Buffer of Thoughts (BoT): A Novel Thought-Augmented Reasoning AI Approach for Enhancing Accuracy, Efficiency, and Robustness of LLMs

A novel thought-augmented reasoning approach called Buffer of Thoughts (BoT) is introduced to enhance the accuracy, efficiency, and robustness of Large Language Models (LLMs). BoT utilizes shared thought-templates to improve precision, streamline reasoning, and enhance LLMs' ability to consistently solve similar issues. Show more