Method for Interpreting Fake News Detection Results Using Large Language Model
Herald of Khmelnytskyi National University. Technical Sciences, Vol. 359, No. 6.2, pp. 149-156
Editorial summary
The method shows which words influence a news classification and how similar articles cluster in the model’s feature space. Analysts can inspect errors and refine the model through an interactive review cycle. DistilBERT is evaluated on LIAR, FakeNewsNet and CONSTRAINT-2021.
Bibliographic reference
Use the publisher’s record for the citation format required by your journal or organisation.
Stefaniia Vovk, Pavlo Radiuk, Tetiana Skrypnyk. "Метод інтерпретування результатів виявлення фейкових новин за великою мовною моделлю". Herald of Khmelnytskyi National University. Technical Sciences, Vol. 359, No. 6.2, pp. 149-156, 2025. https://doi.org/10.31891/2307-5732-2025-359-91
Test a method against your business task.
Describe the decision your system needs to support and the data you have. We assess whether a method fits, define a prototype and agree how to measure its quality, response time and operating cost.