Scientific software is, by its very nature, complex. It is mathematical and highly optimized which makes it prone to subtle bugs not as easily detected by traditional testing. We outline how symbolic execution can be used to write tests similar to traditional unit tests while providing stronger verification guarantees and apply this methodology to a sparse matrix algorithm.
@article{arxiv.2510.13424,
title = {Verifying a Sparse Matrix Algorithm Using Symbolic Execution},
author = {Alexander C. Wilton},
journal= {arXiv preprint arXiv:2510.13424},
year = {2025}
}