Self-consistency
Asking a model the same question several times and taking the most common answer.
Also: majority voting, best-of-n
Self-consistency, described by Xuezhi Wang and colleagues at Google in 2022, samples several chains of thought for one question and picks the final answer that appears most often. Different reasoning paths tend to agree on the right answer more than on any particular wrong one, so accuracy rises. A related method, best-of-n, picks the answer a reward model or verifier scores highest. Both are simple forms of test-time compute and are sometimes used, with many samples, in reported benchmark results.