Analysis of the Impact of Information Retrieval Methods on Reducing Hallucinations in Large Language Models
Analysis of the Impact of Information Retrieval Methods on Reducing Hallucinations in Large Language Models
fahime bagheri1
1) M.Sc. in Computer Engineering, Specialization in Artificial Intelligence and Robotics Shahrood University of Technology, Iran
IOI Code
:
ACCA-ACZGBD
Publication :
4th International Congress on Science, engineering & New Technologies(secongress.com)
Abstract :
One of the fundamental challenges in deploying Large Language Models (LLMs) in sensitive scientific environments is the occurrence of information hallucinations, which can lead to the dissemination of incorrect data. This paper presents a systematic analysis of various Information Retrieval (IR) methods developed to ground model responses and mitigate hallucinations. The study examines diverse approaches—including traditional text-based retrieval, semantic search, and Retrieval-Augmented Generation (RAG) frameworks—evaluating the strengths and weaknesses of each when dealing with complex scientific texts. The analysis indicates that combining multi-stage retrieval methods with re-ranking techniques constitutes the most effective strategy for ensuring the scientific accuracy of responses. Finally, by proposing a conceptual model, this paper outlines future directions for achieving highly reliable language models for scientific applications.
Keywords: Large Language Models, Hallucination, Information Retrieval, Systematic Analysis, Scientific Accuracy.
Keywords : large language models hallucination information retrieval systematic analysis scientific accuracy