TPTP, CASC and the development of a semantically guided theorem prover

Abstract

The first-order theorem prover SCOTT has been through a series of versions over some ten years. The successive provers, while retaining the same underlying technology, have used radically different algorithms and shown wide differences of behaviour. The development process has depended heavily on experiments with problems from the TPTP library and has been sharpened by participation in CASC each year since 1997. In the present paper, we outline some of the difficulties inherent in designing and refining a theorem prover as complex as SCOTT, and explain our experimental methodology. While SCOTT is not one of the systems which have been highly optimised for CASC, it does help to illustrate the influence of both CASC and the TPTP library on contemporary theorem proving research. Show more