This model is a fine-tuned version of distilbert-base-german-cased, specifically optimized for the detection of German-language disinformation and fake news.
Model Description
The model distinguishes between "Real" and "Fake" news articles.
Model Type: Transformer-based text classification
Language: German (de)
Base Model:distilbert-base-german-cased
Training Data
The model was trained on a consolidated dataset of approximately 387,000 German news articles, sourced from various datasets including:
GermanFakeNC
GERMA
GermanNewsDataset
Example
input_text: "Die Bundesregierung plant eine neue Steuer auf Atemluft ab 2026." output: "fake"
input_text: "Der Bundestag hat heute über den neuen Haushaltsplan für das kommende Jahr debattiert." output: "real"
