Back to Glossary

What is Autoregressive?

Technical Glossary term: Autoregressive
Short Answer

Models that generate text one token at a time, using previous tokens to predict the next.

Autoregressive models generate text sequentially, where each new token is predicted based on all previously generated tokens. This approach allows the model to maintain coherence and context throughout the generation process.

Key characteristics:

  • Sequential generation: One token at a time
  • Context dependency: Each token depends on previous tokens
  • Coherence maintenance: Natural flow and consistency
  • Generation control: Can be guided by prompts and parameters

⚙️ Technical Details

Autoregressive models use probability distributions over the vocabulary to predict the next token, often with techniques like temperature sampling and top-k sampling to control randomness.