EUVIMEDEUVIMED entdecken
Crossref

Unveiling AI Safety in Fine-tuning Quantized Model

Hai Le

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (48) · 41492-41494

DOI: 10.1609/aaai.v40i48.42317 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Hai Le.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Post-training quantization is widely used to compress large language models (LLMs) for efficient deployment in resource-constrained environments. However, recent work shows that quantization, especially aggressive schemes such as 4-bit QLoRA, can substantially degrade safety alignment, making models more vulnerable to harmful completions and jailbreaks. In this work, we investigate these safety risks and propose a mitigation strategy: projecting quantized parameters back into safety-aligned subspaces. First, we empirically measure safety degradation on benchmark datasets using both safety and utility metrics. Next, we explore projection-based restoration methods to recover alignment-preserving directions in the LoRA adapters of quantized models. Finally, we study how quantization affects mechanistic safety neurons and how hybrid-precision designs can preserve them. By foregrounding the safety implications of model compression, this work aims to support more robust, deployment-ready, and ethically aligned LLMs.

Zitieren

APA

Le, H. (2026). Unveiling AI Safety in Fine-tuning Quantized Model. Proceedings of the AAAI Conference on Artificial Intelligence, 40(48), 41492-41494. https://doi.org/10.1609/aaai.v40i48.42317

Vancouver

Le H. Unveiling AI Safety in Fine-tuning Quantized Model. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(48):41492-41494. doi:10.1609/aaai.v40i48.42317

BibTeX
@article{Le2026,
  author = {Le, Hai},
  title = {Unveiling AI Safety in Fine-tuning Quantized Model},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {48},
  pages = {41492--41494},
  doi = {10.1609/aaai.v40i48.42317}
}
Kostenlos starten →