True/False

When training a reward model on segment-level scores using a regression loss, the primary objective is to ensure the model's predicted scores for different segments maintain the same relative order (ranking) as the target scores, even if the absolute values of the predictions are consistently different from the targets.

0

1

Updated 2025-10-10

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