Glossary Library

Technical GlossaryGenerative AI and LLM

Autoregressive Decoding

A generation mode in which the model produces output token by token using previous outputs as context.

Autoregressive decoding is the most common generation mechanism in modern large language models. Each new token is produced on top of the context generated so far. While this provides flexible and powerful generation, it also introduces major engineering issues such as latency, error accumulation, and the cost of long outputs.

You Might Also Like

Explore these concepts to continue your artificial intelligence journey.

Glossary Cover

yapay-zeka-temelleri

Generative AI

A class of AI systems capable of generating new content such as text, images, audio, video, or code.

Glossary Cover

veri-bilimi-ve-veri-yonetimi

GAN-Based Synthetic Data

A synthetic data approach based on generating new data samples similar to the real distribution using generative adversarial networks.

Glossary Cover

veri-bilimi-ve-veri-yonetimi

Diffusion-Based Synthetic Data

A modern synthetic data generation approach that reconstructs data distributions through noise injection and reverse sampling.