Computer Science · Artificial Intelligence

Consider the following statements about the process of training Large Language Models (LLMs): A. LLMs are initially trained using a self-supervised method that involves masking tokens and predicting them. B. The neural network layers in LLMs are responsible for making predictions less accurate at each stage. C. The initial pretraining phase equips LLMs only to predict the most probable next token, not necessarily the most helpful one. Which of the following statements is/are correct?