Fetching the paper…

On Linear Representations and Pretraining Data Frequency in Language Models · Around