Proving and Disproving Equivalence of Functional Programming Assignments
summary
A tool implemented on top of the Stainless verifier automatically proves or disproves student Scala submissions by generating inductive equivalence proofs or concrete counterexamples, evaluating over 4,000 programs from real courses and 17 literature pairs, with 86% overall classification success and ≈99% success on incorrect programs. It handles recursion via functional induction, aligns helpers with function call matching, and scales across hundreds of submissions using clustering leveraging transitivity to discover intermediate reference solutions.
AI-written summary · about