For language models, analogies are a tough nut to crack, study shows

Open in new window