It's clear with papers like "Attention is All You Need" and "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks" that the authors didn't realize how much of an impact they would have. What underrated papers have you read that you speculate will have a similar impact in the future?