LB-SAT is a two-stage solver for MAX-SAT. At the first stage, it invokes a local search procedure to
calculate an approximate optimal solution. At the second stage, taking the approximate value as an initial
upper bound, a branch and bound routine is called to find the exact solution. At each search node, like UP
and Maxsatz, LB-SAT exploits unit propagation to compute a lower bound. The lower bound is computed
in an incremental style, i.e. at each node, instead of computing the lower bound from scratch, LB-SAT
reuses the information from the previous search nodes to boost the computation and improve the lower
bound.