Alleviating Hallucination in Large Vision-Language Models with Active Retrieval Augmentation | Synapse