TRANSFORMERS & LLMs • 57 / 397
Follow text from characters to tokens, attention, transformers and generation.
Next-Token Prediction
Next-token prediction is the task of estimating which token should come next given the current context.
Think of it like
Think of Next-Token Prediction as part of a very fast reader that repeatedly decides what information matters and what should come next.
Real life
Autocomplete, translation and chat assistants depend on concepts such as Next-Token Prediction.
SRE lens
The model repeatedly predicts the next piece of an incident explanation until it finishes.
Remember thisNext-token prediction is the task of estimating which token should come next given the current context.