Abstract During the past decade of continuous development, the theorem prover Vampire has become an automated solver for the combined theories of commonly-used data structures. Vampire now supports arithmetic, induction, and higher-order logic. These advances have been made to meet the demands of software verification, enabling Vampire to effectively complement SAT/SMT solvers and aid proof assistants. We explain how best to use Vampire in practice and review the main changes Vampire has undergone since its last tool presentation, focusing on the engineering principles and design choices we made during this process.

Abstract To achieve the best performance, automatic theorem provers often rely on schedules of diverse proving strategies to be tried out (either sequentially or in parallel) on a given problem. In this paper, we report on a large-scale experiment with discovering strategies for the Vampire prover, targeting the FOF fragment of the TPTP library and constructing a schedule for it, based on the ideas of Andrei Voronkov’s system Spider. We examine the process from various angles, discuss the difficulty (or ease) of obtaining a strong Vampire schedule for the CASC competition, and establish how well a schedule can be expected to generalize to unseen problems and what factors influence this property.

Clause selection plays a crucial role in modern saturation-based automatic theorem provers. A commonly used heuristic suggests prioritizing small clauses, i.e., clauses with few symbol occurrences. More generally, we can give preference to clauses with a low weighted symbol occurrence count, where each symbol’s occurrence count is multiplied by a respective symbol weight. Traditionally, a human domain expert would supply the symbol weights. In this paper, we propose a system based on a graph neural network that learns to predict symbol weights with the aim of improving clause selection for arbitrary first-order logic problems. Our experiments demonstrate that by advising the automatic theorem prover Vampire on the first-order fragment of TPTP using a trained neural network, the prover’s problem solving capability improves by 6.6% compared to uniformly weighting symbols and by 2.1% compared to a goal-directed variant of the uniformly weighting strategy.

Abstract The state-of-the-art superposition-based theorem provers for first-order logic rely on simplification orderings on terms to constrain the applicability of inference rules, which in turn shapes the ensuing search space. The popular Knuth-Bendix simplification ordering is parameterized by symbol precedence—a permutation of the predicate and function symbols of the input problem’s signature. Thus, the choice of precedence has an indirect yet often substantial impact on the amount of work required to complete a proof search successfully. This paper describes and evaluates a symbol precedence recommender, a machine learning system that estimates the best possible precedence based on observations of prover performance on a set of problems and random precedences. Using the graph convolutional neural network technology, the system does not presuppose the problems to be related or share a common signature. When coupled with the theorem prover Vampire and evaluated on the TPTP problem library, the recommender is found to outperform a state-of-the-art heuristic by more than 4 % on unseen problems.