On The Origin of Cultural Biases in Language Models: From Pre-training Data to Linguistic Phenomena

Open in new window