Multi-Object Hallucination in Vision Language Models