What are token-level predictions?

Token-level predictions are when a computer guesses what comes next in a sentence, one word at a time.

Imagine you're reading a storybook, and you're trying to figure out what the next word is. A token is like one of those words. So, token-level prediction is like the computer trying to guess each word as it goes along.

Think of writing a sentence like stacking bricks. Each brick is a token, maybe a letter or a whole word. The computer looks at what’s already there and tries to predict which brick comes next. It doesn’t know the whole story yet, just one piece at a time.

For example, if you write "The cat sat on the"_ and stop, the computer might guess "mat*" is the next token because it's common in stories like this.

This kind of guessing helps computers understand sentences better. They can predict not just one word but many, helping them read or even write whole paragraphs, all while learning from each guess they make!