LLM System Design Interview, issue 13, Nov 10, 2025

The Weight Decay Illusion

ML Engineer interview at Anthropic, and the interviewer asks:

You’re pre-training a 500B parameter model. You’re only doing one epoch over a 10T token dataset, so you’re clearly not overfitting. Why on earth are you still using weight decay?

Don’t say: Weight decay is a standard regularizer. It’s just good practice to prevent overfitting.

When overfitting isn’t the problem, weight decay isn’t a regularizer - it’s an optimizer control knob shaping your convergence path.

The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.

Read it on Substack

Get the next one

Free on Substack. Unsubscribe in one click.