Other author here, This started as an experiment to see how much the performance of models improves when you give them examples — basically, how big of a difference do examples actually make? We also wanted to explore whether there’s an ideal number of examples that gives the best results. Was quite fun and scalable to battle any LLMs you want…

We have a short video walkthrough of the setup here https://www.youtube.com/watch?v=z1MhXgmHbwk