Metraipolis › archive › dc_f2fa546c1fd6

What to Forget in Unlearning? Forget Set Curation for Language Models — arXiv:2608.14855

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.14855
License (as archived)
CC0-1.0
Archived
2026-10-18T18:56:36.143Z
SHA-256 of the archived text
9c02bb970bad28f4f5d6eca60b40acfacff3d970a47b5fda4b7b2963b9a07430
Stable URL of this page
https://metraipolis.com/documento/dc_f2fa546c1fd6

How to cite

What to Forget in Unlearning? Forget Set Curation for Language Models — arXiv:2608.14855 — arXiv (en), https://arxiv.org/abs/2608.14855 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f2fa546c1fd6 (Metraipolis archive dc_f2fa546c1fd6, sha256 9c02bb970bad28f4f5d6eca60b40acfacff3d970a47b5fda4b7b2963b9a07430)
BibTeX
@misc{dcf2fa546c1fd6,
  title = {What to Forget in Unlearning? Forget Set Curation for Language Models — arXiv:2608.14855},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.14855},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f2fa546c1fd6; sha256 9c02bb970bad28f4f5d6eca60b40acfacff3d970a47b5fda4b7b2963b9a07430},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.14855. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Machine unlearning aims to remove targeted data or behaviors from a trained model without retraining from scratch. Yet most evaluations assume that the examples to forget are already known. In realistic language-model deployments, a requester may ask a model to stop reproducing a song or book without knowing which spans, documents, quotations, or near-duplicates in a trillion-token corpus support that behavior. We study this missing upstream problem, forget set curation: mapping a suppression request to the data passed to an unlearning algorithm. We introduce CleanSlate, a benchmark for verbatim output suppression over songs and books, with model-specific extraction profiles, content-grounded QA, and capability-retention evaluations. CleanSlate exposes two failure modes. Natural lexical and exact-substring curators often yield forget sets that lead to weak suppression. An evaluation-aware curator suppresses requested continuations almost completely, but causes collateral regression on non-requested content and model-dependent capability loss. These results show that practical unlearning is not only an optimization problem once a forget set is given: the data chosen for forgetting determines both what can be unlearnt and what else is damaged.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.14855 — 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 2026-10-18T18:06:57.831Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).