Improving the Stability of Type Safety Proofs in Dafny
summary
Presents a concrete method for stabilizing Dafny type-soundness proofs by making the safety predicate and then the typechecker opaque, proving small safety lemmas and inversion lemmas instead of unfolding huge defs. Applies to a Cedar-style language and completes the full soundness proof; experiments run 500 verifications show mean resource use drops and variance falls (coefficient of variation for opaque proofs < 0.2 vs > 0.2 for direct proofs, resources measured in millions). The approach noticeably improves Z3/Dafny verification stability.
AI-written summary · about