Improving automation in interactive theorem provers by efficient encoding of lambda-abstractions

Abstract

Hammers are tools for employing external automated theorem provers (ATPs) to improve automation in formal proof assistants. Strong automation can greatly ease the task of developing formal proofs. An essential component of any hammer is the translation of the logic of a proof assistant to the format of an ATP. For- malisms of state-of-the-art ATPs are usually first-order, so some method for eliminating lambda-abstractions is needed. We present an experimental comparison of several combinatory abstraction al- gorithms for HOL(y)Hammer – a hammer for HOL Light. The al- gorithms are compared on problems involving non-trivial lambda- abstractions selected from the HOL Light core library and a library for multivariate analysis. We succeeded in developing algorithms which outperform both lambda-lifting and the simple Scho ̈nfinkel’s algorithm used in Sledgehammer for Isabelle/HOL. This increases the ATPs’ success rate on translated problems, thus enhancing au- tomation in proof assistants.