What is Self-consistency?
Self-consistency runs the same reasoning prompt several times and takes the majority answer — trading a few extra calls for noticeably fewer wrong answers.
A single chain-of-thought run can take a wrong turn. Sampling several independent runs and voting cancels out one-off mistakes; on maths and logic benchmarks it adds a reliable accuracy gain over a single run.
It's a developer technique (you need to run the prompt multiple times and compare), but the everyday version works too: ask the same hard question twice in fresh chats and see whether the answers agree.
How to use it well
- Use it only on questions with a checkable answer.
- Three to five runs is usually enough.
Go deeper
Related terms
Use it right now
Ask our brain anything on the homepage — it remembers the whole conversation — or write a brief in the Studio and see the prompt it compiles to.