List your company
ResearchFormerly WizWand

Table Fact Verification on TabFact 1.0 (test)

65.3AccuracyLPA-Ranking w/ Discriminator (Caption)
49.80453.82757.8561.873Sep 5, 2019
Updated 8mo ago

Evaluation results

MethodLinks
LPA-Ranking w/ Discriminator (Caption)Input=Caption as sequence2019.0965.3
Table-BERT-Horizontal-F+T-TemplateLinearization scan direction=Horizontal, Sequence order=Fact then Table, Linearization strategy=Template2019.0965.1
Table-BERT-Horizontal-T+F-TemplateLinearization scan direction=Horizontal, Sequence order=Table then Fact, Linearization strategy=Template2019.0965.1
LPA-Ranking w/ DiscriminatorDiscriminator usage=Ranking discriminator2019.0965
NSM w/ LPA-guided ML + RLLearning strategy=LPA-guided ML + RL2019.0963.5
LPA-Weighted-VotingDiscriminator usage=Confidence weights2019.0963.1
LPA-Voting w/o DiscriminatorDiscriminator usage=None2019.0958.2
Table-BERT-Vertical-T+F-TemplateLinearization scan direction=Vertical, Sequence order=Table then Fact, Linearization strategy=Template2019.0957
Table-BERT-Vertical-F+T-TemplateLinearization scan direction=Vertical, Sequence order=Fact then Table, Linearization strategy=Template2019.0956.2
NSM w/ RL (Binary Reward)Learning strategy=Reinforcement Learning2019.0954.1
BERT classifier w/o TableInput=Fact only2019.0950.5
Table-BERT-Horizontal-F+T-ConcatenateLinearization scan direction=Horizontal, Sequence order=Fact then Table, Linearization strategy=Concatenate2019.0950.4