HellaSwag: Can a Machine Really Finish Your Sentence?

Benchmark Model Rank Results
sentence-completion-on-hellaswagBERT-Large 340M#188Accuracy: 47.3
sentence-completion-on-hellaswagGPT-1 117M#196Accuracy: 41.7
sentence-completion-on-hellaswagBERT-Base 110M#203Accuracy: 40.5
sentence-completion-on-hellaswagLSTM + BERT-Base#207Accuracy: 36.2
sentence-completion-on-hellaswagESIM + ElMo#213Accuracy: 33.3
sentence-completion-on-hellaswagLSTM + GloVe#215Accuracy: 31.7
sentence-completion-on-hellaswagfastText#216Accuracy: 31.6
sentence-completion-on-hellaswagLSTM + ElMo#217Accuracy: 31.4
sentence-completion-on-hellaswagRandom chance baseline#221Accuracy: 25