Publications

Preprints

• 2026
  Large-scale online deanonymization with LLMs

  S. Lermen*, D. Paleka*, J. Swanson, M. Aerni, N. Carlini, F. Tramèr

  blogpdftweetpress [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

Conferences

• ICLR 2026
  Modal Aphasia: Can Unified Multimodal Models Describe Images From Memory?

  M. Aerni*, J. Swanson*, K. Nikolić, F. Tramèr

  blogpdfcodetweet
• ICLR 2025
  Measuring Non-Adversarial Reproduction of Training Data in Large Language Models

  M. Aerni*, J. Rando*, E. Debenedetti, N. Carlini, D. Ippolito, F. Tramèr

  blogpdfcodetweet
• CCS 2024
  Evaluations of Machine Learning Privacy Defenses are Misleading

  M. Aerni*, J. Zhang*, F. Tramèr

  blogpdfcodetweetdoi
• ICLR 2023
  Strong inductive biases provably prevent harmless interpolation

  M. Aerni*, M. Milanta*, K. Donhauser, F. Yang

  pdfcode
• NeurIPS 2021
  Interpolation can hurt robust generalization even when there is no noise

  K. Donhauser*, A. Țifrea*, M. Aerni, R. Heckel, F. Yang

  pdfcodeposter

Workshops

• DLSP 2025
  Membership Inference Attacks on Sequence Models

  L. Rossi, M. Aerni, J. Zhang, F. Tramèr

  pdf
• ICML OPPO 2021
  Surprising benefits of ridge regularization for noiseless regression

  K. Donhauser*, A. Țifrea*, M. Aerni, R. Heckel, F. Yang

  pdfcode
• ICML AML 2021
  Maximizing the robust margin provably overfits on noiseless data

  K. Donhauser*, A. Țifrea*, M. Aerni, R. Heckel, F. Yang

  pdfcode