SATGL: An Open-Source Graph Learning Toolkit for Boolean Satisfiability
Hongtao Cheng, Jiawei Liu, Jianwang Zhai, Mingyu Zhao, Cheng Hong Yang, Chuan Shi · 2024
As the first proven NP-complete problem, the Boolean Satisfiability (SAT) problem holds significant theoretical value and has wide-ranging practical applications. It has also led to the development of numerous SAT-related tasks, such as MaxSAT and UNSAT core prediction. Due to the high complexity of handling these SAT-related tasks and the natural conversion of SAT formulas into graph structures, researchers have recently developed various graph learning methods to assist in prediction. However, these methods are often experimented on different datasets, with different approaches and different tasks, making it challenging to conduct unified evaluations and develop new algorithms. In this paper, we introduce the SATGL toolkit, the first open-source graph learning toolkit for the SAT problem. We expect SATGL to contribute to the advancement of artificial intelligence (AI) for SAT, facilitating SAT solving and new algorithm design.