ART: Attention Replacement Technique to Improve Factuality in LLMs

ArXi:2604.06393v1 Announce Type: new Hallucination in large language models (LLMs) continues to be a significant issue, particularly in tasks like question answering, where models often generate plausible yet incorrect or irrelevant information. Although various methods have been proposed to mitigate hallucinations, the relationship between attention patterns and hallucinations has not been fully explored.