Luca Eyring

I am currently a Student Researcher at Google DeepMind with Paul Vicol working on video generation, and an ELLIS PhD student in the EML Group at the Technical University of Munich under the supervision of Zeynep Akata and Alexey Dosovitskiy (Recursive, Inceptive). As part of my ELLIS exchange, I had the pleasure to spend time at Inceptive working with Alexey and Jakob Uszkoreit. I am broadly interested in generative modeling, reward-alignment/post-training, representation learning, and optimal transport. My research is generously supported by the Google PhD Fellowship 2025 in ML Foundations.

Previously, I pursued a Bachelor in Informatics at LMU Munich and a Master specializing in Machine Learning at Technical University of Munich. I completed my Master’s thesis on unbalanced neural optimal transport for single-cell trajectory inference at Helmholtz Munich, supervised by Niki Kilbertus and Fabian Theis.

Selected Publications

2026

ICML

Diamond Maps: Efficient Reward Alignment via Stochastic Flow Maps

Peter Holderrieth*, Douglas Chen*, Luca Eyring*, and 7 more authors

In International Conference on Machine Learning (ICML), 2026

Best Paper Award @ ReALM–GEN ICLR 2026

Bib PDF Code

@inproceedings{holderrieth2026diamondmaps,
  author = {Holderrieth, Peter and Chen, Douglas and Eyring, Luca and Shah, Ishin and Anantharaman, Giri and He, Yutong and Akata, Zeynep and Jaakkola, Tommi and Boffi, Nicholas Matthew and Simchowitz, Max},
  title = {Diamond Maps: Efficient Reward Alignment via Stochastic Flow Maps},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2026},
}

ICML

Beyond the Final Layer: Attentive Multilayer Fusion for Vision Transformers

Laure Ciernik, Marco Morik, Lukas Thede, and 4 more authors

In International Conference on Machine Learning (ICML), 2026

And in Sci4DL Workshop @ ICLR 2026

Bib PDF Code

@inproceedings{ciernik202attentivemultilayer,
  author = {Ciernik, Laure and Morik, Marco and Thede, Lukas and Eyring, Luca and Nakajima, Shinichi and Akata, Zeynep and Muttenthaler, Lukas},
  title = {Beyond the Final Layer: Attentive Multilayer Fusion for Vision Transformers},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2026},
}

2025

NeurIPS

Noise Hypernetworks: Amortizing Test-Time Compute in Diffusion Models

Luca Eyring, Shyamgopal Karthik, Alexey Dosovitskiy, and 2 more authors

In Neural Information Processing Systems (NeurIPS), 2025

Bib PDF Demo Code

@inproceedings{eyring2025hypernoise,
  author = {Eyring, Luca and Karthik, Shyamgopal and Dosovitskiy, Alexey and Ruiz, Nataniel and Akata, Zeynep},
  title = {Noise Hypernetworks: Amortizing Test-Time Compute in Diffusion Models},
  booktitle = {Neural Information Processing Systems (NeurIPS)},
  year = {2025},
  url = {https://openreview.net/forum?id=DbzREoPwmM},
}

ICLR

Disentangled Representation Learning with the Gromov-Monge Gap

Théo Uscidda*, Luca Eyring*, Karsten Roth, and 3 more authors

In International Conference on Learning Representations (ICLR), 2025

And in SPIGM Workshop @ ICML 2024

Bib PDF

@inproceedings{uscidda2025gromovmonge,
  title = {Disentangled Representation Learning with the Gromov-Monge Gap},
  author = {Uscidda, Théo and Eyring, Luca and Roth, Karsten and Theis, Fabian and Akata, Zeynep and Cuturi, Marco},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2025},
  url = {https://openreview.net/forum?id=ehr4oTe6XI},
}

2024

NeurIPS

ReNO: Enhancing One-step Text-to-Image Models through Reward-based Noise Optimization

Luca Eyring*, Shyamgopal Karthik*, Karsten Roth, and 2 more authors

In Neural Information Processing Systems (NeurIPS), 2024

Bib PDF Demo Code

@inproceedings{eyring2024reno,
  title = {ReNO: Enhancing One-step Text-to-Image Models through Reward-based Noise Optimization},
  author = {Eyring, Luca and Karthik, Shyamgopal and Roth, Karsten and Dosovitskiy, Alexey and Akata, Zeynep},
  booktitle = {Neural Information Processing Systems (NeurIPS)},
  year = {2024},
  url = {https://openreview.net/forum?id=MXY0qsGgeO},
}

ICLR

Unbalancedness in Neural Monge Maps Improves Unpaired Domain Translation

Luca Eyring*, Dominik Klein*, Théo Uscidda*, and 4 more authors

In International Conference on Learning Representations (ICLR), 2024

Bib PDF Talk Code

@inproceedings{eyring2024unbalancedness,
  title = {Unbalancedness in Neural Monge Maps Improves Unpaired Domain Translation},
  author = {Eyring, Luca and Klein, Dominik and Uscidda, Th{\'e}o and Palla, Giovanni and Kilbertus, Niki and Akata, Zeynep and Theis, Fabian J},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2024},
  url = {https://openreview.net/forum?id=2UnCj3jeao},
}