The concept of integrating large language models and regularizing bayesian approach

Problem statement. The creation of transparent and explicable artificial intelligence systems is a prerequisite for their widespread use in industry. The task of developing approaches to assess the quality of solutions, ensure the required level of accuracy and stability, and take into account risks when implementing intelligent systems is particularly relevant.

Goal. To consider the possibility of integrating large language models (LLM) and the methodology of the regularizing Bayesian approach (RBP) to create a new class of applied artificial intelligence systems with transparency, interpretability and practical usefulness for enterprises.

Results. The theoretical concepts of LLM and RBP integration are presented. It is shown that the regularizing Bayesian approach provides tools for assessing the quality of decisions made, ensuring their accuracy and stability, as well as forming a unified approach to creating interpreted intelligent data processing systems. The possibilities of using RBP to increase the explainability and reliability of systems based on large language models are outlined.

Practical significance. The integration of LLM and RBP can contribute to the creation of applied intelligent systems more suitable for industrial use due to consistency, the ability to accurately assess risks, increase confidence in the results of models, as well as compliance with the principles and standards of metrology. Such systems can be used by enterprises for more reliable and informed decision support.