Skip to content
Road to Intelligence
← The Lab

Adjust a model's scores for the next word and watch cross-entropy loss and perplexity respond — the exact quantity language models are trained to minimize.

Try it · toy model

Confident and Wrong

Know well6 min