2026· Conference and Labs of the Evaluation Forum· pp. 5603-5610· 1 citation· 21 references
Computer Science
TL;DR
An authorship verification method for the PAN 2026 Voight-Kampff Generative AI Detection task that combines word-level TF-IDF, word-boundary character n-gram TF-IDF, and stylometric features to capture complementary textual information.
Ebed-KNN is presented, a system that fine-tunes RoBERTa-base with a supervised contrastive objective to produce geometrically separable embeddings of human and AI-generated text, indicating that contrastively learned embedding spaces offer a competitive and robust foundation for AI-generated text detection.
Anish Thapaliya, Shushanta Pudasaini, B. Bal· Conference and Labs of the E...· 1 citation
The Sinai submission to Subtask 1 of the PAN 2026 Voight-Kampff Generative AI Detection shared task is described, a binary classification challenge covering three genres: fiction, news, and essays, by combining mStyleDistance, a multilingual stylometric encoder that captures writing style independently of content, with...
César Espin-Riofrio, Arturo Montejo Ráez, Janis Lara-Jama et al.· Conference and Labs of the E...· 1 citation
This study presents a lightweight hybrid framework that combines word-level TF–IDF features using 1–3-grams, character-level TF–IDF features using 3–6-grams, calibrated linear classification, and GPT-2-based language-model surprisal through a weighted ensemble to demonstrate the potential of combining multi-granularity...
Pravangkar Boruah· Journal of Scientific Resear...· 0 citations