Perplexity is a statistical measure derived from the probability of the next word appearing in a sequence. A lower perplexity indicates that the model assigns higher probabilities to the more correct words in a given sequence. A higher perplexity means than higher probabilities are given to less correct words. A common metaphor used to explain perplexity is “surprise”. A model with higher perplexity is said to be more “surprised” by the output of the model.