{"id":"arxiv/2604.08625","url":"https://arxiv.org/abs/2604.08625","tags":["stat.ML","cs.LG","math.ST"],"score":0,"title":"A Transport Geometry of Benign Overfitting","domain":"stat.ML","source":"arxiv","authors":["Olaf Yunus Laitinen Imanov","Rana Irem Turhan","Isgandar Jafarli"],"reviews":0,"subject":"stat.ML","summary":"Benign overfitting, the phenomenon that interpolating noisy data can be statistically harmless, is usually characterized through effective-rank conditions on the covariance spectrum. These conditions are fragile: a perturbation of vanishing relative operator and Frobenius norm can turn a benign covariance into one whose minimum-norm interpolator has non-vanishing excess risk. We argue that the natural geometry of benign overfitting is optimal transport of the log-eigenvalues, measured in units of the sample size $n$. In log-spectral coordinates, the deterministic equivalent of the risk…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-04-09","reactions":{}}