Researchers have proposed a new framework to reduce hallucinations in large language models (LLMs). Hallucinations occur when LLMs make incorrect predictions based on incomplete or missing information. The new approach, called causal prompt engineering, involves creating a mental model of an expert's decision-making process and using it as input for the LLM. This allows the LLM to reason more accurately and reduce hallucination rates. The framework has been tested in various domains, including grant proposal evaluation, cybersecurity design, and clinical diagnosis. --- Why it matters: This matters because LLMs are increasingly being used in high-stakes applications where accuracy is crucial. Reducing hallucinations can improve the reliability of these models and prevent costly mistakes. Source: https://arxiv.org/abs/2509.10818