Tree of Thoughts: Deliberate Problem Solving with Large Language Models

1 Introduction

1. What limitation of language model inference does Tree of Thoughts address?
   Generation makes token-level decisions one at a time, left to right, so models fall short on tasks needing exploration, strategic lookahead, or where early decisions are pivotal; e.g. GPT-4 with chain-of-thought solved only 4% of hard Game of 24 problems.
   • Yao et al., 2023, Tree of Thoughts (Advances in Neural Information Processing Systems 36 (NeurIPS 2023)), §1
2. Which ideas from cognitive science and early AI inspire Tree of Thoughts?
   Dual-process theory: token-by-token generation resembles fast, associative “System 1” and could be complemented by deliberate “System 2” planning. And Newell, Shaw and Simon’s view of problem solving as search through a combinatorial problem space represented as a tree.
   • Yao et al., 2023, Tree of Thoughts (Advances in Neural Information Processing Systems 36 (NeurIPS 2023)), §1