Benedikt Seiter

I am a Master’s student in Mathematics at ETH Zurich. My research interests include domain generalization, causality, robust representation learning, and mechanistic interpretability.

My BSc thesis, supervised by Prof. Jonas Peters, resulted in Anchor PCA, a method for invariant dimensionality reduction across multiple environments. The paper has been accepted at NeurIPS 2026.

Contacts: bseiter@ethz.ch Google Scholar GitHub LinkedIn

News

Sep, 2026	🎉 Anchor PCA has been accepted at NeurIPS 2026! I will present the poster on Wednesday, December 9, 2026, 12:30–14:30 CET, at the Palais des Congrès de Paris.
Jun, 2026	Jonas briefly introduced Anchor PCA’s objective and one result during his Florence Nightingale Lecture at Oxford. The Anchor PCA segment begins at 39:23 in the recording.
Jun, 2026	Our paper Anchor PCA is now available on arXiv! We introduce a method for robust dimensionality reduction across multiple environments.

Selected Publications

NeurIPS ’26
Anchor PCA

Benedikt Seiter, Anya Fries, Julius von Kügelgen, and 1 more author

In Advances in Neural Information Processing Systems (NeurIPS), 2026

Camera-ready version in preparation.

Abs arXiv Bib PDF Code

Principal component analysis (PCA) is one of the most widely used unsupervised dimension reduction techniques. We study PCA for data from multiple related domains. Since principal components generally differ across domains, one way to obtain a shared low-rank embedding is to perform PCA on the pooled data. However, this approach can focus on spurious directions that exhibit high variation in only a few domains. To find a robust embedding that still explains most variance in unseen but similar domains, we propose instead to focus on shared directions of variation. To this end, we introduce Anchor PCA which trades off overall explained variance with agreement between the shared and domain-specific low-rank embeddings. Anchor PCA amounts to PCA on a modified target matrix and thus can be solved efficiently. Moreover, we show that Anchor PCA recovers a maximal invariant subspace and admits a minimax reconstruction interpretation under bounded domain-specific covariance inflations. On simulated and real-world data with temporal distribution shift, we demonstrate, respectively, that Anchor PCA recovers the maximal invariant subspace and yields embeddings that explain more variance on unseen domains than the pooling baseline and a worst-case alternative. Taken together, these findings establish Anchor PCA as a promising approach to robust unsupervised dimension reduction from multi-domain data.
@inproceedings{seiter2026anchorpca, title = {Anchor PCA}, author = {Seiter, Benedikt and Fries, Anya and {von Kügelgen}, Julius and Peters, Jonas}, year = {2026}, archiveprefix = {arXiv}, primaryclass = {stat.ML}, booktitle = {Advances in Neural Information Processing Systems (NeurIPS)}, note = {Camera-ready version in preparation.}, }

Anchor PCA in 113 seconds:

A short visual introduction to Anchor PCA.