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

Optimization of Adaptive Prompt Engineering for Large Language Models via Bayesian Inference in Low-Resource Settings

Arthur J. Williams, Katherine L. McDough, Thomas S. Halloway

Frontiers in Artificial Intelligence Research · 30.04.2026 · 3 (1) · 46-53

DOI: 10.71465/fair601 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Artificial Intelligence Research, online erschienen am 30.04.2026. Verfasst von Arthur J. Williams und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid proliferation of Large Language Models (LLMs) has necessitated the development of effective prompt engineering strategies to harness their full potential. However, the stochastic nature of LLM outputs and the vast, discrete combinatorial space of natural language make manual prompt design a laborious and often suboptimal process. While automated prompt optimization techniques exist, they typically require significant computational resources, massive datasets, or access to model gradients, rendering them inaccessible for low-resource environments and edge computing applications. This paper proposes a novel framework for optimizing adaptive prompt engineering using Bayesian Inference. We introduce a sample-efficient methodology that treats prompt selection as a black-box optimization problem, utilizing Gaussian Processes to model the latent manifold of prompt effectiveness. By strategically balancing exploration and exploitation through acquisition functions, our approach converges on high-performing prompts with significantly fewer model queries than traditional grid search or reinforcement learning paradigms. We demonstrate that this method achieves state-of-the-art performance on reasoning benchmarks while reducing token consumption by an order of magnitude, making advanced prompt engineering feasible for academic laboratories and limited-budget applications.

Zitieren

APA

Williams, A. J., McDough, K. L., & Halloway, T. S. (2026). Optimization of Adaptive Prompt Engineering for Large Language Models via Bayesian Inference in Low-Resource Settings. Frontiers in Artificial Intelligence Research, 3(1), 46-53. https://doi.org/10.71465/fair601

Vancouver

Williams AJ, McDough KL, Halloway TS. Optimization of Adaptive Prompt Engineering for Large Language Models via Bayesian Inference in Low-Resource Settings. Frontiers in Artificial Intelligence Research. 2026;3(1):46-53. doi:10.71465/fair601

BibTeX
@article{Williams2026,
  author = {Williams, Arthur J. and McDough, Katherine L. and Halloway, Thomas S.},
  title = {Optimization of Adaptive Prompt Engineering for Large Language Models via Bayesian Inference in Low-Resource Settings},
  journal = {Frontiers in Artificial Intelligence Research},
  year = {2026},
  volume = {3},
  number = {1},
  pages = {46--53},
  doi = {10.71465/fair601}
}
Kostenlos starten →