Training: pre-training, post-training and reasoning models
Comes up in: AI engineering
This concept is being written. It gets the full five steps — explanation, predict, interview angle, practice with tests, and review cards. Here is exactly what it will cover, so you can study it elsewhere in the meantime.
What it covers
- pre-training on large text
- instruction tuning
- RLHF and related methods (RLAIF, constitutional approaches — at concept level)
- reasoning/"thinking" models and adaptive thinking/effort controls (facts)
- multimodal input
Words you will know
What you will practise
- explain to a PM why the model "knows" some things and not others
Interview questions this answers
- What is the difference between pre-training and fine-tuning?