Metraipolis › archive › dc_ecac72a033f9
ASCENT: First-Order Optimal Fine-Tuning with Recalibration for Safety--Utility Co-Enhancement — arXiv:2610.08061
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.08061
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T11:35:13.681Z
- SHA-256 of the archived text
3fca208fb804145ea8805a5043a11ea10c86d1a162c32019d436c81e54e9fb2a- Stable URL of this page
https://metraipolis.com/documento/dc_ecac72a033f9
How to cite
ASCENT: First-Order Optimal Fine-Tuning with Recalibration for Safety--Utility Co-Enhancement — arXiv:2610.08061 — arXiv (en), https://arxiv.org/abs/2610.08061 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ecac72a033f9 (Metraipolis archive dc_ecac72a033f9, sha256 3fca208fb804145ea8805a5043a11ea10c86d1a162c32019d436c81e54e9fb2a)
BibTeX
@misc{dcecac72a033f9,
title = {ASCENT: First-Order Optimal Fine-Tuning with Recalibration for Safety--Utility Co-Enhancement — arXiv:2610.08061},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.08061},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ecac72a033f9; sha256 3fca208fb804145ea8805a5043a11ea10c86d1a162c32019d436c81e54e9fb2a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.08061. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Supervised fine-tuning can substantially improve the downstream utility of large language models (LLMs) but may compromise their safety. Existing safety-preserving methods constrain downstream updates using safety-related parameters or subspaces, but mainly focus on safety preservation rather than joint safety and utility enhancement, lack a theoretical characterization of the optimal safety-related subspace and safety-preserving task update, and typically rely on a static safety subspace that may become outdated during fine-tuning. To address these limitations, we propose ASCENT, a downstream fine-tuning framework for safety--utility co-enhancement through first-order optimal safety-aware periodic calibration and task optimization. We model safety as a function of LLM parameters $S(\theta)$ and use its first-order approximation to characterize safety changes under parameter updates. Under a fixed rank and Frobenius-norm budget, we prove that the update constructed from the top-$r$ singular components of the safety-function gradient maximizes the estimated safety change, and use it for periodic calibration to preserve and improve safety. We further derive a unique safety-preserving task update that stays close to the original task update while penalizing negative effects on the estimated safety change. ASCENT alternates these optimal task and calibration updates to jointly enhance safety and utility. Experiments across multiple LLM families and downstream tasks show that ASCENT improves downstream utility by up to 20.3\% and reduces attack success rate by up to 35.5\%, achieving state-of-the-art safety and utility across all evaluated settings. Our code is available at this https URL . Tema degli esploratori: fiducia digitale e crittografia (elenco arXiv cs.CR, nuovi lavori del giorno). Categorie arXiv: Cryptography and Security (cs.CR). Pagina del lavoro: https://arxiv.org/abs/2610.08061 — 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 2026-10-07T11:18:39.681Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).