Machine Learning at the Flatiron Institute Seminar: Alberto Bietti

Date & Time

Title: Transformers and Associative Memories

Abstract: Large language models based on transformers have achieved great empirical successes. However, as they are deployed more widely, there is a growing need to better understand their internal mechanisms in order to make them more reliable. These models appear to store vast amounts of knowledge from their training data, and to adapt quickly to new information provided in their context or prompt. We study how transformers balance these two types of knowledge by considering a synthetic setup where tokens are generated from either global or context-specific bigram distributions. By a careful empirical analysis of the training process on a simplified two-layer transformer, we illustrate the fast learning of global bigrams and the slower development of an “induction head” mechanism for the in-context bigrams. We highlight the role of weight matrices as associative memories, provide theoretical insights on how gradients enable their learning during training, and study the role of data-distributional properties

About the Speaker

Alberto is a research scientist at the Center for Computational Mathematics. He received his Ph.D. in applied mathematics from Inria and Université Grenoble Alpes in 2019, and was a Faculty Fellow at the NYU Center for Data Science from 2020 to 2022. He also spent time at Inria Paris, Microsoft Research, and Meta AI. His research focuses on the theoretical foundations of deep learning.

Advancing Research in Basic Science and MathematicsSubscribe to Flatiron Institute announcements and other foundation updates