Kimia Hamidieh

I am a third-year PhD student at MIT CSAIL, working with Antonio Torralba. My research focuses on data-centric approaches to improving capabilities and reliability of AI models. I am broadly interested in how choices about data–what we pretrain on, how we shape learning targets and feedback signals, and how we evaluate–affect model behavior. I build methods that reliably predict model performance from data and other learning signals, and use these predictions to improve model capabilities. Among other directions, I have previously worked on problems in uncertainty quantification, data-efficient post-training, and training data composition.

During my PhD, I have interned at Microsoft Research New England and Google DeepMind. I previously completed my M.Sc. in Computer Science from the University of Toronto and Vector Institute and B.Sc. in Computer Engineering from Sharif University of Technology.

news

Jul 08, 2026	Our work on data synergy was accepted to COLM 2026!
Mar 19, 2026	Our work on uncertainty quantification was accepted to ICLR 2026 and featured on MIT News.
Jun 01, 2025	Started research internship at Microsoft Research New England working with David Alvarez-Melis on data-centric AI.

selected publications

COLM

Domain-Aware Scaling Laws Uncover Data Synergy

Kimia Hamidieh, Lester Mackey, and David Alvarez-Melis

arXiv preprint arXiv:2607.11052, 2026

arXiv BIB Code

@article{hamidieh2026domain,
  title = {Domain-Aware Scaling Laws Uncover Data Synergy},
  author = {Hamidieh, Kimia and Mackey, Lester and Alvarez-Melis, David},
  journal = {arXiv preprint arXiv:2607.11052},
  year = {2026},
}

ICLR

Complementing Self-Consistency with Cross-Model Disagreement for Uncertainty Quantification

Kimia Hamidieh, Veronika Thost, Walter Gerych, Mikhail Yurochkin, and Marzyeh Ghassemi

In The Fourteenth International Conference on Learning Representations, 2026

arXiv BIB HTML

@inproceedings{hamidieh2026complementing,
  title = {Complementing Self-Consistency with Cross-Model Disagreement for Uncertainty Quantification},
  author = {Hamidieh, Kimia and Thost, Veronika and Gerych, Walter and Yurochkin, Mikhail and Ghassemi, Marzyeh},
  booktitle = {The Fourteenth International Conference on Learning Representations},
  year = {2026},
}

NeurIPS

Improving Subgroup Robustness via Data Selection

Saachi Jain^*, Kimia Hamidieh^*, Kristian Georgiev^*, Andrew Ilyas, Marzyeh Ghassemi, and Aleksander Madry

Advances in Neural Information Processing Systems, 2024

arXiv BIB

@article{jain2024improving,
  title = {Improving Subgroup Robustness via Data Selection},
  author = {Jain, Saachi and Hamidieh, Kimia and Georgiev, Kristian and Ilyas, Andrew and Ghassemi, Marzyeh and Madry, Aleksander},
  journal = {Advances in Neural Information Processing Systems},
  volume = {37},
  pages = {94490--94511},
  year = {2024},
}

ICLR

Views Can Be Deceiving: Improved SSL Through Feature Space Augmentation

Kimia Hamidieh, Haoran Zhang, Swami Sankaranarayanan, and Marzyeh Ghassemi

In The Twelfth International Conference on Learning Representations, 2024

arXiv BIB

@inproceedings{hamidiehviews,
  title = {Views Can Be Deceiving: Improved SSL Through Feature Space Augmentation},
  author = {Hamidieh, Kimia and Zhang, Haoran and Sankaranarayanan, Swami and Ghassemi, Marzyeh},
  booktitle = {The Twelfth International Conference on Learning Representations},
  year = {2024},
}