Metraipolis › archive › dc_0edf476c2cfe

Loss is Not Behavior: A Unified Output-Space Analysis of Gradient-Based Machine Unlearning — arXiv:2610.38784

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.38784
License (as archived)
CC0-1.0
Archived
2026-11-01T02:24:53.943Z
SHA-256 of the archived text
c35a0541b2678f9aba44918c0623138e2109ed0fac6b897ad12e7e0f2e309b3d
Stable URL of this page
https://metraipolis.com/documento/dc_0edf476c2cfe

How to cite

Loss is Not Behavior: A Unified Output-Space Analysis of Gradient-Based Machine Unlearning — arXiv:2610.38784 — arXiv (en), https://arxiv.org/abs/2610.38784 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0edf476c2cfe (Metraipolis archive dc_0edf476c2cfe, sha256 c35a0541b2678f9aba44918c0623138e2109ed0fac6b897ad12e7e0f2e309b3d)
BibTeX
@misc{dc0edf476c2cfe,
  title = {Loss is Not Behavior: A Unified Output-Space Analysis of Gradient-Based Machine Unlearning — arXiv:2610.38784},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.38784},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0edf476c2cfe; sha256 c35a0541b2678f9aba44918c0623138e2109ed0fac6b897ad12e7e0f2e309b3d},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.38784. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Gradient-based machine unlearning methods typically ``unlearn'' by optimizing forget and retain losses, but similar losses do not necessarily imply similar output-space behavior. Prior work has shown that high forget loss alone does not certify that unlearning has produced retraining-like output behavior. We therefore ask: when a gradient-based unlearner differs from the model that would have been learned on only the retain data, what structure does the resulting output-space gap have? We derive closed-form expressions for this gap in linear regression, with an extension to nonlinear models via a generalized Gauss-Newton and local linearity approximations. Our analysis puts various gradient-based methods under a unified framework parametrized by geometry, reweighting, and projection. Two leverage matrices emerge that capture the forget-retain entanglement under the retrained model geometry: forget self-leverage and retain cross-leverage. We show both theoretically and empirically that these two leverage matrices drive the impact of the forget residual on the forget and retain outputs, and existing methods account for this geometry only partially. Our framework yields precomputable leverage diagnostics for auditing forget sets to identify what type of gradient-based unlearning is likely to be successful.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.38784 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-01T02:12:44.136Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).