Trustworthy Deep Learning Theory: Private Over-Parameterized Models and Robust LLMs

Project Period: 2024-10-01 – 2025-09-14

Funder: Google Research

Principal Investigator

Department(s)

Mondelli Group

Funder

Google Research

Funder Schema

Google-PhPF

Funder Registry

2026 | Published | Thesis | PhD | IST-REx-ID: 22857 |

Trustworthy machine learning in high dimensions
S. Bombari, Trustworthy Machine Learning in High Dimensions, Institute of Science and Technology Austria, 2026.

[Published Version] View | Files available | DOI

2026 | Published | Conference Paper | IST-REx-ID: 22894 |

A law of data reconstruction for random features (and beyond)
L. Iurada, S. Bombari, T. Tommasi, M. Mondelli, in:, 14th International Conference on Learning Representations, OpenReview, 2026, pp. 145275–145314.

[Published Version] View | Files available | arXiv

2025 | Published | Journal Article | IST-REx-ID: 19627 |

Privacy for free in the overparameterized regime
S. Bombari, M. Mondelli, Proceedings of the National Academy of Sciences 122 (2025).

2025 | Published | Conference Paper | IST-REx-ID: 21324 |

Spurious correlations in high dimensional regression: The roles of regularization, simplicity bias and over-parameterization
S. Bombari, M. Mondelli, in:, Proceedings of the 42nd International Conference on Machine Learning, ML Research Press, 2025, pp. 4839–4873.

[Published Version] View | Files available | arXiv