Large language models can generate fabricated content that is not grounded in either the provided context or their pre-training data. There are two types of hallucination, but this post focuses on 'extrinsic' hallucinations where the model's output should be verifiable by external world knowledge. To avoid these hallucinations, LLMs need to produce factual responses and acknowledge when they don't know an answer. --- Why it matters: Understanding extrinsic hallucinations is crucial for developing reliable AI systems that can provide accurate information without spreading misinformation. Improving the ability of LLMs to distinguish between known and unknown facts will have a significant impact on their trustworthiness in applications such as customer service chatbots, language translation tools, and expert knowledge retrieval systems. Source: https://lilianweng.github.io/posts/2024-07-07-hallucination/