Pre-training: refining the weights in an existing model using more training data.
Post-training: Adding some training data to the prompt (RAG, basically).