Nathan Stromberg

Nathan Stromberg

Fourth year PhD student at Arizona State University in Electrical Engineering. Currently researching robust finetuning of large models for fairness.

Papers Accepted to NeurIPS 2025

less than 1 minute read

Published:

I am thrilled to announce that two of our papers have been accepted to NeurIPS 2025!

  • Thumb on the Scale elucidates the role of overparameterization during last layer retraining. We provide optimal weights which depend on the parameterization ratio $\delta=\frac{d}{n}$ and show that this weighting scheme leads to stronger generalization under distribution shift than standard weights.
  • CORAL is a lightweight adaptation of standard diffusion architectures which utilizes a contrastive loss to discourage latent feature entanglement between classes. The discovery of this entanglement was quite surprising and the fix is quite elegant – nice job Esther!

I’m excited to see folks in person in San Diego! Please reach out if you are interested in either work!

You May Also Enjoy

Recognized as a Horejsi Scholar by the ARCS Foundation

less than 1 minute read

Published:

I am excited to announce that I was selected by the ARCS Foundation as a 2025-2026 Horejsi Scholar. This is a fantastic honor and validates the importance of our work in fair and robust machine learning. I am extremely grateful to the ARCS Foundation and Mr. Stewart Horejsi whose generous donation made the award possible.

Paper Accepted to TMLR

less than 1 minute read

Published:

I am happy to announce that my paper entitled “Robustness to Subpopulation Shift with Domain Label Noise via Regularized Annotation of Domains” has been accepted to the Transactions on Machine Learning Research (TMLR). It’s a nice work that introduces a new two-stage last layer model intervention which encourages group-fair predictions without knowledge of the groups.

Papers Accepted to NeurIPS 2024

less than 1 minute read

Published:

I am excited to announce that a paper on which I am first author was accepted to NeurIPS 2024. I am grateful to my collaborators: Rohan Ayyagari (ASU), Monica Welfert (ASU), Sanmi Koyejo (Stanford), Richard Nock (Google), and Lalitha Sankar (ASU). It is a really cool work on making state-of-the-art fairness interventions robust to labels noise. Find the paper here. I am looking forward to meeting with some fantastic (future) collaborators!