cs.LGMay 8, 2026

Disagreement-Regularized Importance Sampling for Adversarial Label Corruption

Authors: Csongor HorváthIda-Maria SintornPrashant Singh

Organizations: Department of Information Technology, Uppsala University, Uppsala, Sweden · Science for Life Laboratory, Uppsala University, Uppsala, Sweden

Abstract

Standard Importance Sampling (IS) collapses under label corruption because high-norm examples, prioritized for variance reduction, are often adversarial outliers. We formalize this misalignment using an ε\varepsilon-contamination model and propose Disagreement-Regularized Importance Sampling (DR-IS), a sub-sampling method based on loss rank-disagreement across independent proxy ensemble. We prove finite-sample concentration bounds showing that the empirical rank disagreement of bulk corrupted examples is bounded above, and that of boundary-clean examples bounded below, both at rate O(log(N/δ)/K)O(\sqrt{\log(N/δ)/K}) with probability 1δ1-δ; when the structural expectation gap ΔΔ' between the two groups is positive and the boundary-clean set is at least as large as the selected subset, these bounds certify strict separation and control the contamination rate of the selected subset. Empirically, DR-IS remains robust under targeted high-norm attacks that break magnitude-based methods such as the Error L2L_2-norm (EL2N) on benchmark datasets. DR-IS complements training-dynamics approaches like Area Under the Margin ranking (AUM), offering improved robustness in the loss-aligned regime alongside explicit finite-sample concentration certificates and a contamination bound limiting noise leakage from the statistical tail of corrupted points.

Explore similar work

CardsList
  1. Learning with Monotone Adversarial Corruptions

    Jan 5, 2026Kasper Green Larsen, Chirag Pabbaraju, Abhishek ShettyEmpirical Risk MinimizationGround Truth

  2. Label Differential Privacy via Aggregation

    Date pendingAnand Brahmbhatt, Rishi Saket, Shreyas Havaldar +3Differential PrivacyCandidate Label Distributions