In Search of Forgotten Domain Generalization

Prasanna Mayilvahanan*
University of Tübingen, MPI-IS, Tübingen AI Center
Roland Zimmermann*
University of Tübingen, MPI-IS, Tübingen AI Center
Thaddäus Wiedemer
MPI-IS, University of Tübingen, Tübingen AI Center
Evgenia Rusak
University of Tübingen, MPI-IS, TÜbingen AI Center
Attila Juhos
University of Tübingen, MPI-IS, TÜbingen AI Center
Matthias Bethge
University of Tübingen, Tübingen AI Center
Wieland Brendel
MPI-IS, ELLIS Institute Tübingen, Tübingen AI Center

tl;dr: CLIP's high performance on style-centric domain shifts is significantly influenced by the presence of such images in its training set.

News

Feb '25 Our paper was accepted at ICLR 2025 as a spotlight!
Oct '24 The pre-print is now available on arXiv.

Abstract

Acknowledgements & Funding

BibTeX

If you find our study helpful, please cite our paper:

@inproceedings{mayilvahanan2025clipdg,
  title={In Search of Forgotten Domain Generalization},
  author={Prasanna Mayilvahanan and Roland Zimmermann and Thaddäus Wiedemer and Evgenia Rusak and Attila Juhos and Matthias Bethge and Wieland Brendel},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025},
  url={https://openreview.net/forum?id=Fk3eod9aaD}
}
Webpage designed using Bootstrap 4.5. Layout courtesy of Roland Zimmermann.