Metraipolis › archive › dc_07cee150fae3
GRACE:Gradient-guided Coreset Selection for LLM Unlearning — arXiv:2608.28361
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.28361
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T10:47:36.807Z
- SHA-256 of the archived text
0b45f7a0ffdc9b27cbea084235d2b5e3e59db025f5673eda956ed0694c868cc0- Stable URL of this page
https://metraipolis.com/documento/dc_07cee150fae3
How to cite
GRACE:Gradient-guided Coreset Selection for LLM Unlearning — arXiv:2608.28361 — arXiv (en), https://arxiv.org/abs/2608.28361 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_07cee150fae3 (Metraipolis archive dc_07cee150fae3, sha256 0b45f7a0ffdc9b27cbea084235d2b5e3e59db025f5673eda956ed0694c868cc0)
BibTeX
@misc{dc07cee150fae3,
title = {GRACE:Gradient-guided Coreset Selection for LLM Unlearning — arXiv:2608.28361},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.28361},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_07cee150fae3; sha256 0b45f7a0ffdc9b27cbea084235d2b5e3e59db025f5673eda956ed0694c868cc0},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.28361. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Machine Unlearning methods for Large Language Models typically assume pre-specified forget and retain sets. In realistic settings, however, requests may provide only a few examples of undesired behavior, requiring forget and retain sets to be inferred from heterogeneous corpora. We study this data-selection problem and propose GRACE , a gradient-guided coreset selection method that constructs both forget and retain sets for LLM unlearning. GRACE first computes a forget direction from seed examples that elicit the undesired behavior, then selects a compact forget coreset whose gradients approximate this direction using non-negative orthogonal matching pursuit. To preserve model utility, it selects retain examples after projecting out the forget direction and applying clustered orthogonal matching pursuit in the remaining gradient space. Across two target domains, two model families, and four unlearning algorithms, GRACE improves model utility while maintaining comparable forget quality, with particularly consistent gains over prior gradient-based selection methods. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-31 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2608.28361 — 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 2026-10-17T10:33:42.421Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-31?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).