Rethinking the Evaluation of Pre-trained Text-and-Layout Models from an Entity-Centric Perspective

Benchmark Model Rank Results
entity-linking-on-ec-funsdGeoLayoutLM#2F1: 86.18
entity-linking-on-ec-funsdLayoutLMv3 (large)#5F1: 78.14
entity-linking-on-ec-funsdLayoutLMv3 (base)#8F1: 67.47
semantic-entity-labeling-on-ec-funsdLayoutLMv3 (large)#4F1: 83.88
semantic-entity-labeling-on-ec-funsdGeoLayoutLM#5F1: 83.62
semantic-entity-labeling-on-ec-funsdLayoutLMv3 (base)#8F1: 82.30