Learn Before
Example

MASS-style Masked Language Modeling

MASS-style masked language modeling is a variant of the mask-predict framework where multiple consecutive tokens in an input sequence are masked, and the model is trained to predict the entire masked span. For example, if the input is provided as [C] The kitten [M] [M] [M] ball ., the model's objective is to independently output the original contiguous masked words, such as is, chasing, and the, at their respective positions. This task can be applied to encoder-only or encoder-decoder architectures.

0

1

Updated 2026-04-17

Contributors are:

Who are from:

Tags

Foundations of Large Language Models

Ch.1 Pre-training - Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences