How much context or “training data” do you give an LLM before asking it a major question? (e.g., 10% of the necessary info, 50%, 90%). Share your input ratio!
Depends what I want it to do. I would say for ChatGPT I would create a project and give it instructions and lots of urls, pdfs, docs then ask it to analyze (no need to give me the output yet) and break it down for me in simple terms what has been analyzed, what concepts and what it thinks I should provide additional resources on. Feed extra resources. Then when I’m happy with the breakdown response ask it for output in a structured way.