A Systematic Review of Approaches Used in Hallucination Detection Among Large Learning Models

Abstract

Generative AI (artificial intelligence) is a leading technology that is multiple sectors like healthcare, finance, education and retail sector. The growth of these sectors is entirely dependent on implementation or adoption of AI based technologies and this adoption will help to make a strong hold a market. LLMs (Large Language Models) like GPT, XLM-R and mBERT have shown tremendous growth in natural language processing field or era. These models are prevalent due to several benefits like solving the hard problem in seconds and answering the user queries in fraction of seconds, despite having this prevalence in different fields these models have a significant reliability issue called as hallucination, this issue arises they produce information that is fluent and convincing in nature but is irrelevant, incorrect, or completely made up of false sentences and facts. In this systematic literature review a subjective analysis is followed to analyze different hallucination techniques used to improve the performance of LLMs by reducing the hallucination. For this purpose, a secondary data collection is done in this analysis that includes the research journals published in recent years. From this broader scan of literature and analysis it is identified that RAG is the best approach to detect the hallucination among the different LLms