abstract
- Ensuring correctness and reliability in research software is often challenging due to the oracle problem: the absence of known, expected results for complex simulations and data analyses. Here, GeoMetaMorph is presented, an automated framework prototype that supports identifying metamorphic relations within research software projects that operate on geometrically representable input data (e.g., images, coordinate grids, volumetric models, etc.). By detecting invariant transformations, GeoMetaMorph automatically derives likely metamorphic relations that serve as effective alternatives for oracles. The approach reduces manual effort in designing test suites, improves fault detection in research software, and promotes reproducible computational research.