Who it's for
Platform
Resources
subtopic
Learn the one big idea: an LLM is a decoder-only Transformer (the GPT family from the NLP module) trained to predict the next token. It writes an answer one token at a time, each time choosing from a list of probabilities — that is why the same question can give different answers.
Builds on: 2.2 The Transformer, 5.4 How text is generated