Next-token prediction

The one thing a language model does - give every possible next token a probability. A token is picked (sampling), added to the text, and the whole text goes back in, again and again until an end token or a length limit. It's why answers stream in, vary between tries, and tend to keep an early mistake.

Depth 1 · Foundation

You can describe the predict-pick-append loop and use it to explain streaming answers, different answers to the same question, and mistakes that snowball.

Depths of understanding

1
Foundation
You can describe the predict-pick-append loop and use it to explain streaming answers, different answers to the same question, and mistakes that snowball.
Read it
- One Token at a TimeAI Agents Level 1: How Agents Work
Opens after
TokenThe model predicts one token at a time, from its tokenizer's vocabulary.ProbabilityAt each step the model gives every possible next token a probability, and one is picked from those chances.

Your path here

What's still between you and Next-token prediction, depth 1, in the order you can learn it.

Where it fits

Builds on

Token ProbabilityProbability

This concept

Next-token predictionLanguage Models

Opens up

Stateless model Thinking tokens

Builds onThis conceptOpens upToken ProbabilityProbability Next-token predictionLanguage Models Stateless model Thinking tokens