What it is
Chain of thought prompting is a technique where you ask a language model to lay out its intermediate reasoning steps before stating a final answer. It often improves accuracy on math, logic, and multi step problems.
Why it helps
Hard problems need several steps, and a model that jumps straight to an answer can skip them.
- Writing the steps gives the model room to break the problem into parts
- Each step conditions the next, so the model builds toward the answer instead of guessing
- The visible reasoning also makes mistakes easier to spot and check
You can trigger it simply, for example by adding a phrase like let us think step by step, which is called zero shot chain of thought.
Costs and cautions
The technique is not free.
- It uses more tokens, which raises latency and cost
- The written steps are a useful story but are not a guaranteed account of the true computation
- For very simple questions it can add overhead without helping
Key idea
Chain of thought prompting asks the model to reason in steps, which lifts accuracy on multi step problems at the cost of more tokens.