This revision is from 2024/07/16 09:04. You can Restore it.

By adding "step by step" the model outputs a rationale. Then the user nurses the model to a better rationale. The fine-turning automates this fine-tuning with many questions.