THE STUDY OF LARGE LANGUAGE MODELS FOR KNOWLEDGE OF THE SPECIFICS OF RAILWAY TRANSPORT

Abstract

Large LLM (Large Language Model) language models have made significant progress in the field of natural language processing in recent years. These models, using advanced neural network techniques and huge amounts of training data, have demonstrated capabilities in understanding and creating human-like text. Language models have great potential for solving urgent tasks of searching for information at the user's request. Large language models can make significant changes in the educational process, adapting to the current level of user training.

Conducted studies on the use of ChatGPT, also known as Chat Generative Pre-Trained Transformer, to solve the tests of the US Medical license exam (USMLE) [16], which confirmed the ability of ChatGPT to cross the threshold for obtaining a license, as well as conducting a number of experiments on the use of various LLMs to solve the Unified State Exam, confirm the high potential of using large language models for solving problems that were previously solved with the required level of quality exclusively with the involvement of human resources. 

 The article presents the results of research aimed at studying the knowledge of the specifics of railway transport using various language models. Four available language models have been tested using railway-related tests.

Based on the results obtained, the following conclusion is formulated: information in the field of railway transport in these models is presented in a limited volume that is not sufficient to build an intelligent system that helps the user study railway transport or be used as a decision support system in the transport industry. In this regard, the task of creating large language models based on the materials of the transport industry is relevant.