Tree of Thoughts

[clear condor]

1 Sentence Summary

The research paper "Tree of Thoughts: Deliberate Problem Solving with Large Language Models" introduces a new framework, "Tree of Thoughts" (ToT), that enhances the problem-solving abilities of language models by allowing them to consider multiple reasoning paths and self-evaluate decisions, significantly improving performance on tasks requiring non-trivial planning or search.

Bullet Points Summary:

• The paper introduces a new framework for language model inference called "Tree of Thoughts" (ToT).
• ToT allows language models to perform deliberate decision-making by considering multiple reasoning paths and self-evaluating choices.
• The framework is based on the concept of "thoughts", coherent units of text that serve as intermediate steps toward problem-solving.
• The model maintains a tree of these thoughts and uses search algorithms to systematically explore this tree.
• The authors tested ToT on three tasks requiring non-trivial planning or search: Game of 24, Creative Writing, and Mini Crosswords.
• The results showed that ToT significantly enhanced the problem-solving abilities of the language models.
• For instance, in the Game of 24, while GPT-4 with chain-of-thought prompting only solved 4% of tasks, the ToT method achieved a success rate of 74%.
• The authors suggest that future work could explore more advanced search algorithms within the ToT framework.

[narrow timber]

Do you have examples?

[clear condor]

There are examples in the original paper (google the exact title to find it since we cant paste links here)

There are also some PoC implementations on github