ORAS

An Operations Research Assisted System for Logic Data Generation

Introduction

Logical reasoning ability is one of the most valued capabilities of Large Language Models (LLMs). To enhance the reasoning level and intelligence of these models, high-quality data is essential. However, obtaining such data has become increasingly challenging. Synthetic data generation has emerged as a solution to this problem. The most difficult aspect of data generation is validating the correctness of the generated data. Therefore, we propose ORAS, an Operations Research Assisted System, to generate reliable data for the training and testing of LLMs.

Part One: Text Logic puzzles

assignment problem

assertion problem

Part Two: Multi Module Two Dimension Puzzles

route problem

board fill problem

board divide problem