Metraipolis › archive › dc_cf83d9ee8867
Balancing Privacy, Utility, and Safety in LLM Alignment through Preference Optimization — arXiv:2608.30141
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.30141
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T08:25:46.302Z
- SHA-256 of the archived text
cc8f8d2439cecbd7eb37ff0d99cf731cb4c0870dbab6b4c75b4e96b7963c1b14- Stable URL of this page
https://metraipolis.com/documento/dc_cf83d9ee8867
How to cite
Balancing Privacy, Utility, and Safety in LLM Alignment through Preference Optimization — arXiv:2608.30141 — arXiv (en), https://arxiv.org/abs/2608.30141 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cf83d9ee8867 (Metraipolis archive dc_cf83d9ee8867, sha256 cc8f8d2439cecbd7eb37ff0d99cf731cb4c0870dbab6b4c75b4e96b7963c1b14)
BibTeX
@misc{dccf83d9ee8867,
title = {Balancing Privacy, Utility, and Safety in LLM Alignment through Preference Optimization — arXiv:2608.30141},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.30141},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cf83d9ee8867; sha256 cc8f8d2439cecbd7eb37ff0d99cf731cb4c0870dbab6b4c75b4e96b7963c1b14},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.30141. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Preference optimization is widely used to align large language models with human preferences, but preference-data composition may also influence privacy-relevant memorization. We examine whether adding synthetic privacy-preference pairs to Direct Preference Optimization (DPO) is associated with lower canary-based memorization signals without modifying the objective or introducing a formal privacy mechanism. We propose Privacy-Pressure Preference Mixing (P3M), a data-composition protocol that varies the amount of privacy-preference data while keeping helpfulness and harmlessness preference data fixed. We evaluate a non-privacy Baseline and privacy-mixing ratios of 0.5, 1.0, and 2.0 using Gemma 3 270M-IT across five random seeds and validate the same four conditions using 4-bit-quantized Gemma 2 2B-IT across three seeds. Overall, under the tested conditions, privacy-preference mixing is associated with lower mean canary suffix log-likelihood proxy values across both model settings and lower aggregate membership-inference attack performance relative to the Baseline in the mixed-source 2B evaluation. Specifically, across the privacy-aware 2B configurations, the mean area under the receiver operating characteristic curve (AUROC) ranges from 0.596 to 0.629, and the mean area under the precision-recall curve (AUPRC) ranges from 0.541 to 0.575, compared with 0.804 and 0.790, respectively, for the Baseline. However, the reduction in membership distinguishability does not hold uniformly across data sources. Moreover, the relationship between the privacy ratio and harmlessness preference accuracy varies by model setting, whereas helpfulness preference accuracy remains broadly stable. These findings suggest that P3M should be viewed as a lightweight empirical protocol for examining privacy-utility-safety trade-offs rather than as a formal privacy guarantee or a defense against extraction attacks. Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-09-01 in cs.CR). Categorie arXiv: Cryptography and Security (cs.CR) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2608.30141 — 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 2026-10-17T08:10:59.883Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).