ML Research Wiki / Benchmarks / Relation Extraction / FUNSD

FUNSD

Relation Extraction Benchmark

Performance Over Time

📊 Showing 9 results | 📏 Metric: F1

Top Performing Models

Rank Model Paper F1 Date Code
1 LayoutLMv3 large EM + BBO + RSF A LayoutLMv3-Based Model for Enhanced Relation Extraction in Visually-Rich Documents 90.81 2024-04-16 -
2 GeoLayoutLM GeoLayoutLM: Geometric Pre-training for Visual Information Extraction 89.45 2023-04-21 📦 alibabaresearch/advancedliteratemachinery
3 RORE (GeoLayoutLM) Modeling Layout Reading Order as Ordering Relations for Visually-rich Document Understanding 88.46 2024-09-29 📦 chongzhangFDU/ROOR
4 LayoutLMv3 large LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking 80.35 2022-04-18 📦 huggingface/transformers 📦 microsoft/unilm 📦 pwc-1/Paper-9 📦 MindSpore-scientific-2/code-14
5 LayoutLMv3 large GeoLayoutLM: Geometric Pre-training for Visual Information Extraction 80.35 2023-04-21 📦 alibabaresearch/advancedliteratemachinery
6 TPP (LayoutMask) Reading Order Matters: Information Extraction from Visually-rich Documents by Token Path Prediction 79.20 2023-10-17 📦 chongzhangfdu/tpp 📦 WinterShiver/Token-Path-Prediction
7 BROS BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents 77.01 2021-08-10 📦 clovaai/bros 📦 2024-MindSpore-1/Code2
8 LayoutLMv2 large LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding 70.57 2020-12-29 📦 huggingface/transformers 📦 PaddlePaddle/PaddleOCR 📦 microsoft/unilm
9 LayoutLM LayoutLM: Pre-training of Text and Layout for Document Image Understanding 42.83 2019-12-31 📦 huggingface/transformers 📦 PaddlePaddle/PaddleOCR 📦 microsoft/unilm

All Papers (9)

alibabaresearch/advancedliteratemachinery
chongzhangFDU/ROOR
huggingface/transformers microsoft/unilm
alibabaresearch/advancedliteratemachinery
chongzhangfdu/tpp WinterShiver/Token-Path-Prediction
clovaai/bros 2024-MindSpore-1/Code2
huggingface/transformers PaddlePaddle/PaddleOCR
huggingface/transformers PaddlePaddle/PaddleOCR