The Surprising Harmfulness of Benign Overfitting for Adversarial Robustness. (arXiv:2401.12236v2 [cs.LG] UPDATED)

Recent empirical and theoretical studies have established the generalization
capabilities of large machine learning models that are trained to
(approximately or exactly) fit noisy data. In this work, we prove a surprising
result that even if the ground truth itself is robust to adversarial examples,
and the benignly overfitted model is benign in terms of the ``standard''
out-of-sample risk objective, this benign overfitting process can be harmful
when out-of-sample data are subject to adversarial manipulation. More
specifically, our main results …

adversarial arxiv capabilities data examples large machine machine learning machine learning models noisy prove result robustness studies truth work