Guided Discrete Diffusion for Constraint Satisfaction Problems

TL;DR

This paper introduces a discrete diffusion guidance method for solving constraint satisfaction problems, specifically demonstrating its effectiveness on Sudoku puzzles without requiring supervised training.

Contribution

It presents a novel discrete diffusion guidance approach tailored for CSPs, enabling unsupervised solving of Sudoku puzzles.

Findings

Successfully solves Sudoku puzzles without supervision

Demonstrates effectiveness of discrete diffusion guidance for CSPs

Provides a new approach for unsupervised problem solving

Abstract

We propose discrete diffusion guidance for constraint satisfaction problems (CSPs) and demonstrate its ability to solve Sudoku puzzles without supervision.