Unsupervised Deep Structured Semantic Models for Commonsense Reasoning

Benchmark Model Rank Results
coreference-resolution-on-winograd-schema-challengeDSSM–Accuracy: 63.0
coreference-resolution-on-winograd-schema-challengeUDSSM-I–Accuracy: 54.5
coreference-resolution-on-winograd-schema-challengeUDSSM-I (ensemble)–Accuracy: 57.1
coreference-resolution-on-winograd-schema-challengeUDSSM-II–Accuracy: 59.2
coreference-resolution-on-winograd-schema-challengeUDSSM-II (ensemble)–Accuracy: 62.4
natural-language-understanding-on-pdp60DSSM–Accuracy: 75.0
natural-language-understanding-on-pdp60UDSSM-I (ensemble)–Accuracy: 76.7
natural-language-understanding-on-pdp60UDSSM-II–Accuracy: 75
natural-language-understanding-on-pdp60UDSSM-II (ensemble)–Accuracy: 78.3