Case Study

Evaluating Model Performance in MLM Training

A language model is being trained to predict masked words in sentences. The training objective for a given sentence is to maximize the sum of the log-probabilities of the correct words at the masked positions. A higher sum (a value closer to zero) indicates the model was more confident in its predictions.

Given the two training steps detailed below, calculate the objective value for each step and determine in which step the model's overall prediction for the masked words was better. Explain your reasoning.

0

1

Updated 2025-10-09

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