Re-reading (RE2) is a surprisingly simple technique: include the question twice in the prompt. By "re-reading" the question, the model pays more attention to key details and constraints, reducing errors from misunderstanding the problem.
Despite its simplicity, RE2 shows consistent improvements across reasoning tasks, especially for complex questions.