Abstract
Recent advances in large language models (LLMs), such as ChatGPT, have led tohighly sophisticated conversation agents. However, these models suffer from"hallucinations," where the model generates false or fabricated information.Addressing this challenge is crucial, particularly with AI-driven platformsbeing adopted across various sectors. In this paper, we propose a novel methodto recognize and flag instances when LLMs perform outside their domainknowledge, and ensuring users receive accurate information. We find that the use of context combined with embedded tags can successfullycombat hallucinations within generative language models. To do this, webaseline hallucination frequency in no-context prompt-response pairs usinggenerated URLs as easily-tested indicators of fabricated data. We observed asignificant reduction in overall hallucination when context was supplied alongwith question prompts for tested generative engines. Lastly, we evaluated howplacing tags within contexts impacted model responses and were able toeliminate hallucinations in responses with 98.88% effectiveness.