Hubert Strauss

hs6702[at]princeton.edu

Hi! I am a Research Engineer at Princeton University, within the Princeton Language & Intelligence Initiative.

My current work focuses on language model post-training. I’m interested in understanding how training choices shape learning across SFT, RL, and distillation. I also like to spend time on systems and infra for more efficient training and generation.

I previously worked at o9Solutions (1.5 years, within the ML team), and at Dematic (internship then part-time, within the Operations Research and Applied AI team).

I obtained my Engineering degree (BSc./MSc.) at Arts et Metiers ParisTech and studied Operations Research (MSc.) at Georgia Tech.

publications

^* denotes equal contribution

When Errors Can Be Beneficial: A Categorization of Imperfect Rewards for Policy Gradient

Shuning Shang^*, Hubert Strauss^*, Stanley Wei, Sanjeev Arora, and Noam Razin

arXiv preprint arXiv:2604.25872, 2026

Bib PDF Code

@article{shang2026when,
  title = {When Errors Can Be Beneficial: A Categorization of Imperfect Rewards for Policy Gradient},
  author = {Shang, Shuning and Strauss, Hubert and Wei, Stanley and Arora, Sanjeev and Razin, Noam},
  journal = {arXiv preprint arXiv:2604.25872},
  year = {2026}
}

Hardware-Efficient Attention for Fast Decoding

Ted Zadouri, Hubert Strauss, and Tri Dao

In Conference on Language Modeling (COLM 2025), 2025

Bib PDF Code

@inproceedings{zadouri2025hardwareefficientattentionfastdecoding,
  title = {Hardware-Efficient Attention for Fast Decoding},
  author = {Zadouri, Ted and Strauss, Hubert and Dao, Tri},
  booktitle = {Conference on Language Modeling},
  year = {2025},
}

What Makes a Reward Model a Good Teacher? An Optimization Perspective

Noam Razin, Zixuan Wang, Hubert Strauss, Stanley Wei, Jason D Lee, and Sanjeev Arora

In Advances in Neural Information Processing Systems (NeurIPS 2025), 2025

Bib PDF Code Poster

@inproceedings{razin2025what,
  title = {What Makes a Reward Model a Good Teacher? An Optimization Perspective},
  author = {Razin, Noam and Wang, Zixuan and Strauss, Hubert and Wei, Stanley and Lee, Jason D and Arora, Sanjeev},
  booktitle = {Advances in Neural Information Processing Systems},
  year = {2025},
}

FutureFill: Fast Generation from Convolutional Sequence Models

Naman Agarwal, Xinyi Chen, Evan Dogariu, Devan Shah, Hubert Strauss, Vlad Feinberg, Daniel Suo, Peter Bartlett, and Elad Hazan

In International Conference on Learning Representations (ICLR 2026), 2024

Bib PDF

@inproceedings{agarwal2025futurefillfastgenerationconvolutional,
  title = {FutureFill: Fast Generation from Convolutional Sequence Models},
  author = {Agarwal, Naman and Chen, Xinyi and Dogariu, Evan and Shah, Devan and Strauss, Hubert and Feinberg, Vlad and Suo, Daniel and Bartlett, Peter and Hazan, Elad},
  booktitle = {International Conference on Learning Representations},
  year = {2024},
}