Skip to main content

Paper on Slicing Accepted at ESOP 2026

· 2 min read
Philipp Schroer
Caesar Developer

Our paper "Error Localization, Certificates, and Hints for Probabilistic Program Verification via Slicing" was accepted at the European Symposium on Programming (ESOP) 2026 to be held in Turin, Italy. The authors are Philipp Schröer, Darion Haase, and Joost-Pieter Katoen. An extended version of the paper is already available on arXiv.

The paper presents theoretical foundations and the implementation of our slicing-based user diagnostics in Caesar, which we dub Brutus. On this website, you can also find the user guide-level documentation on slicing in Caesar. See also the announcement of Caesar 2.0, which introduced the first slicing implementation in Caesar.

The theory is about slices, i.e. smaller sub-programs, to provide

  1. error localization (error-witnessing slices),
  2. proof simplification (verification-witnessing slices), and
  3. preserving successful verification results (verification-preserving slices).

These notions are defined on the quantitative intermediate verification language HeyVL, and thus the theory and implementation work for all statements expressible in HeyVL, such as assume and assert statements. Based on these notions, specialized slicing-based diagnostics are provided for various proof rules, such as the induction proof rule or procedure calls (called specification statements in the paper). The correctness of the above is formally proven in the paper.

The paper also reports on the implementation of these ideas. We explain the algorithms to compute the different slice types efficiently. For error reporting, we use a binary search-based algorithm to find minimal error-witnessing slices. For the other two slice types, we compare different algorithms based on unsat core extraction, minimal unsat subset enumeration, and a direct SMT encoding of the slicing problem. Our empirical evaluation of Brutus on existing and new benchmarks shows that we can find slices that are both small and informative.