Short Answer

Fate of Models in a Two-Part Pre-training Scheme

Consider a pre-training method where a smaller 'generator' model corrupts a text by replacing some words, and a larger 'discriminator' model is trained to identify which words were replaced. After this training process is complete, describe the typical fate of both the generator and the discriminator models, and explain the reasoning behind these decisions.

0

1

Updated 2025-10-06

Contributors are:

Who are from:

Tags

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

Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences

Analysis in Bloom's Taxonomy

Cognitive Psychology

Psychology

Social Science

Empirical Science

Science