Pretraining, fine-tuning and RLHF

After this lesson you can: Name the three stages that turn raw text prediction into an assistant.

First: Attention and the transformer

Watch

• Reinforcement Learning from Human Feedback (RLHF) Explained - IBM Technology, 11 min. Focused RLHF explainer from a reliable technical channel covers the final assistant-alignment stage.
• Fine Tuning LLM Explained Simply - codebasics, 7 min. Short English explainer likely clarifies fine-tuning, the key middle step between base model and assistant.

Notes

A modern chat assistant is not created in one step. The usual story has three stages: pretraining, supervised fine-tuning, and RLHF or a related preference-optimization step. Pretraining teaches a model to predict the next token across a huge text corpus, producing a base model that has broad language and world-pattern knowledge but no built-in obligation to behave like a helpful assistant.

Unlock the full lesson notes, the exercises and the graded checkpoint - ₹5,000 once.