Multimodal large language models (MLLMs) are increasingly used as OCR systems in document and knowledge-processing pipelines, but their ability to faithfully…
机构:华东师范大学
来源:arXiv 2608.18586 | AI4Papers 论文推荐平台