AI Paper Atlas

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

Unifying Vision, Text, and Layout for Universal Document Processing

Zineng Tang, Ziyi Yang, Guoxin Wang, Yuwei Fang, Yang Liu, Chenguang Zhu, Michael Zeng, Cha Zhang, Mohit Bansal

CVPR 2023 · Vision-language and multimodal › Human and embodied modeling · approach: Layout-aware document understanding · 119 citations (OpenAlex)

Read the paper

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 the proposed self-supervised pretraining objectives instead of MLM alone (ablation)UDOP · DocVQA validationscore: 79.7 $\pm$ 0.4 -> 84.4 $\pm$ 0.2Improved
Add the proposed self-supervised pretraining objectives instead of MLM alone (ablation)UDOP · RVL-CDIP validationscore: 95.3 $\pm$ 0.3 -> 96.2 $\pm$ 0.2Improved
Add Visual Text Recognition to Joint Text-Layout pretraining (ablation)UDOP · DocVQA validationscore: 82.8 $\pm$ 0.1 -> 83.3 $\pm$ 0.2Improved
Add Layout Modeling after Visual Text Recognition (ablation)UDOP · DocVQA validationscore: 83.3 $\pm$ 0.2 -> 84.0 $\pm$ 0.3Improved
Add Layout Modeling after Visual Text Recognition (ablation)UDOP · RVL-CDIP validationscore: 95.4 $\pm$ 0.2 -> 95.6 $\pm$ 0.2Improved
Add Masked Image Reconstruction to text-layout objectives (ablation)UDOP · RVL-CDIP validationscore: 95.6 $\pm$ 0.2 -> 96.2 $\pm$ 0.2Improved
Add supervised pretraining data after self-supervised objectives (ablation)UDOP · DocVQA validationscore: 84.4 $\pm$ 0.2 -> 85.0$\pm$ 0.2Improved
Add supervised pretraining data after self-supervised objectives (ablation)UDOP · RVL-CDIP validationscore: 96.2 $\pm$ 0.2 -> 96.3 $\pm$ 0.1Improved
Use a unified encoder instead of separate modality-specific encoders (ablation)UDOP · DUE Benchmarkaverage score: 64.7$\pm$0.3 -> 65.1$\pm$0.2Improved
Use a unified encoder instead of separate modality-specific encoders (ablation)UDOP · WTQscore: 46.7$\pm$0.4 -> 47.2$\pm$0.2Hurt
Remove image input embeddings (ablation)UDOP · DocVQAscore: 84.7 -> 84.4Hurt
Remove image input embeddings (ablation)UDOP · InfoVQAscore: 47.4 -> 45.0Hurt

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