AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

XYLayoutLM: Towards Layout-Aware Multimodal Networks for Visually-Rich Document Understanding

Zhangxuan Gu, Changhua Meng, Ke Wang, Jun Lan, Weiqiang Wang, Ming Gu, Liqing Zhang

CVPR 2022 · Vision-language and multimodal › Human and embodied modeling · approach: Layout-aware document understanding · 84 citations (OpenAlex)

Check your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add Augmented XY Cut and DCPE (ablation)LayoutXLM · XFUN ZH SERF1: 0.8924 -> 0.9176Improved
Add Augmented XY Cut and DCPE (ablation)LayoutXLM · XFUN ES SERF1: 0.7550 -> 0.7687Improved
Add Augmented XY Cut and DCPE (ablation)LayoutXLM · XFUN ZH REF1: 0.7073 -> 0.7445Improved
Add Augmented XY Cut and DCPE (ablation)LayoutXLM · XFUN ES REF1: 0.6896 -> 0.7259Improved
Replace the position-embedding MLP with CPE without correcting reading order (ablation)LayoutXLM · XFUN SER and REF1: 0.8924 -> 0.8776; 0.7550 -> 0.7306; 0.7073 -> 0.7059; 0.6896 -> 0.6812Hurt
Add standard XY Cut reading order (ablation)LayoutXLM · XFUN ZH and ES REF1: 0.7073 -> 0.7281; 0.6896 -> 0.7175Improved
Use Augmented XY Cut rather than standard XY Cut (ablation)LayoutXLM · XFUN ZH and ES SERF1: 0.8903 -> 0.9023; 0.7562 -> 0.7570Improved
Remove position embeddings (ablation)LayoutXLM · XFUN ZH SERF1: 0.8924 -> 0.8842Hurt
Sort tokens by X then Y instead of the default reading order (ablation)LayoutXLM · XFUN ZH SERF1: 0.8924 -> 0.8561Hurt
Use Augmented XY Cut thresholds (0.5,0.5,10) rather than (0.5,0.5,5) (ablation)LayoutXLM · XFUN ZH SERF1: 0.9023 -> 0.8702Hurt
Use dilated rather than standard convolutions for both text and image DCPE (ablation)XYLayoutLM · XFUN ZH SERF1: 0.9037 -> 0.9176Improved
Use standard 2D convolution for text instead of 1D convolution in DCPE (ablation)XYLayoutLM · XFUN ZH SERF1: 0.9140 -> 0.9037Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy