Max Springer

I am a theoretical computer scientist and algorithms researcher focusing on large language model reasoning, safety, and algorithmic theory. 

Currently, I am a postdoctoral research fellow at Princeton's Center for Information Technology Policy. I work on the science and safety of large language models, wherein I develop rigorous frameworks for evaluating their reliability, fairness, and trustworthiness, and studying how fine-tuning and alignment reshape model behavior and reasoning. My approach to artificial reasoning is deeply informed by my early research in computational neuroscience at Yale School of Medicine and the NIH, where I worked on biological theories of cognition.

This research builds on my work in approximation algorithms and algorithmic game theory, where I design provably efficient methods for clustering, matching, and allocation with guarantees on fairness and proportionality. Across my work, I aim to combine these theoretical guarantees with empirical evaluation of AI systems. 

I obtained my PhD in Applied Mathematics from the University of Maryland, where I was advised by MohammadTaghi Hajiaghayi and my dissertation, titled "The Price of Fairness in Algorithmic Decision Making," was supported by an NSF Graduate Research Fellowship. Previously, I held research roles at Google Research and Nokia Bell Labs. I completed my undergraduate studies at Cornell University with a B.A. in Mathematics and minors in Cognitive Science and Biological Sciences.

Beyond my research, I am currently a consultant to the United Nations' Independent International Scientific Panel on AI wherein I am working as a technical editor and drafter for their inaugural reports. Moreover, I frequently contribute mathematics explainer pieces and news articles to Scientific American, serve as a fact-checker for several science news outlets, and was selected as a 2024 AAAS Mass Media Fellow.

Selected Publications

The Geometry of Alignment Collapse: When Fine-Tuning Breaks Safety

M. Springer, C.P. Lee, B. Turbal, B. Metevier, H. Jung, J. Castleman, Z. Shen, A. Korolova.

in review with NeurIPS 2026

Greedy Coordinate Diffusion: Effective and Semantically Coherent Adversarial Attacks via Diffusion Guidance

B. Turbal, M. Springer, B. Metevier, A. Korolova.

ICML 2026

Bi-criteria Metric Distortion

K. Banihashem, D. Chakraborty, S.C. Jahan, I. Gholami, M.T. Hajiaghayi, M. Mahdavi, M. Springer

ICLR 2026

Service

I serve on the inaugural AAAI Student Committee, which was formed to increase involvement from student researchers on matters pertaining to the responsible use of Artificial Intelligence. I also volunteer as an editor for Street Sense Media, Washington DC's weekly street newspaper sold by self-employed homeless distributors.

In my free time, I like to take mediocre photos on my Nikon FE.