List your company
ResearchFormerly WizWand

Machine Reading Comprehension on Daily Mail (val)

81.5AccuracyGA Reader
62.57267.48672.477.314Mar 4, 2016Apr 14, 2016May 25, 2016Jul 5, 2016Aug 15, 2016Sep 25, 2016Nov 5, 2016
Updated 8mo ago

Evaluation results

MethodLinks
GA Readervariant=GA (update L(w))2016.0681.5
GA Readervariant=GA (+feature, update L(w))2016.0680.7
GA Readervariant=GA (fix L(w))2016.0680.4
BiDAFensemble=false2016.1180.3
BIDAF2016.0680.3
Stanford ARensemble=true2016.1180.2
Stanford AR (relabeling)ensemble=true2016.0680.2
GA Readervariant=GA (+feature, fix L(w))2016.0680
GAReaderensemble=true2016.1179.1
AS ReaderAggregation=greedy ensemble2016.0378.7
AS Readerensemble=true2016.1178.7
AS Readerensemble=true2016.0678.7
AS ReaderAggregation=avg ensemble2016.0378.1
Stanford ARensemble=false2016.1177.6
ReasoNetensemble=false2016.1177.6
Stanford AR (relabeling)2016.0677.6
ReasoNet2016.0677.6
ReasoNetTmax=52016.0977.6
Chen et al. (2016)Model configuration=single model2016.0376.9
Stanford AR2016.0976.9
GAReaderensemble=false2016.1176.7
GA Readervariant=GA--2016.0676.7
GA Reader2016.0976.7
AS ReaderModel configuration=single model2016.0375
AS Readerensemble=false2016.1175
AS Reader2016.0675
AS Reader2016.0975
AS ReaderAggregation=avg for top 20%2016.0374.5
Attentive ReaderModel type=single model2016.0370.5
Attentive Readerensemble=false2016.1170.5
Attentive Reader2016.0670.5
Attentive Reader2016.0970.5
Impatient ReaderModel type=single model2016.0369
Impatient Reader2016.0669
Deep LSTM Reader2016.0663.3
Deep LSTM Reader2016.0963.3