List your company
ResearchFormerly WizWand

Question Answering on CNN (test)

78.6AccuracyGA + MAGE
62.37666.58870.875.012Nov 5, 2016Nov 25, 2016Dec 15, 2016Jan 5, 2017Jan 25, 2017Feb 14, 2017Mar 7, 2017
Updated 8mo ago

Evaluation results

MethodLinks
GA + MAGEBase Architecture=GA, Coreference augmentation=MAGE, Coreference state size=322017.0378.6
GA + one-hotBase Architecture=GA, Coreference augmentation=one-hot2017.0378.1
GA + MAGEBase Architecture=GA, Coreference augmentation=MAGE, Coreference state size=162017.0378.1
GABase Architecture=GA2017.0377.9
Stanford ARensemble=true2016.1177.6
GAReaderensemble=true2016.1177.4
BiDAFBase Architecture=BiDAF2017.0376.9
BiDAFensemble=false2016.1176.9
Iterative Attentionensemble=true2016.1175.7
AS Readerensemble=true2016.1175.4
ReasoNetensemble=false2016.1174.7
AoA Readerensemble=false2016.1174.4
EpiReaderensemble=false2016.1174
GAReaderensemble=false2016.1173.8
Stanford ARensemble=false2016.1173.6
Iterative Attentionensemble=false2016.1173.3
BiGRU + MAGEBase Architecture=BiGRU, Coreference augmentation=MAGE, Coreference state size=482017.0372.9
DER Networkensemble=false2016.1172.9
BiGRUBase Architecture=BiGRU2017.0370.4
BiGRU + one-hotBase Architecture=BiGRU, Coreference augmentation=one-hot2017.0370.1
AS Readerensemble=false2016.1169.5
MemNNensemble=true2016.1169.4
MemNNensemble=false2016.1166.8
Attentive Readerensemble=false2016.1163