Knowledge-in-Context: Towards Knowledgeable Semi-Parametric Language Models

Benchmark Model Rank Results
common-sense-reasoning-on-winograndeKiC-770M–Accuracy: 55.30
coreference-resolution-on-winograd-schema-challengeKiC-770M–Accuracy: 65.40
natural-language-inference-on-anli-testKiC-770M–A1: 36.30A2: 35.00A3: 37.60
natural-language-inference-on-rteKiC-770M–Accuracy: 74.00
question-answering-on-copaKiC-770M–Accuracy: 85.30
question-answering-on-storyclozeKiC-770M–Accuracy: 94.40
sentence-completion-on-hellaswagKiC-770M–Accuracy: 29.6
word-sense-disambiguation-on-words-in-contextKiC-770M–Accuracy: 52.40