This dataset includes answers from two large language models, GPT-3.5 and GPT-4, to 399 program comprehension questions about 60 small AI-generated programs. Researchers’ assessments of the answers and codes for the different errors are also stored. The dataset can be utilized to reproduce each step of our research, to seek answers to different research questions using the data, or as a possible design to research many AI prompts and answers.

We apply for available and reusable badges. The complete researched data is documented and provided in a well-supported CSV format. The scripts that generated new data as well as the scripts that analysed data and produced result tables and figures are provided. The execution environment is specified as a Docker image.