perplexity objective

In language modeling, the perplexity objective is a metric used to evaluate how well a probability model predicts a sample, with lower perplexity indicating better performance in terms of predictive accuracy.