Perplexity
Domain: NLP Fundamentals
🎯 Motivation
Measures how surprised a language model is by text. Lower perplexity means the model expected the text better.
📋 Concrete Example
A good language model has low perplexity on well-written English and high perplexity on random gibberish.
🔗 Analogy
Like a prediction game - if you guessed the text right, you are not perplexed; if it shocked you, perplexity is high.