Learning for Dynamic subsumption

TL;DR

This paper introduces a dynamic subsumption technique for Boolean CNF formulas that enhances SAT solver performance by simplifying clauses during conflict analysis, leading to improved results especially on crafted instances.

Contribution

It presents a novel dynamic subsumption method that uses conflict analysis to simplify clauses in real-time within SAT solvers, improving efficiency.

Findings

Improved solver performance on crafted instances

Effective clause reduction during conflict analysis

Integration with Minisat and Rsat enhances results

Abstract

In this paper a new dynamic subsumption technique for Boolean CNF formulae is proposed. It exploits simple and sufficient conditions to detect during conflict analysis, clauses from the original formula that can be reduced by subsumption. During the learnt clause derivation, and at each step of the resolution process, we simply check for backward subsumption between the current resolvent and clauses from the original formula and encoded in the implication graph. Our approach give rise to a strong and dynamic simplification technique that exploits learning to eliminate literals from the original clauses. Experimental results show that the integration of our dynamic subsumption approach within the state-of-the-art SAT solvers Minisat and Rsat achieves interesting improvements particularly on crafted instances.