Metraipolis › archive › dc_333fd884a33b
RADNPO: Reference-free Adaptive Negative Preference Optimization for LLM Unlearning — arXiv:2609.34251
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.34251
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T22:34:47.538Z
- SHA-256 of the archived text
f3b8c711b1c6e16bd51464deca46f9c0b4ad281e63f91ea511ed5ea8aaaad1dc- Stable URL of this page
https://metraipolis.com/documento/dc_333fd884a33b
How to cite
RADNPO: Reference-free Adaptive Negative Preference Optimization for LLM Unlearning — arXiv:2609.34251 — arXiv (en), https://arxiv.org/abs/2609.34251 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_333fd884a33b (Metraipolis archive dc_333fd884a33b, sha256 f3b8c711b1c6e16bd51464deca46f9c0b4ad281e63f91ea511ed5ea8aaaad1dc)
BibTeX
@misc{dc333fd884a33b,
title = {RADNPO: Reference-free Adaptive Negative Preference Optimization for LLM Unlearning — arXiv:2609.34251},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.34251},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_333fd884a33b; sha256 f3b8c711b1c6e16bd51464deca46f9c0b4ad281e63f91ea511ed5ea8aaaad1dc},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34251. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language models (LLMs) can memorize sensitive, private, or copyrighted content during pre-training, making machine unlearning necessary for removing targeted knowledge. Recent preference optimization (PO)-based unlearning methods improve stability over gradient ascent (GA)-based methods by introducing alignment-style objectives, which effectively suppress the probability of forget targets. However, target suppression alone does not sufficiently constrain the next-token distribution after unlearning. Existing methods provide limited control over how suppressed probability mass is redistributed and insufficiently adapt forgetting strength to target confidence and distributional concentration. Even after target suppression, probability mass may remain concentrated on a few non-target tokens, potentially producing repetitive or uninformative outputs. To address these limitations, we propose Reference-free ADaptive Negative Preference Optimization (RADNPO), which explicitly guides next-token probability redistribution. Specifically, RADNPO contrasts each forget target with alternative tokens favored by the current next-token distribution and adaptively modulates token-level forgetting strength using target confidence and next-token concentration. Experiments on TOFU and MUSE demonstrate that RADNPO achieves a better trade-off between forgetting quality and model utility than current baselines. Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CR). Categorie arXiv: Cryptography and Security (cs.CR). Pagina del lavoro: https://arxiv.org/abs/2609.34251 — 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 2026-10-07T22:13:03.717Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).