What is Tree of Thoughts (ToT)?
TL;DR
A prompting technique that branches LLM reasoning into a tree structure and explores the optimal thinking path.
Tree of Thoughts (ToT): Definition & Explanation
Tree of Thoughts (ToT) is an advanced prompting technique that extends Chain-of-Thought (CoT) by branching the LLM's thinking process into a tree structure rather than a linear sequence, exploring multiple reasoning paths in parallel. At each branching point, the model itself evaluates whether a thinking path is promising and prunes unpromising branches early, efficiently reaching the optimal solution. ToT is particularly powerful for complex problems requiring trial and error, such as puzzles, mathematical reasoning, and planning tasks. By combining search algorithms like BFS (breadth-first search) and DFS (depth-first search), ToT has been reported to solve problems that traditional CoT cannot. Implementation requires multiple API calls, so the balance between cost and accuracy must be considered.