Short Answer

Optimizing a Model's Training Strategy

An engineer is preparing bilingual sentence pairs for a model. The process involves taking a packed sequence (e.g., [CLS] sentence_one [SEP] sentence_two [SEP]) and randomly replacing some words with a [MASK] symbol. The model is then trained to predict the original words. The engineer notices the model becomes very good at predicting common grammatical words (like 'is', 'a', 'the') but performs poorly on important content-carrying words (like specific nouns and verbs). Describe a modification to the word replacement strategy that would compel the model to improve its performance on these content-carrying words. Justify your reasoning.

0

1

Updated 2025-10-04

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