Broken Neural Scaling Laws

TL;DR — Proposes a single functional form that fits how AI model performance changes with scale, including the bends and breaks plain power laws miss.

How much weight it carries: A peer-reviewed ICLR 2023 paper; useful, though the fits are descriptive rather than explanatory.

Where this came from

• Will Manidis’s “most underrated PDF” thread, July 2023 — routing record: The Most Underrated PDF on the Internet (July 2023)

38 pages. A copy is archived locally against link rot; the header links the original source.