Visual interventions provide a training-free way to reduce hallucinations in multimodal large language models. However, an intervention may have little lever…
机构:西湖大学
来源:arXiv 2610.12981 | AI4Papers 论文推荐平台