Perplexity: Visualizing a Model's Uncertainty

Perplexity (PPL) is a measurement of how "surprised" a language model is by a sequence of words. It can be thought of as the model's confusion level when predicting the next word.

• A low perplexity score indicates the model is confident and has a narrow, focused set of likely next words.
• A high perplexity score means the model is uncertain, considering a wide range of words as equally possible.

Use the slider below to see how a model's "cone of possibilities" changes as its perplexity shifts from confident to confused.