Quick reference

LLM glossary

Token

A chunk of text chosen by a tokenizer; it may be a word, part of a word, or punctuation.

Model weights

The learned numerical parameters that determine a model's predictions.

Inference

Using already-trained weights to generate or evaluate text.

Next-token prediction

Assigning probabilities to possible next tokens given the preceding context.

Transformer

The neural-network architecture that processes tokens using attention and other layers.

Attention

A mechanism that lets a token weigh other relevant tokens in the current context.