Branch-and-cut solution of inference problems in propositional logic

We describe and test computationally a branch-and-cut algorithm for solving inference problems in propositional logic. The problem is written as an integer program whose variables correspond to atomic propositions. We generate cuts for the integer program using a separation algorithm based on the resolution method for theorem proving. We find that the algorithm substantially reduces the size of the search tree when it is large. It is faster than Jeroslow and Wang's method on hard problems and slower on easy problems.