David Litman

Hi, I'm David. I'm interested in the structure of representations, human vs machine cognition, and affective neuroscience. I studied math and philosophy at the University of Chicago and finished my BS in math at UCSB.

I work in Terry Sejnowski's computational neuroscience lab at the Salk Institute. My aim is to do meaningful work at the intersection of AI and neuroscience.

I also construct crosswords (with puzzles published in the New York Times and the Wall Street Journal), and in my free time I read non-fiction, play tennis and chess, and try to learn bass.

I study how transformers represent things, mostly chess engines. My latest paper shows that raising Maia-3's skill parameter (Elo) pushes its computations deeper in the trained network: Increasing Skill Level Recruits Deeper Attention Layers in a Frozen Chess Transformer (arXiv:2609.23917).

I built chessformer-lens, an open-source library for mechanistic interpretability of transformer-based chess models.

The chessformer-lens app: a chess position with policy, attention, neuron, and logit-lens panels — Interactive app frontend for chessformer_lens

About

LinkedIn
Email is the best way to reach me: [first name] @ davidlitman.com