Advanced Prompting Techniques

Once you know the basics, these techniques unlock more reliable and powerful results.

Chain-of-thought prompting

Ask the model to reason through intermediate steps before answering. This dramatically improves performance on math, logic, and multi-step tasks.

Few-shot with format anchors

Provide examples that lock in an exact output structure — like JSON or a table — so results are easy to parse and consistent.

System / role prompts

Define a persistent role and rules at the start: tone, constraints, what to avoid. This shapes the entire conversation.

Decomposition

Break a big task into smaller prompts and chain them. Each step is easier to verify and correct.

Self-critique

Ask the model to review and improve its own answer: 'Check your work for errors and revise.'

Key takeaways

• Chain-of-thought helps complex reasoning.
• Format anchors make outputs predictable.
• Decomposition and self-critique improve reliability.