Pretraining, Supervised Fine-Tuning, and Preference Optimization
Explain how training targets shape representations, instruction following, and preferred behavior without equating preference with truth.
Explain how training targets shape representations, instruction following, and preferred behavior without equating preference with truth.