EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Scalable Retrieval-Augmented Generation for Technical Documents with Hallucination Risk Control

Elizabeth Mitchell, Xiuying Wang

International Journal of Computational and Biological Sciences · 31.12.2025 · 2 (02) · 241-251

DOI: 10.66238/ijcbs29 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Computational and Biological Sciences, online erschienen am 31.12.2025. Verfasst von Elizabeth Mitchell und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The integration of Large Language Models (LLMs) into technical knowledge management systems has revolutionized information accessibility. However, the deployment of Generative AI in critical technical domains—such as aerospace, medical device engineering, and semiconductor manufacturing—is severely hindered by the phenomenon of hallucination, where models generate plausible but factually incorrect information. This paper proposes a novel framework for Scalable Retrieval-Augmented Generation (RAG) specifically tailored for complex technical documentation. We introduce a Hierarchical Contextual Indexing (HCI) mechanism that preserves the structural integrity of technical manuals, coupled with a dual-stage Hallucination Risk Control (HRC) protocol. The HRC protocol utilizes a lightweight Natural Language Inference (NLI) model to perform post-hoc consistency verification between the retrieved context and the generated response. Furthermore, we address the scalability challenges associated with vectorizing millions of technical artifacts by implementing a hybrid sparse-dense retrieval architecture optimized for latency. Our experimental results on a curated dataset of industrial specifications demonstrate that our approach reduces hallucination rates by 43% compared to standard RAG baselines while maintaining sub-second query latency. The findings suggest that enforcing structural context and rigorous consistency checks is essential for the safe deployment of LLMs in high-stakes technical environments.

Zitieren

APA

Mitchell, E., & Wang, X. (2025). Scalable Retrieval-Augmented Generation for Technical Documents with Hallucination Risk Control. International Journal of Computational and Biological Sciences, 2(02), 241-251. https://doi.org/10.66238/ijcbs29

Vancouver

Mitchell E, Wang X. Scalable Retrieval-Augmented Generation for Technical Documents with Hallucination Risk Control. International Journal of Computational and Biological Sciences. 2025;2(02):241-251. doi:10.66238/ijcbs29

BibTeX
@article{Mitchell2025,
  author = {Mitchell, Elizabeth and Wang, Xiuying},
  title = {Scalable Retrieval-Augmented Generation for Technical Documents with Hallucination Risk Control},
  journal = {International Journal of Computational and Biological Sciences},
  year = {2025},
  volume = {2},
  number = {02},
  pages = {241--251},
  doi = {10.66238/ijcbs29}
}
Kostenlos starten →