No abstract

Delete Resolution Asymmetric Tautology (DRAT) proofs have become a de facto standard to certify unsatisfiability results from SAT solvers with inprocessing. However, DRAT shows behaviors notably different from other proof systems: DRAT inferences are nonmonotonic, and clauses that are not consequences of the premises can be derived. In this paper, we clarify some discrepancies on the notions of reverse unit propagation (RUP) clauses and asymmetric tautologies (AT), and furthermore develop the concept of resolution consequences. This allows us to present an intuitive explanation of RAT in terms of permissive definitions. We prove that a formula derived using RATs can be stratified into clause sets depending on which definitions they require, which give a strong invariant along RAT proofs. We furthermore study its interaction with clause deletion, characterizing DRAT derivability as satisfiability-preservation.

Unsatisfiability proofs in the DRAT format became the de facto standard to increase the reliability of contemporary SAT solvers. We consider the problem of generating proofs for the XOR reasoning component in SAT solvers and propose two methods: direct translation transforms every XOR constraint addition inference into a DRAT proof, whereas T-translation avoids the exponential blow-up in direct translations by using fresh variables. T-translation produces DRAT proofs from Gaussian elimination records that are polynomial in the size of the input CNF formula. Experiments show that a combination of both approaches with a simple prediction method outperforms the BDD-based method.

No abstract

No abstract