Hallucinations in LLMs: Strategies for monitoring
Sevval Gölbasi
Hallucinations in large language models (LLMs) refer to instances where the model generates information that is either incorrect, incomplete, or unfaithful to the input. These hallucinations can take many forms, from giving wrong answers to leaving out important details. Additionally, when evaluating model performance, it’s important to consider other quality metrics, such as the tone of voice, levels of toxicity, bias, or even hatefulness.
