List your company
ResearchFormerly WizWand

Natural Language Inference on MultiNLI mismatched (test)

81.4AccuracyLM-Transformer
63.82468.38772.9577.513Jun 1, 2017Aug 17, 2017Nov 2, 2017Jan 18, 2018Apr 5, 2018Jun 21, 2018Sep 7, 2018
Updated 8mo ago

Evaluation results

MethodLinks
LM-Transformerensemble=false2018.0581.4
DRCN+ELMoensemble=true, embedding=ELMo2018.0581.4
DRCNensemble=true2018.0579.5
CAFEensemble=true2018.0579
DIINExtra SNLI training set=true2017.1178.8
DIINEnsemble=true2017.0978.7
DIINensemble=true2018.0578.7
DRCNensemble=false2018.0578.4
CAFEensemble=false2018.0577.9
CAFE2017.1177.9
DIINEnsemble=false2017.0977.8
DIINensemble=false2018.0577.8
KIM2017.1176.4
ESIM2017.1175.8
Gated-Att BiLSTMEnsemble=true2017.0974.9
BiLSTM with generalized pooling# Params=18M2018.0974
ESIM GloVeAuxiliary Information=GloVe vectors2017.0673.92
2-layer Bi-CAS-LSTM# Params=6.8M2018.0973.7
Deep Gated Attn. BiLSTM encodersSNLI Mix=false, Parameter count=11.6m2017.1273.6
Gated-Att BiLSTMEnsemble=false2017.0973.6
Shortcut-Stacked encoderEnsemble=false2017.0973.6
Deep Gated Attn. BiLSTM encoders2018.0873.6
Shortcut stacked BiLSTM# Params=140M, Additional training data=SNLI dataset2018.0973.6
BiLSTM with gated pooling# Params=12M2018.0973.6
Gated BiLSTM2017.1173.6
SS BILSTM2017.1173.6
Shortcut-Stacked BiLSTM encodersSNLI Mix=true, Parameter count=140.2m2017.1273.5
Shortcut-Stacked BiLSTM2018.0873.5
3-layer Bi-CAS-LSTM# Params=8.6M2018.0973.4
2-layer CAS-LSTM# Params=2.9M2018.0973.3
3-layer CAS-LSTM# Params=4.8M2018.0973.1
HBMPdimensions=600D2018.0873
Distance-based Self-Attention NetworkSNLI Mix=false, Parameter count=4.7m2017.1272.9
InnerAttEnsemble=true2017.0972.8
BiLSTM + Inner-attentionSNLI Mix=true2017.1272.1
InnerAttEnsemble=false2017.0972.1
ESIMEnsemble=false2017.0972.1
ESIMensemble=false2018.0572.1
BiLSTM + Inner-attention2018.0872.1
Directional Self-Attention NetworkSNLI Mix=false, Parameter count=2.4m2017.1271.4
DiSAN2017.1171.4
BiLSTM + enhanced embedding + max poolingSNLI Mix=false2017.1270.8
BiLSTM + enh embed + max pooling2018.0870.8
ESIM dictionaryAuxiliary Information=dictionary definitions2017.0670.7
ESIM spellingAuxiliary Information=spelling2017.0669.76
ESIM baselineAuxiliary Information=none2017.0668.57
Cha-level Intra-attention BiLSTM encodersSNLI Mix=true2017.1268.2
BiLSTMEnsemble=false2017.0967.6
BiLSTMSNLI Mix=true2017.1267.1
BiLSTM2018.0867.1
BiLSTM# Params=null2018.0966.9
BILSTM2017.1166.9
CBOWSNLI Mix=true2017.1264.6
CBOW2018.0864.6
CBOW# Params=null2018.0964.5
CBOW2017.1164.5