Reasoning with Latent Structure Refinement for Document-Level Relation Extraction

Benchmark Model Rank Results
relation-extraction-on-gdaLSR w/o MDP Nodes#6F1: 82.2
relation-extraction-on-cdrLSR w/o MDP Nodes#7F1: 64.8
relation-extraction-on-docredLSR+BERT-base#30F1: 59.05Ign F1: 56.97
relation-extraction-on-docredLSR+GloVe#44F1: 54.18Ign F1: 52.15