Day: January 18, 2026

Evaluating Perplexity on Language Models

A language model is a probability distribution over sequences of tokens. When you train a language model, you want to measure how accurately it predicts human language use. This