In our last few posts, we explored what large language models (LLMs) are, how their data is collected and curated, and how that data is broken down into tokens. Now comes the question at the heart of it all: How do these models actually learn?