We propose a way of mitigating generative AI’s (GAI) inherent untrustworthiness by exploiting its ability to generate multiple versions of code and tests, facilitating comparative analysis across versions. Instead of relying on the quality of a single test or code module, this “Differential GAI” approach promotes more reliable quality evaluation through version diversity.
Dieser Eintrag ist Teil der Universitätsbibliographie.
Search Authors in
