List your company
ResearchFormerly WizWand

Natural Language Inference on MultiNLI mismatched (cross-domain) RepEval 2017 (test)

75.8AccuracyOur ESIM
64.04867.09970.1573.201Aug 4, 2017Aug 10, 2017Aug 17, 2017Aug 24, 2017Aug 31, 2017Sep 7, 2017Sep 14, 2017
Updated 8mo ago

Evaluation results

MethodLinks
Our ESIMimplementation=authors2017.0875.8
Ensembledensemble=true, RepEval 2017 participant=true2017.0874.9
Singleensemble=false, RepEval 2017 participant=true2017.0873.6
Single (Input Gate)gate_type=input, RepEval 2017 participant=true2017.0873.6
Single - Gated-Attgated_attention=removed2017.0873.6
Shortcut-Stacked Encoder2017.0873.6
YixinNie-UNC-NLPRepEval 2017 participant=true2017.0873.5
Single - CharCNNcharacter_composition=removed2017.0873.5
Single (Output Gate)gate_type=output2017.0873.4
Single (Forget Gate)gate_type=forget2017.0873.1
RivercornersRepEval 2017 participant=true, ensemble=true2017.0872.8
ESIM2017.0872.1
RivercornersRepEval 2017 participant=true, ensemble=false2017.0872.1
DiSAN2017.0971.402
biLSTM-Max Encoderreimplementation=true2017.0871.2
LCT-MALTARepEval 2017 participant=true2017.0870.8
Single - AbsDiff/Productelementwise_vector_comparison=removed2017.0869.2
TALP-UPCRepEval 2017 participant=true2017.0868.2
Bi-LSTM2017.0967.248
biLSTM Encoder2017.0867.1
BiLSTM2017.0866.9
Single - Word Embeddingword_level_embedding=removed2017.0866
CBOW2017.0964.759
CBOW2017.0864.6
CBOW2017.0864.5