As narratives shape public opinion and influence societal actions, distinguishing between truthful and misleading narratives has become a significant challenge. To address this, we introduce the EU DisinfoTest, a novel benchmark designed to evaluate the efficacy of Language Models in identifying disinformation narratives. Developed through a Human-in-the-Loop methodology and grounded in research from EU DisinfoLab, the EU DisinfoTest comprises more than 1,300 narratives. Our benchmark includes persuasive elements under Logos, Pathos, and Ethos rhetorical dimensions. We assessed state-of-the-art LLMs, including the newly released GPT-4o, on their capability to perform zero-shot classification of disinformation narratives versus credible narratives. Our findings reveal that LLMs tend to regard narratives with authoritative appeals as trustworthy, while those with emotional appeals are frequently incorrectly classified as disinformative. These findings highlight the challenges LLMs face in nuanced content interpretation and suggest the need for tailored adjustments in LLM training to better handle diverse narrative structures.

In our article, we present the systems developed for SemEval-2023 Task 3, which aimed to evaluate the ability of Natural Language Processing (NLP) systems to detect genres and persuasion techniques in multiple languages. We experimented with several data augmentation techniques, including machine translation (MT) and text generation. For genre detection, synthetic texts for each class were created using the OpenAI GPT-3 Davinci language model. In contrast, to detect persuasion techniques, we relied on augmenting the dataset through text translation using the DeepL translator. Fine-tuning the models using augmented data resulted in a top-ten ranking across all languages, indicating the effectiveness of the approach. The models for genre detection demonstrated excellent performance, securing the first, second, and third positions in Spanish, German, and Italian, respectively. Moreover, one of the models for persuasion techniques’ detection secured the third position in Polish. Our contribution constitutes the system architecture that utilizes DeepL and GPT-3 for data augmentation for the purpose of detecting both genre and persuasion techniques.