{"ID":6023403,"CreatedAt":"2026-07-08T01:00:23.257252134Z","UpdatedAt":"2026-07-10T06:38:11.380144103Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.05898","arxiv_id":"2607.05898","title":"Auditing of Unlearning Algorithms","abstract":"Evaluating whether unlearning algorithms truly remove training data influence remains an open challenge. We propose a practical auditor that computes data-dependent lower bounds on the unlearning parameter $\\varepsilon$ using membership inference attacks. Evaluating multiple unlearning algorithms, we find a sharp separation: algorithms with rigorous guarantees, such as model clipping and rewind-to-delete, achieve very small $\\varepsilon$ bounds that do not falsify their unlearning guarantees, whereas empirical methods such as Hessian-based unlearning, interleaved ascent-descent, ascent on the forget set, and fine-tuning on the retain set exhibit large bounds, indicating poor unlearning. Our auditor provides a practical tool for empirically falsifying unlearning claims through a hypothesis-testing framework, and we validate it on CIFAR-100 and Shakespeare text.","short_abstract":"Evaluating whether unlearning algorithms truly remove training data influence remains an open challenge. We propose a practical auditor that computes data-dependent lower bounds on the unlearning parameter $\\varepsilon$ using membership inference attacks. Evaluating multiple unlearning algorithms, we find a sharp separ...","url_abs":"https://arxiv.org/abs/2607.05898","url_pdf":"https://arxiv.org/pdf/2607.05898v1","authors":"[\"Sahasrajit Sarmasarkar\",\"Anastasia Koloskova\",\"Sanmi Koyejo\"]","published":"2026-07-07T06:51:55Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.CR\"]","methods":"[]","has_code":false}