Hallucination remains a significant challenge in Large Vision-Language Models (LVLMs). Existing training-free methods generally mitigate hallucinations throu…
机构:上交
来源:arXiv 2610.11907 | AI4Papers 论文推荐平台