Sample many, keep the majority

Different samples take different reasoning paths but tend to converge on the same correct answer, while the wrong answers scatter. Voting over the final answers turns that asymmetry into accuracy.

It costs you N times the compute, so reserve it for the questions that matter. A useful middle ground is to escalate to self-consistency only when the first answer looks low-confidence.