Dataset for Textual and Functional Diversity Analysis of LLM-Generated Programming Exercises

This dataset contains generated beginner-level C programming exercise outputs produced by six large language models under controlled prompting conditions. The dataset includes model identifiers, run identifiers, temperature settings, prompt categories, prompts, raw model replies, parsed task fields, JSON validity indicators, markdown-block indicators, response-length information, timing and token metadata, and manually assigned functional grouping labels. The manual_group column was used to support the functional-diversity analysis reported in the associated manuscript.