Reflection on search trees (RoT) is a framework that enhances the performance of tree-search-based prompting methods by summarizing guidelines from previous search experiences to help large language models avoid repeating past mistakes.
