What’s in Common? Multimodal Models Hallucinate When Reasoning Across Scenes | Read Paper on Bytez