I am a research scientist at the Helmholtz Institute for Human-Centered AI, working with Dr. Eric Schulz.
My research is focused on using large language models (LLMs) for scientific hypothesis generation.
I obtained my PhD at UC Berkeley (2024) and my undergraduate degree at Brown University (2018).

Large Language Models (LLMs)

I am broadly interested in LLMs' ability to reason through data sets of human behavior, and generate plausible hypotheses with regards to underlying processes that explain the data.

Published work:

Rmus, M., Jagadish, A.K., Mathony, M., Ludwig, T., & Schulz, E. (2025). Generating Computational Cognitive Models using Large Language Models PDF
Binz, M., Jagadish, A.K., Rmus, M., & Schulz, E. (2025). Automated scientific minimization of regret PDF

Computational cognitive modeling

My PhD research focused on developing computational models that translate cognitive theories into concise algorithms, with parameters that map onto cognitive mechanisms impacting observable behavior. This work involves quantifying individual differences in human information processing through model parameters, pinpointing the most suitable theories embedded within various cognitive models, and investigating the interactions among cognitive processes like executive function, working memory, and reinforcement learning.

Published work:

Rmus, M., Pan, T., Xia, L., & Collins, A. G. E. (2024). Artificial neural networks for model identification and parameter estimation in computational cognitive models. PLOS Computational Biology PDF
Rmus, M., He, M., Baribault, B., Walsh, E. G., Festa, E. K., Collins, A. G. E. & Nassar, M. R. (2022). Age-related differences in prefrontal glutamate are associated with increased working memory decay that gives the appearance of learning deficits. eLife. PDF
Rmus, M., Zou, A. & Collins, A. G. E. (2023). Choice type impacts human reinforcement learning. Journal of Cognitive Neuroscience. PDF
Rmus, M., McDougle, S. D. & Collins, A. G. E. (2021). The role of executive function in shaping reinforcement learning. Current Opinion in Behavioral Sciences. PDF
Rmus, M., Ritz, H., Hunter, L. E., Bornstein, A. M. & Shenhav, A. (2022). Humans can navigate complex graph structures acquired during latent learning. Cognition. PDF

Model fitting tools

Cognitive science researchers heavily rely on tools to connect computational cognitive models with data, often through likelihood estimation. However, numerous models do not have tractable likelihood functions, thus limiting the range of testable theories. My work has also focused on implementing an alternative model-fitting approach using light-weight artificial neural networks that enable parameter estimation and model identification solely based on data sequences - circumventing the need for computing likelihood.

Published work:

Rmus, M., Pan, T., Xia, L. & Collins, A. G. E. (2024). Artificial neural networks for model identification and parameter estimation in computational cognitive models. PLOS Computational Biology PDF
Rmus, M., Xia, J., Collins, J. & Collins, A. G. E. (2022). Using Deep Learning tools for fitting Reinforcement Learning Models. Conference on Computational Cognitive Neuroscience Proceedings. PDF

Resume for more information.

Get in touch!


Email: milena_rmus@berkeley.edu
Twitter: @milenamr7
Bluesky: @milenamr7bsky.social
My GitHub page is here.

Tattooing

I am learning how to tattoo in my free time. Here are a few of my practice tattoos: