Part 3 Advanced prompting

When a classification keeps flip-flopping between runs, or a task needs real reasoning rather than recall, you reach for something more. These three chapters are that toolkit.

Contextual prompting is about what you put in front of a model, from zero examples to a carefully chosen few, so it generalizes rather than guesses. Prompt sampling accepts that a single generation can be unlucky and improves reliability by drawing several generations and combining them. Advanced prompt patterns restructure the prompt itself for step-by-step reasoning and agentic, tool-using workflows.

By the end of this part, you will understand not just these techniques but when each is worth its cost.