Monte Carlo Tree of Thoughts Search
The following work has been done on top of the groundbreaking paper - Tree of Thoughts: Deliberate Problem Solving with Large Language Models [Yao et al] Link to the paper- https://arxiv.org/abs/2305.10601
The paper uses Game of 24 to create a Tree of Thoughts to improve the problem-solving skills of an LLM. The prompting technique was a drastic improvement from IO, COT and Self-Consistency as you can see in the results section.
This project aims to improve the Tree of Thoughts Technique by adding a Monte Carlo Tree Search algorithm to the Tree structure. This enables freely exploring and exploiting any branch that may lead to accurate results.
The following implementation is on Game of 24. I shall continue working on the other types of tasks and hopefully turn this into a more detailed blog.