Comment by lelanthran

I would guess:

Pre-training: refining the weights in an existing model using more training data.

Post-training: Adding some training data to the prompt (RAG, basically).