Beyond Binary Erasure: Soft-Weighted Unlearning for Fairness and Robustness
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (29) · 24936-24944
DOI: 10.1609/aaai.v40i29.39681 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Xinbao Qiao und 3 weiteren Autor:innen.
Abstract (Originalsprache)
Machine unlearning, as a post-hoc processing technique, has gained widespread adoption in addressing challenges like bias mitigation and robustness enhancement. However, existing non-privacy unlearning-based solutions persist in using a binary data removal framework designed for privacy-driven motivation, even when repurposed for fairness or robustness improvements. This leads to significant utility loss, a phenomenon known as “over-unlearning”. While over-unlearning has been largely described in many studies as primarily causing utility degradation, we investigate deeper insights in this work through counterfactual leave-one-out analysis. Based on insights, we introduce a soft weighting strategy that assigns tailored weights to each sample by solving a convex quadratic programming problem analytically, which enables fine-grained model adjustments to address the over-unlearning. We demonstrate that the proposed soft-weighted scheme can be seamlessly integrated into most existing unlearning algorithms. Extensive experiments show that in fairness- and robustness-driven tasks, the soft-weighted scheme significantly outperforms hard-weighted schemes in fairness/robustness metrics and alleviates the decline in utility metric, thereby enhancing unlearning algorithm as an effective correction solution.
Zitieren
APA
Qiao, X., Ding, N., Cheng, Y., & Zhang, M. (2026). Beyond Binary Erasure: Soft-Weighted Unlearning for Fairness and Robustness. Proceedings of the AAAI Conference on Artificial Intelligence, 40(29), 24936-24944. https://doi.org/10.1609/aaai.v40i29.39681
Vancouver
Qiao X, Ding N, Cheng Y, Zhang M. Beyond Binary Erasure: Soft-Weighted Unlearning for Fairness and Robustness. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(29):24936-24944. doi:10.1609/aaai.v40i29.39681
BibTeX
@article{Qiao2026,
author = {Qiao, Xinbao and Ding, Ningning and Cheng, Yushi and Zhang, Meng},
title = {Beyond Binary Erasure: Soft-Weighted Unlearning for Fairness and Robustness},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {29},
pages = {24936--24944},
doi = {10.1609/aaai.v40i29.39681}
}
EUVIMED