Theory Exploration Powered by Deductive Synthesis

Abstract This paper presents a symbolic method for automatic theorem generation based on deductive inference. Many software verification and reasoning tasks require proving complex logical properties; coping with this complexity is generally done by declaring relevant sub-properties. gives rise to the challenge of discovering useful sub-properties that can assist automated proof process. known as theory exploration problem, so far, predominant solutions emerged rely evaluation using concrete values. limits applicability these techniques programs properties. In work, we introduce new technique exploration, capable (offline) library lemmas from base set inductive data types recursive definitions. Our approach introduces abstraction overcome above limitations, combining it synthesis reason about abstract implementation has shown find more than prior art, avoiding redundant (in terms provability), while being faster in most cases. abstraction-based step toward applying synthesis.