New 'Tree of Thoughts' Strategy Guides AI Through Reasoning Steps for Improved Performance
-
Researchers developed a new prompting strategy called Tree of Thoughts (ToT) that guides language models through reasoning steps to produce better outputs.
-
ToT allows exploring multiple reasoning paths in a tree structure, evaluating each step before settling on a final decision.
-
ToT is inspired by dual process theory of human cognition with fast, intuitive vs slower, analytical thinking.
-
ToT outperformed other prompting strategies like input-output, chain of thoughts, and self-consistency on a math game.
-
Combining deliberative search from classical AI with strengths of language models is a promising direction.