How an LLM is trained
Three stages turn random weights into a helpful assistant. Pick a stage to watch it step by step.
Stage 1
Pretraining
Learn language from raw text
→
Stage 2
SFT
Learn to follow instructions
→
Stage 3
RLHF
Learn what people prefer
❚❚ Pause
↺ Restart
0.25×
0.5×
0.75×
1×
2×