← All notes

Summary

Scaling Laws for Neural Language Models (2020) gave the first widely used quantitative picture of how language-model loss falls as you scale parameters, dataset size, and compute. The paper turned a loose bigger-is-better intuition into a set of power-law trends that labs could use to plan training runs.

What it introduced: empirical scaling laws relating cross-entropy loss to model size, data, and training compute, and an analysis of the tradeoffs those curves imply for compute-efficient training.

What it led to: the later compute-optimal work often called Chinchilla, which is a pointer note in this batch, and the scale-up path behind GPT-3, Llama, and other foundation models. Those later papers are easier to read if you start here.

Where to read it: https://arxiv.org/abs/2001.08361 (v1, 23 Jan 2020).

The paper is posted on arXiv under the arXiv non-exclusive distribution license. This note links to it rather than reproducing it. The note itself is CC BY 4.0.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.

@misc{iil_note_v8vd4sm9,
  author = {{'iinventedlight (AI-assisted)'}},
  title = {{Scaling Laws for Neural Language Models}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-11},
  url = {https://iinventedlight.com/library/notes/2001-08361-scaling-laws-for-neural-language-models/},
  note = {Permanent ID iil:note:v8vd4sm9. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2001-08361-scaling-laws-for-neural-language-models.md.