next-token prediction

Next-token prediction is a common task in natural language processing where a model predicts the next word or token in a sequence based on the preceding context. It is the foundation for many language models, including transformers.

9 papers