近日，字节跳动与中国科学技术大学共同研究的多模态文档大模型DocPedia取得重大突破，成功刷新了现有模型分辨率极限。DocPedia的分辨率达到了2560×2560，相较于之前的先进模型有显著提升。这款模型不仅能精准识别图像信息，还能根据用户需求调用知识库回答问题，展现了高分辨率多模态文档理解的强大能力。
作为一款创新性的多模态文档大模型，DocPedia的研制成功标志着我国在人工智能领域取得了重要进展。字节跳动与中科大的联合研究，旨在推动深度学习技术的发展，为用户提供更智能、更高效的服务。
英文翻译：
Title:ByteDance and USTC Collaborate to Develop High-Resolution Multimodal Large Model
Keywords: ByteDance, USTC, Multimodal Document Large Model, DocPedia, Resolution Limit
News Content:
Recently, ByteDance and USTC jointly developed a significant breakthrough in the multimodal document large model DocPedia. The model successfully surpassed the resolution limit of existing models, reaching 2560×2560, significantly improving upon advanced models. DocPedia not only accurately recognizes image information but also answers questions based on user demands by calling upon the knowledge base, demonstrating the powerful capabilities of high-resolution multimodal document understanding.
As an innovative multimodal document large model, the successful development of DocPedia marks an important advancement in the field of artificial intelligence in China. The collaboration between ByteDance and USTC aims to promote the development of deep learning technology and provide users with smarter and more efficient services.
【来源】https://www.qbitai.com/2023/12/103190.html
Views: 1