Metraipolis › archive › dc_fe60d57c07cc
Preemptive LLM Unlearning against Forbidden Capability Acquisition via Gradient Sealing — arXiv:2609.39866
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.39866
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T13:46:46.105Z
- SHA-256 of the archived text
7161d7eeb088b1f94086ba35f75f618c8c54ce74afa756ded09b3ff498b24266- Stable URL of this page
https://metraipolis.com/documento/dc_fe60d57c07cc
How to cite
Preemptive LLM Unlearning against Forbidden Capability Acquisition via Gradient Sealing — arXiv:2609.39866 — arXiv (en), https://arxiv.org/abs/2609.39866 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fe60d57c07cc (Metraipolis archive dc_fe60d57c07cc, sha256 7161d7eeb088b1f94086ba35f75f618c8c54ce74afa756ded09b3ff498b24266)
BibTeX
@misc{dcfe60d57c07cc,
title = {Preemptive LLM Unlearning against Forbidden Capability Acquisition via Gradient Sealing — arXiv:2609.39866},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.39866},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fe60d57c07cc; sha256 7161d7eeb088b1f94086ba35f75f618c8c54ce74afa756ded09b3ff498b24266},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39866. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Open-weight LLMs are released not only as fixed products but also as substrates for downstream fine-tuning. This openness, however, creates legal and ethical risks because users may misuse fine-tuning to instill illicit knowledge or enable hostile operations. Model providers therefore need apre-release defense against such acquisition, motivating the problem of preemptive unlearning. Unlike retrospective unlearning, which removes capabilities already present in a fixed model, preemptive unlearning seeks to prevent their acquisition under unseen attack data and future fine-tuning procedures. Despite its practical importance, this setting remains largely unexplored, presents distinct challenges, and is therefore the central focus of our work. We first verify that existing retrospective methods provide insufficient pre-release protection. Even when forbidden capabilities are suppressed in current outputs, forbidden-domain data can still induce gradients through internal pathways, enabling later acquisition. Motivated by this finding, we propose a gradient-sealing principle that blocks these pathways by pushing relevant pre-activations into the negative region, where ReLU-family activations exhibit zero or near-zero derivatives. Experiments across multiple LLM families demonstrate our stronger resistance to downstream acquisition than retrospective baselines, validating gradient sealing as an effective mechanism for pre-release protection. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.39866 — 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 2026-10-07T13:18:10.481Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).