The paper introduces TIMECHARA, a benchmark designed to evaluate point-in-time character hallucination in role-playing large language models (LLMs). Point-in-time role-playing, where characters are situated at specific moments in a narrative, enhances user immersion, avoids spoilers, and fosters engagement in fandom role-playing. However, current LLMs often display knowledge that contradicts their characters' identities and historical timelines, a phenomenon known as character hallucination. TIMECHARA, comprising 10,895 instances, reveals significant hallucination issues in state-of-the-art LLMs like GPT-4o. To address this, the authors propose NARRATIVE-EXPERTS, a method that decomposes reasoning steps and utilizes narrative experts to reduce point-in-time character hallucinations. Despite these efforts, the findings highlight ongoing challenges in maintaining spatiotemporal consistency, emphasizing the need for further research. The paper also discusses related work, dataset construction, and experimental results, providing insights into the current state and future directions of point-in-time character hallucination in role-playing LLMs.The paper introduces TIMECHARA, a benchmark designed to evaluate point-in-time character hallucination in role-playing large language models (LLMs). Point-in-time role-playing, where characters are situated at specific moments in a narrative, enhances user immersion, avoids spoilers, and fosters engagement in fandom role-playing. However, current LLMs often display knowledge that contradicts their characters' identities and historical timelines, a phenomenon known as character hallucination. TIMECHARA, comprising 10,895 instances, reveals significant hallucination issues in state-of-the-art LLMs like GPT-4o. To address this, the authors propose NARRATIVE-EXPERTS, a method that decomposes reasoning steps and utilizes narrative experts to reduce point-in-time character hallucinations. Despite these efforts, the findings highlight ongoing challenges in maintaining spatiotemporal consistency, emphasizing the need for further research. The paper also discusses related work, dataset construction, and experimental results, providing insights into the current state and future directions of point-in-time character hallucination in role-playing LLMs.