Scaling Laws for Neural Language Models

Loss falls as a power law in parameters, data, and compute.

What this paper explains

Loss falls as a power law in parameters, data, and compute.

What to notice

• What problem the paper set out to solve, and why earlier approaches stalled there.
• The key mechanism or formula introduced, in the authors' own terms.
• Which results held up, and which assumptions later work relaxed.

How to read it

Read the abstract and introduction for the problem setup, then the method section for the core mechanism. Skim experiments for what actually moved.

Sources

• Authors: Kaplan et al. (2020)