Multiple Choice

A language model is given the prompt 'Despite the initial positive reviews, the film's box office performance was ultimately disappointing because...' and is tasked with generating a continuation. Consider two different ways the model could process this prompt before generating the next token:

  • Method 1: When processing the prompt, the token 'disappointing' can directly see and incorporate information from the token 'positive' at the beginning of the sentence.
  • Method 2: When processing the prompt, the token 'disappointing' can only see and incorporate information from the preceding tokens, such as 'ultimately' and 'was'.

Which of the following statements best analyzes the fundamental difference in how these two methods build an understanding of the prompt?

0

1

Updated 2025-09-26

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