FeverAlbert is a classifier model that predicts if evidence is related to query claim. The model achieved F1 score of 88.33% with test dataset "mwong/fever-claim-related". Using pretrained albert-base-v2 model, the classifier head is trained on Fever dataset.
