Publications

Michael Aerni, Javier Rando, Edoardo Debenedetti, Nicholas Carlini, Daphne Ippolito, Florian Tramèr (2025). Measuring Non-Adversarial Reproduction of Training Data in Large Language Models. In ICLR 2025.

PDF Cite Code Blog post Twitter

Michael Aerni, Jie Zhang, Florian Tramèr (2024). Evaluations of Machine Learning Privacy Defenses are Misleading. In CCS 2024.

PDF Cite Code Blog post Twitter ACM DOI

Michael Aerni, Marco Milanta, Konstantin Donhauser, Fanny Yang (2022). Strong inductive biases provably prevent harmless interpolation. In ICLR 2023.

PDF Cite Code

Konstantin Donhauser, Alexandru Țifrea, Michael Aerni, Reinhard Heckel, Fanny Yang (2021). Interpolation can hurt robust generalization even when there is no noise. In NeurIPS 2021.

PDF Cite Code Poster

Konstantin Donhauser, Alexandru Țifrea, Michael Aerni, Reinhard Heckel, Fanny Yang (2021). Surprising benefits of ridge regularization for noiseless regression. In ICML 2021 Workshop OPPO.

PDF Cite Code

Konstantin Donhauser, Alexandru Țifrea, Michael Aerni, Reinhard Heckel, Fanny Yang (2021). Maximizing the robust margin provably overfits on noiseless data. In ICML 2021 Workshop AML.

PDF Cite Code