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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add Augmented XY Cut and DCPE (ablation) | LayoutXLM · XFUN ZH SER | F1: 0.8924 -> 0.9176 | Improved |
| Add Augmented XY Cut and DCPE (ablation) | LayoutXLM · XFUN ES SER | F1: 0.7550 -> 0.7687 | Improved |
| Add Augmented XY Cut and DCPE (ablation) | LayoutXLM · XFUN ZH RE | F1: 0.7073 -> 0.7445 | Improved |
| Add Augmented XY Cut and DCPE (ablation) | LayoutXLM · XFUN ES RE | F1: 0.6896 -> 0.7259 | Improved |
| Replace the position-embedding MLP with CPE without correcting reading order (ablation) | LayoutXLM · XFUN SER and RE | F1: 0.8924 -> 0.8776; 0.7550 -> 0.7306; 0.7073 -> 0.7059; 0.6896 -> 0.6812 | Hurt |
| Add standard XY Cut reading order (ablation) | LayoutXLM · XFUN ZH and ES RE | F1: 0.7073 -> 0.7281; 0.6896 -> 0.7175 | Improved |
| Use Augmented XY Cut rather than standard XY Cut (ablation) | LayoutXLM · XFUN ZH and ES SER | F1: 0.8903 -> 0.9023; 0.7562 -> 0.7570 | Improved |
| Remove position embeddings (ablation) | LayoutXLM · XFUN ZH SER | F1: 0.8924 -> 0.8842 | Hurt |
| Sort tokens by X then Y instead of the default reading order (ablation) | LayoutXLM · XFUN ZH SER | F1: 0.8924 -> 0.8561 | Hurt |
| Use Augmented XY Cut thresholds (0.5,0.5,10) rather than (0.5,0.5,5) (ablation) | LayoutXLM · XFUN ZH SER | F1: 0.9023 -> 0.8702 | Hurt |
| Use dilated rather than standard convolutions for both text and image DCPE (ablation) | XYLayoutLM · XFUN ZH SER | F1: 0.9037 -> 0.9176 | Improved |
| Use standard 2D convolution for text instead of 1D convolution in DCPE (ablation) | XYLayoutLM · XFUN ZH SER | F1: 0.9140 -> 0.9037 | Hurt |
Closest papers by meaning
- DocLayLLM: An Efficient Multi-modal Extension of Large Language Models for Text-rich Document Understanding (CVPR 2025 · similarity 0.89)
- A Simple yet Effective Layout Token in Large Language Models for Document Understanding (CVPR 2025 · similarity 0.88)
- LayoutLLM: Layout Instruction Tuning with Large Language Models for Document Understanding (CVPR 2024 · similarity 0.88)
- Relation-Rich Visual Document Generator for Visual Information Extraction (CVPR 2025 · similarity 0.87)
- MosaicDoc: A Large-Scale Bilingual Benchmark for Visually Rich Document Understanding (AAAI 2026 · similarity 0.87)
- DocKylin: A Large Multimodal Model for Visual Document Understanding with Efficient Visual Slimming (AAAI 2025 · similarity 0.86)
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.
