SMT-LIB

SMT-LIB
Name	SMT-LIB
Genre	Standardization effort
Established	2000s
Main location	International
Website	SMT-LIB

SMT-LIB SMT-LIB is a standardized input language and benchmark repository for satisfiability modulo theories, widely used in formal methods, automated reasoning, and software verification. The project coordinates contributions from research groups and industrial partners to support interoperability among theorem provers, model checkers, and static analyzers, aligning with initiatives in academic conferences and laboratories. SMT-LIB influences toolchains across universities and companies, informing competitions and workshops organized by professional societies and consortia.

History

The origins trace to collaborations among researchers affiliated with institutions such as Stanford University, SRI International, IBM, Microsoft Research, and NASA Ames Research Center, converging with participants from Carnegie Mellon University, École Polytechnique, CNRS, University of Oxford, and University of California, Berkeley. Early milestones involved coordination with conference series like CADE and CP, workshops at Dagstuhl and meetings tied to events such as CAV and TACAS, alongside input from projects funded by agencies including NSF, DARPA, and the European Research Council. Development evolved through drafts and revisions influenced by prominent researchers associated with groups at MIT, ETH Zurich, University of Cambridge, University of Toronto, and University of Illinois Urbana-Champaign. The format matured alongside solver progress from teams at Z3 (Microsoft Research), CVC (Stanford/INRIA), Yices (SRI), Boolector (Boolector team), and academic tools emerging from University of Waterloo and University of Freiburg.

Syntax and Language Features

The language specifies a parenthesized s-expression syntax drawing on traditions from Lisp, with definitions for symbols, sorts, and terms inspired by work at Princeton University, Brown University, Columbia University, University of Pennsylvania, and Cornell University. Commands for declaring functions, asserting formulas, checking satisfiability, and producing models align with interactive protocols used by engines developed at Microsoft Research, Google Research, NVIDIA Research, Siemens, and Airbus. Typed identifiers and sort declarations reflect type theory influences explored at Harvard University, Yale University, University of Chicago, and New York University. Extensions for quantifiers, datatypes, arrays, and uninterpreted functions match capabilities in solvers from teams at INRIA, KTH Royal Institute of Technology, EPFL, and Max Planck Institute for Informatics. The format supports producing models and proofs useful for verification workflows in groups at Bell Labs, Oracle, Red Hat, and Intel Research. Compatibility with input/output conventions used by tools at Facebook AI Research, Adobe Research, Toyota Research Institute, and Xerox PARC helped adoption.

Logic and Theories

SMT-LIB formalizes many logics and theories including propositional logic, linear integer and real arithmetic, bit-vectors, arrays, uninterpreted functions, algebraic datatypes, and floating-point arithmetic, topics investigated at Caltech, Johns Hopkins University, Duke University, Rutgers University, and University of Maryland. Theories map to solver implementations from teams at SRI International, Microsoft Research, IBM Research, Amazon AWS, and Google DeepMind. Specialized theories for strings and regular expressions reflect collaborations with researchers at University of Texas at Austin, University of Pennsylvania, Peking University, and Tsinghua University. Semantics and decidability results cited by contributors from Princeton University, ETH Zurich, University of Munich, and University of Pisa shaped the repository’s logic profiles. Benchmarks exercise decision procedures developed in groups at Verimag, University of Freiburg, UCLA, and University of Southampton.

Standard Library and Benchmarks

The SMT-LIB repository contains a corpus of benchmarks, benchmarks suites, and a standard library of sorts and functions curated with input from SV-COMP, SAT Competition, SMT-COMP, CAV, and ETAPS organizers, along with datasets from teams at Google Research, Microsoft Research, MITRE, and NASA Jet Propulsion Laboratory. Benchmarks cover verification tasks, synthesis instances, and symbolic execution problems produced by groups at Carnegie Mellon University, University of Illinois, ETH Zurich, and University of Oxford. Standardized scripts and benchmark metadata follow practices promoted by committees including members from IEEE, ACM, INRIA, and leading laboratories at LAAS-CNRS. The collection facilitates reproducible evaluation efforts by participants from Siemens, Bosch, Toyota, Vodafone, and academic labs at Imperial College London.

Tools, Solvers, and Adoption

Major solvers that accept the language include projects such as Z3, CVC variants, Yices, MathSAT, and Boolector, developed by teams at Microsoft Research, Stanford University, SRI International, Fondazione Bruno Kessler, and Ulm University, used in products and research by Google, Amazon, Intel, ARM Holdings, and Qualcomm. Integration targets include verification frameworks like Dafny, SPIN, CBMC, Frama-C, KLEE, and Coq toolchains, with connectors created by groups at ETH Zurich, Inria, University of Cambridge, and University of Southampton. Industrial adoption spans verification workflows at NASA, Airbus, Siemens, Thales, and Lockheed Martin, while academic curricula at MIT, Stanford, Cambridge, Oxford, and Tsinghua University incorporate SMT-LIB-based tooling.

Community and Governance

Governance and community stewardship involve representatives from universities, companies, and research institutes including Microsoft Research, Stanford University, INRIA, SRI International, ETH Zurich, IBM Research, and NASA Ames Research Center, coordinating through competitions and workshops at SMT-COMP, CAV, TACAS, CADE, and ETAPS. Stakeholders participate in mailing lists and working groups connected to professional bodies such as ACM, IEEE, ERC, and national funding agencies like NSF and EPSRC. The project’s development reflects collaborative norms similar to those in open-source ecosystems hosted by organizations like GitHub and governance practices mirrored in consortia including W3C and IETF.

Category:Satisfiability Modulo Theories