What does perplexity measure, and when are comparisons misleading?

Instruction: Give a clear explanation, a practical example, and the main limitation.

Context: Explain a language-model metric with a small calculation and the controls needed for fair comparisons.

Updated

Official answer available

Read the opening below, then unlock the full answer and practical guidance.

For an autoregressive language model, perplexity is the exponential of the average negative log probability assigned to the evaluated tokens...

Upgrade to view official answer

Related Questions