Entity Extraction on FUNSD (test)
93.2Entity F1 ScoreLayoutMask Large
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LayoutMask LargeParameters=404M, Modality=T+L2023.05 | 93.2 | — | — | |
| ERNIE-Layout_largeModel Size=Large2022.10 | 93.12 | — | — | |
| ERNIE-Layout LargeModality=T+L+I2023.05 | 93.12 | — | — | |
| LayoutMask BaseParameters=182M, Modality=T+L2023.05 | 92.91 | — | — | |
| LayoutLMv3_LARGEParameters=368M + α, Pretrain Data Size=11M2023.03 | 92.08 | — | — | |
| LayoutLMv3 LargeParameters=368M, Modality=T+L+I2023.05 | 92.08 | — | — | |
| ESPParameters=50M, Pretrain Data Size=0.9M2023.03 | 91.12 | — | — | |
| ESPmultilingual model=false2023.03 | 91.1 | — | — | |
| LayoutLMv3_BASEParameters=133M + α, Pretrain Data Size=11M2023.03 | 90.29 | — | — | |
| LayoutLMv3 BaseParameters=133M, Modality=T+L+I2023.05 | 90.29 | — | — | |
| LayoutLM_BASEParameters=160M + α, Pretrain Data Size=11M2023.03 | 88.41 | — | — | |
| LiLT BaseModality=T+L2023.05 | 88.41 | — | — | |
| UDocParameters=272M + α, Pretrain Data Size=11M2023.03 | 87.93 | — | — | |
| LiLT_BASEPretrain Data Size=11M2023.03 | 85.74 | — | — | |
| StructuralLM_LARGEModality=Text + Layout + Cell*, # Params=355M2021.08 | 85.14 | 83.52 | 86.81 | |
| StructuralLM_largeModel Size=Large2022.10 | 85.14 | — | — | |
| FormNetImage=false, #Params=217M, Pre-training Size=0.7M (9GB)2022.03 | 84.69 | 85.21 | 84.18 | |
| DocFormerImage=true, #Params=536M, Pre-training Size=5M2022.03 | 84.55 | 82.29 | 86.94 | |
| DocFormer_largeModel Size=Large2022.10 | 84.55 | — | — | |
| DocFormer_LARGEParameters=536M + α, Pretrain Data Size=5M2023.03 | 84.55 | — | — | |
| DocFormer LargeParameters=536M, Modality=T+L+I2023.05 | 84.55 | — | — | |
| BROS_LARGEModality=Text + Layout, # Params=340M2021.08 | 84.52 | 82.81 | 86.31 | |
| BROS LARGEVisual Features=false, Number of Parameters=340M2021.08 | 84.52 | — | — | |
| BROS_LARGEText block order=With order, Modality=Text + Layout2021.08 | 84.52 | — | — | |
| BROS_LARGEParameters=340M + α, Pretrain Data Size=11M2023.03 | 84.52 | — | — | |
| BROS LargeParameters=340M, Modality=T+L2023.05 | 84.52 | — | — | |
| LayoutLMv2Model size=LARGE2020.12 | 84.2 | — | — | |
| LayoutLMv2Image=true, #Params=426M, Pre-training Size=11M2022.03 | 84.2 | 83.24 | 85.19 | |
| LayoutLMv2_LARGEModality=Text + Layout + Image*, # Params=426M2021.08 | 84.2 | 83.24 | 85.19 | |
| LayoutLMv2 LARGEVisual Features=true, Number of Parameters=426M2021.08 | 84.2 | — | — | |
| LayoutLMv2_largeModel Size=Large2022.10 | 84.2 | — | — | |
| LayoutLMv2_LARGEParameters=426M + α, Pretrain Data Size=11M2023.03 | 84.2 | — | — | |
| LiLTbackbone=InfoXLM2023.03 | 84.2 | — | — | |
| LayoutLMv2 LargeParameters=426M, Modality=T+L+I2023.05 | 84.2 | — | — | |
| LayoutLMv2_LARGEText block order=With order, Modality=Text + Layout + Image2021.08 | 83.59 | — | — | |
| SelfDocModality=Text + Layout + Image*, # Params=137M2021.08 | 83.36 | — | — | |
| SelfDocPretrain Data Size=0.32M2023.03 | 83.36 | — | — | |
| XYLayout_BASEParameters=345M + α, Pretrain Data Size=30M2023.03 | 83.35 | — | — | |
| DocFormer_BASEModality=Text + Layout + Image*, # Params=183M2021.08 | 83.34 | 80.76 | 86.09 | |
| DocFormer BaseParameters=183M, Modality=T+L+I2023.05 | 83.34 | — | — | |
| DocFormerImage=false, #Params=502M, Pre-training Size=5M2022.03 | 83.33 | 81.33 | 85.44 | |
| StrucTexTModality=Text + Layout + Image*, # Params=107M+2021.08 | 83.09 | 85.68 | 80.97 | |
| StrucTextParameters=107M + α, Pretrain Data Size=0.9M2023.03 | 83.09 | — | — | |
| BROS_BASEModality=Text + Layout, # Params=110M2021.08 | 83.05 | 81.16 | 85.02 | |
| BROS BASEVisual Features=false, Number of Parameters=110M2021.08 | 83.05 | — | — | |
| BROS_BASEText block order=With order, Modality=Text + Layout2021.08 | 83.05 | — | — | |
| BROS_BASEParameters=110M + α, Pretrain Data Size=11M2023.03 | 83.05 | — | — | |
| BROS BaseParameters=110M, Modality=T+L2023.05 | 83.05 | — | — | |
| LayoutLMv2Model size=BASE2020.12 | 82.76 | — | — | |
| LayoutLMv2_BASEModality=Text + Layout + Image*, # Params=200M2021.08 | 82.76 | 80.29 | 85.39 | |
| LayoutLMv2 BASEVisual Features=true, Number of Parameters=200M2021.08 | 82.76 | — | — | |
| LayoutLMv2_BASEParameters=200M + α, Pretrain Data Size=11M2023.03 | 82.76 | — | — | |
| LayoutLMv2 BaseParameters=200M, Modality=T+L+I2023.05 | 82.76 | — | — | |
| LayoutLMv2_BASEText block order=With order, Modality=Text + Layout + Image2021.08 | 81.89 | — | — | |
| BROS2020.12 | 81.21 | — | — | |
| DocFormer_BASEModality=Text + Layout, # Params=149M2021.08 | 80.54 | 77.63 | 83.69 | |
| LayoutXLMlanguage-specific fine-tuning=true2023.03 | 79.4 | — | — | |
| LayoutLMv1Image=true, #Params=160M, Pre-training Size=11M2022.03 | 79.27 | 76.77 | 81.95 | |
| LayoutLM_BASEModality=Text + Layout + Image*, # Params=160M2021.08 | 79.27 | 76.77 | 81.95 | |
| LayoutLM_LARGEText block order=With order, Modality=Text + Layout2021.08 | 79.27 | — | — | |
| LayoutLM BaseParameters=160M, Modality=T+L+I2023.05 | 79.27 | — | — | |
| LayoutLMModel size=LARGE2020.12 | 78.95 | — | — | |
| LayoutLM_LARGEModality=Text + Layout, # Params=343M2021.08 | 78.95 | 75.96 | 82.19 | |
| LayoutLM LARGEVisual Features=false, Number of Parameters=343M2021.08 | 78.95 | — | — | |
| LayoutLM_largeModel Size=Large2022.10 | 78.95 | — | — | |
| LayoutLMModel size=BASE2020.12 | 78.66 | — | — | |
| LayoutLM_BASEModality=Text + Layout, # Params=113M2021.08 | 78.66 | 75.97 | 81.55 | |
| LayoutLM BASEVisual Features=false, Number of Parameters=113M2021.08 | 78.66 | — | — | |
| LayoutLM_BASEText block order=With order, Modality=Text + Layout2021.08 | 78.54 | — | — | |
| LayoutLMv1Image=false, #Params=343M, Pre-training Size=11M2022.03 | 77.89 | 75.36 | 80.61 | |
| LayoutLM_LARGEParameters=343M + α, Pretrain Data Size=11M2023.03 | 77.89 | — | — | |
| LayoutLM LargeParameters=343M, Modality=T+L2023.05 | 77.89 | — | — | |
| BROS# Data=10, Epochs=100, Batch size=42021.08 | 72.6 | — | — | |
| UniLMv2Model size=LARGE2020.12 | 72.57 | — | — | |
| UniLMv2_largeModel Size=Large2022.10 | 72.57 | — | — | |
| UniLMv2 LargeParameters=355M, Modality=T2023.05 | 72.57 | — | — | |
| UniLMv2Image=false, #Params=355M, Pre-training Size=160GB2022.03 | 70.72 | 67.8 | 73.91 | |
| RoBERTa_largeModel Size=Large2022.10 | 70.72 | — | — | |
| RoBERTa_LARGEParameters=355M + α2023.03 | 70.72 | — | — | |
| ROBERTA LargeParameters=355M, Modality=T2023.05 | 70.72 | — | — | |
| SPADEImage=false, #Params=110M, Pre-training Size=BERT-multilingual2022.03 | 70.5 | — | — | |
| LayoutLMv2*# Data=10, Epochs=100, Batch size=42021.08 | 69.92 | — | — | |
| UniLMv2Model size=BASE2020.12 | 68.9 | — | — | |
| UniLMv2 BaseParameters=125M, Modality=T2023.05 | 68.9 | — | — | |
| BROS# Data=5, Epochs=100, Batch size=42021.08 | 68.35 | — | — | |
| ROBERTa2023.03 | 66.7 | — | — | |
| RoBERTa_BASEParameters=125M + α2023.03 | 66.48 | — | — | |
| ROBERTA BaseParameters=125M, Modality=T2023.05 | 66.48 | — | — | |
| BERTModel size=LARGE2020.12 | 65.63 | — | — | |
| BERT_LARGEModality=Text, # Params=340M2021.08 | 65.63 | 61.13 | 70.85 | |
| BERT_largeModel Size=Large2022.10 | 65.63 | — | — | |
| BERT_LARGEParameters=340M + α2023.03 | 65.63 | — | — | |
| BERT LargeParameters=340M, Modality=T2023.05 | 65.63 | — | — | |
| LayoutLMv2*# Data=5, Epochs=100, Batch size=42021.08 | 64.26 | — | — | |
| BERT_LARGEText block order=With order, Modality=Text2021.08 | 64.17 | — | — | |
| LayoutLM*# Data=10, Epochs=100, Batch size=42021.08 | 62.5 | — | — | |
| BERT_BASEText block order=With order, Modality=Text2021.08 | 60.92 | — | — | |
| BERT_BASEParameters=110M + α2023.03 | 60.92 | — | — | |
| BERTModel size=BASE2020.12 | 60.26 | — | — | |
| BERT_BASEModality=Text, # Params=110M2021.08 | 60.26 | 54.69 | 67.1 |