Uloženo v:
Podrobná bibliografie
Hlavní autor: Liu, Xiaoming
Médium: Recurso digital
Jazyk:angličtina
Vydáno: Zenodo 2026
Témata:
On-line přístup:https://doi.org/10.5281/zenodo.19253962
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Obsah:
  • <div>Mutation score (MS) is a single scalar value that summarizes the effectiveness of a test suite in detecting artificially seeded bugs. While widely adopted, MS loses critical subcategory-level information about how different mutation operator types contribute to the overall score. This paper proposes operator-level R´enyi entropy fingerprints as a fine-grained information-theoretic characterization of mutation testing landscapes. We conduct experiments on 5 controlled Python projects and 4 real-world Python modules.</div> <div> Our key findings are:</div> <div>(1) projects with identical MS can exhibit significantly different fingerprints, revealing that MS alone obscures important structural differences in mutation distributions;</div> <div>(2) the diversity gap metric, defined as the difference between Hartley and min-entropy, predicts operator-level vulnerability, identifying which operator categories are inadequately covered by test suites;</div> <div>(3) Literal Constant Replacement (LCR) dom inates mutation landscapes, accounting for 61% of all mutants in the dateutil-relativedelta module, indicating that test suites primarily need validation of boundary and literal values.</div> <div>The R´enyi entropy frame work provides a structured diagnostic perspective that complements traditional mutation score reporting.</div>