Phase 6 · Modern LLM engineering · Post-training
Base models vs assistants
What post-training adds to a pretrained model.
Why it matters
Every assistant you use started as a base model: a document completer trained on raw text. The behavior you think of as "the model" (answering questions, following instructions, using a turn structure, declining some requests, a consistent voice) is added afterwards, in post-training. This phase is about that layer, which is also the layer you can change yourself with fine-tuning.
What you'll be able to do
- Contrast a base model's behavior with an assistant's on the same prompt
- Explain how a conversation becomes one token sequence through a chat template
- List what post-training adds (and what it mostly doesn't) and the usual stages
Before this lesson
Next in this phase: Supervised fine-tuning