Learn Before
Concept

Power-Law Curve of Performance Scaling

A scaling law curve, which plots test error against a variable of interest such as training dataset size, can typically be divided into three phases. At the beginning, test errors decrease slowly for a short period. In the second phase, test errors decrease drastically, forming a power law curve. In the third phase, the reduction in error slows down again as the model encounters irreducible errors that cannot be eliminated regardless of the amount of training data.

Image 0

0

1

Updated 2026-04-21

Contributors are:

Who are from:

Tags

Foundations of Large Language Models

Ch.2 Generative Models - Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences