This repository implements a Sudoku solver using the Tree-of-Thought (ToT) framework, designed to enhance the reasoning and problem-solving capabilities of Large Language Models (LLMs). It targets researchers and developers looking to explore advanced LLM reasoning techniques beyond standard auto-regressive generation, offering a structured approach to complex tasks.

How It Works

The ToT framework augments an LLM with specialized modules: a prompter agent, a checker, a memory module, and a ToT controller. These components orchestrate multi-round interactions with the LLM, simulating human-like trial-and-error problem-solving. This approach allows the system to explore different solution paths and backtrack when necessary, overcoming the limitations of purely sequential token generation.

Quick Start & Requirements

• Install: Clone the repository and install dependencies via pip install -r requirements.txt.
• Prerequisites: Python 3.9+, OpenAI API key, and a configured config.yaml file.
• Run: Execute puzzles with python run_tot.py "puzzle_description".
• Experiments: Compare ToT with other methods using python run_expr.py <solver_type> <data_file>.
• Details: Large Language Model Guided Tree-of-Thought preprint

Highlighted Details

• Implements the Tree-of-Thought (ToT) framework for LLM reasoning.
• Includes modules for prompting, checking, memory, and control.
• Enables backtracking and exploration of solution space.
• Supports comparison against zero-shot, one-shot-cot, and few-shot-cot methods.

Maintenance & Community

The project is authored by Jieyi Long. Further community or maintenance details are not specified in the README.

Licensing & Compatibility

The repository does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The project relies on OpenAI's API, requiring an API key and incurring associated costs. The README does not detail performance benchmarks or specific LLM compatibility beyond mentioning OpenAI models.