Technical Report: Impact of Position Bias on Language Models in Token Classification

Mehdi Ben Amor, Michael Granitzer, Jelena Mitrović

Language Models (LMs) have shown state-of-the-art performance in Natural Language Processing (NLP) tasks. Downstream tasks such as Named Entity Recognition (NER) or Part-of-Speech…