Short Answer

Analyzing a Flawed Log-Likelihood Calculation

A junior data scientist proposes calculating the total conditional log-likelihood for a two-token sequence y = (y₁, y₂) given an input x using the following formula:

log Pr(y|x) = log Pr(y₁|x) + log Pr(y₂|x)

Identify the fundamental error in this approach and explain why it fails to correctly model the sequential nature of text generation.

0

1

Updated 2025-10-08

Contributors are:

Who are from:

Tags

Ch.4 Alignment - 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