Mitigating LLM Hallucination in the Banking Domain

Abstract

Large Language Models (LLMs) offer significant potential in the banking sector, particularly for applications such as fraud detection, credit approval, and enhancing customer experience. However, their tendency to "hallucinate"—generating plausible but inaccurate information—poses a critical challenge. This thesis examines existing strategies for mitigating LLM hallucinations and proposes a novel approach to reduce hallucinations in the context of predicting customer churn using LLMs.