Perplexity as an Evaluation Metric

Medium

llms OpenAI Google Anthropic

What does perplexity measure when evaluating a language model?

Multiple Choice

The number of parameters in the model How surprised the model is by the test text, with lower values indicating better predictions The ratio of correct to incorrect token predictions The average length of generated sequences

Correct!

Incorrect — try again next time!

Upgrade to Pro to unlock solutions and explanations

Perplexity as an Evaluation Metric

Multiple Choice

Related problems

Deep-dive guides