TRANSFORMERS & LLMs • 56 / 397
Follow text from characters to tokens, attention, transformers and generation.

Autoregressive Generation

Autoregressive generation produces one token, appends it to context, then predicts the next token repeatedly.

Think of it like

Think of Autoregressive Generation 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 Autoregressive Generation.

SRE lens

Long answers take time because generation happens token by token.

Remember thisAutoregressive generation produces one token, appends it to context, then predicts the next token repeatedly.
AIForSREJump to a concept
Search is optional. The main journey is simply ↓ Next.