Certifying Without Loss of Generality Reasoning in Solution-Improving Maximum Satisfiability

Jeremias Berg, Bart Bogaerts, Jakob Nordström, Andy Oertel, Tobias Paxian, Dieter Vandesande
30th International Conference on Principles and Practice of Constraint Programming (CP 2024), 2024
to appear

Experimental data


Proof logging has long been the established method to certify correctness of Boolean satisfiability (SAT) solvers, but has only recently been introduced for SAT-based optimization (MaxSAT). The focus of this paper is solution-improving search (SIS), in which a SAT solver is iteratively queried for increasingly better solutions until an optimal one is found. A challenging aspect of modern SIS solvers is that they make use of complex “without loss of generality” arguments that are quite involved to understand even at a human meta-level, let alone to express in a simple, machine-verifiable proof.

In this work, we develop pseudo-Boolean proof logging methods for solution-improving MaxSAT solving, and use them to produce a certifying version of the state-of-the-art solver Pacose with VeriPB proofs. Our experimental evaluation demonstrates that this approach works in practice. We hope that this is yet another step towards general adoption of proof logging in MaxSAT solving.