Prompt Engineering · TRACK 02/05
Reasoning Techniques
Making models think step by step — and knowing when that helps.
// THE TRACK
01 · START HEREChain-of-ThoughtUnderstand why "think step by step" became the most famous phrase in prompting, what it actually does, and which tasks it helps most.BEGINNERSelf-ConsistencyLearn how running the same prompt several times and voting on the answer trades tokens for accuracy — and how to decide if it is worth it.INTERMEDIATETree-of-ThoughtSee how tree-of-thought turns one linear reasoning chain into an explored, scored search tree — and why it is overkill for most tasks.INTERMEDIATEReasoning ModelsLearn what a reasoning model actually does with its hidden thinking tokens, what that costs you, and when a standard model is the smarter pick.BEGINNERStep-Back PromptingLearn how a single 'first, what general principle applies here?' step lifts a model's accuracy on hard questions, and where it adds value over plain chain-of-thought.INTERMEDIATELeast-to-Most PromptingSee how splitting a hard task into ordered subproblems and feeding each answer forward helps models tackle problems too big for a single reasoning pass.INTERMEDIATESelf-Critique & Self-RefineLearn the critique-then-revise loop that catches a model's own mistakes, plus the honest limits — including when self-correction makes answers worse.INTERMEDIATE