I would be curious to understand the "How to Prompt" section more. As someone who does not interact with LLMs regularly I have no idea why this looks like a templating language.
Would anyone be able to explain what's going on in that section or point to resources that explain what the goal is / why this looks so programmatic?
It looks programmatic because this is the simplest way to describe what parameters and sentences structure the authors have found works well. I find its best to remember that LLMs are infrence engines - they are stoy telling machines. LLMs that work via instruction are building probible narrative, not responding to their instructions like orders but as a writer being asked to continue a scrip.
I would be curious to understand the "How to Prompt" section more. As someone who does not interact with LLMs regularly I have no idea why this looks like a templating language.
Would anyone be able to explain what's going on in that section or point to resources that explain what the goal is / why this looks so programmatic?