English(EN) Enhancing Large Multimodal Models in Key Information Extraction via Scene-Aware Document Synthesis

新框架SAYRE合成数据以提升多模态KIE模型

作者 PulseAugur 编辑部 · [2 个来源] · 2026-07-06 03:39

研究人员开发了SAYRE，一个用于合成训练数据以提高大型多模态模型（LMMs）关键信息提取（KIE）能力的新框架。这种场景感知的合成方法从示例文档中生成文档-模式-标注三元组，捕捉内容模式和布局约定。SAYRE还结合了错误驱动生成，以基于真实世界的失败案例创建具有挑战性的训练示例。实验表明，SAYRE显著增强了Qwen3-VL等模型，提高了性能，特别是在设备端LMMs和开放类别提取任务上。 AI

影响增强了多模态模型的数据生成能力，有望改善KIE系统的实际部署。

排序理由该集群描述了一篇研究论文，其中详细介绍了一个用于数据合成以提高AI模型性能的新框架。

在 Hugging Face Daily Papers 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

Hugging Face Daily Papers TIER_1 English(EN) · 2026-07-06 03:39

Enhancing Large Multimodal Models in Key Information Extraction via Scene-Aware Document Synthesis

Key Information Extraction (KIE) converts visually rich documents into structured data, but practical deployment remains challenging: strong performance often relies on costly on-server Large Multimodal Models (LMMs), while compact locally deployable models lack sufficient KIE su…
arXiv cs.CV TIER_1 English(EN) · Zhipeng Xu, Zulong Chen, Qing Liu, Junhao Ji, Jinxin Hu, Yipeng Yu, Jianqiang Wan, Jun Tang, Zhao Li · 2026-07-07 04:00

Enhancing Large Multimodal Models in Key Information Extraction via Scene-Aware Document Synthesis

arXiv:2607.04636v1 Announce Type: new Abstract: Key Information Extraction (KIE) converts visually rich documents into structured data, but practical deployment remains challenging: strong performance often relies on costly on-server Large Multimodal Models (LMMs), while compact …

报道来源 [2]

Enhancing Large Multimodal Models in Key Information Extraction via Scene-Aware Document Synthesis

Enhancing Large Multimodal Models in Key Information Extraction via Scene-Aware Document Synthesis

相关实体

相关话题