ChainMPQ: Interleaved Text-Image Reasoning Chains for Mitigating Relation Hallucinations

Yike Wu, Yiwei Wang, Yujun Cai

While Large Vision-Language Models (LVLMs) achieve strong performance in multimodal tasks, hallucinations continue to hinder their reliability. Among the three categories of halluc…