G4SATBench
收藏资源简介:
G4SATBench是由多伦多大学和上海交通大学的研究团队创建的综合性SAT解决数据集,包含7个不同来源和难度的数据集。该数据集旨在为基于图神经网络的SAT解决方法提供一个公平的评估框架。数据集涵盖了从随机问题到组合问题的广泛范围,每个数据集都有三个难度级别:简单、中等和困难。G4SATBench不仅包括了先前的数据集,还引入了新的难度级别,以支持更细致的分析。数据集的创建过程经过精心设计,以避免生成平凡的案例,并确保数据集的质量和多样性。该数据集的应用领域包括机器学习和人工智能,特别是在解决布尔可满足性问题(SAT)方面,旨在提高解决策略的效率和准确性。
G4SATBench is a comprehensive SAT solving dataset developed by research teams from the University of Toronto and Shanghai Jiao Tong University, which includes 7 datasets with distinct sources and difficulty characteristics. This dataset is designed to offer a fair evaluation framework for graph neural network-based SAT solving approaches. It covers a broad spectrum of problem types spanning from random instances to combinatorial problems, with each dataset containing three difficulty levels: easy, medium, and hard. Beyond incorporating existing prior datasets, G4SATBench introduces new difficulty tiers to support more fine-grained analysis. The dataset was carefully constructed during its development to avoid trivial cases and guarantee its quality and diversity. Its application areas cover machine learning and artificial intelligence, specifically for Boolean Satisfiability Problem (SAT) solving, with the objective of enhancing the efficiency and accuracy of SAT solving strategies.

- 1G4SATBench: Benchmarking and Advancing SAT Solving with Graph Neural Networks多伦多大学 · 2024年



