This dataset is inspired by the response structure and formatting style of Perplexity's Sonar model.
It contains 20 Sonar reference examples and 427 synthetic examples generated with the GPT-5 Nano model using a few-shot
prompt, which is also included in this repository.
The dataset creation process was as follows: