Learning Hard Constraints in Max-SAT⋆

Solving over-constrained problems with Max-SAT solvers typically consists of finding an assignment that satisfies all the hard constraints and the maximum number of soft constraints. Despite the relevance of clause learning in SAT for solving structured instances, this technology has not yet been extended to Max-SAT. In this paper, we have incorporated a module that learns hard clauses in a branch and bound Max-SAT solver, and performed an empirical investigation that provides evidence that clause learning gives raise to considerable speedups in Max-SAT.