Id: arxiv/2607.09547 Url: https://arxiv.org/abs/2607.09547 Tags: math.ST, math.PR, stat.ML Score: 0 Title: High-Dimensional Interpolators Can Be Fragile: Heavy Tails and High-Dimensional Large Deviations Domain: math.ST Source: arxiv Authors: Youheng Zhu, Yiping Lu Reviews: 0 Subject: math.ST Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-07-10 Page: https://delphi.science/r/arxiv/2607.09547 High-dimensional interpolation is common in modern machine learning, but its tail risk is less understood than its expected prediction risk. Existing theory shows that interpolating models can perform well in expectation, yet such guarantees do not determine the probability of rare, severe errors. In operations research and stochastic decision-making applications, rare estimation errors can have disproportionate downstream effects, so tail behavior matters alongside average performance. We study the fragility of high-dimensional linear interpolators using large-deviation methods. We focus on…