Autoregressive Models
Models that generate output one token at a time, left to right.
Autoregressive models predict the next token from all previous ones, generating text sequentially. This next-token-prediction objective is what powers GPT-style language models and their open-ended generation.