Advanced NLP Interview Questions, issue 3, Dec 11, 2025
The Attention Entropy Illusion
Senior ML Engineer interview at OpenAI, and the interviewer asks:
“How do we use the Attention mechanism’s weights to measure the model’s uncertainty?”
Attention looks like a probability distribution, but it cannot tell you how certain the model is, here’s the hidden flaw.
The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.