When Less Is Not More: Large Language Models Normalize Less-Frequent Terms with Lower Accuracy

Open in new window