
Australian Radio Talkback (ART) is a set of transcribed recordings of samples of national, regional and commercial Australian talkback radio from 2004 to 2006. It consists of transcriptions of 27 audi
