TextBox 2.0: A Text Generation Library with Pre-trained Language Models

Benchmark Model Rank Results
abstractive-text-summarization-on-cnn-daily-2BART (TextBox 2.0)#1ROUGE-1: 44.47ROUGE-2: 21.5ROUGE-L: 41.35
data-to-text-generation-on-webnlgBART (TextBox 2.0)#15METEOR: 47.78BLEU-4: 67.33ROUGE-L: 76.83
machine-translation-on-wmt2016-english-1BART (TextBox 2.0)#20BLEU-4: 37.2
machine-translation-on-wmt2016-romanianBART (TextBox 2.0)#19BLEU-4: 37.48
question-answering-on-squad11BART (TextBox 2.0)#48F1: 93.04Exact Match: 86.44
question-generation-on-squad11BART (TextBox 2.0)#2BLEU-4: 25.08METEOR: 26.73ROUGE-L: 52.55
text-generation-on-commongen-1BART (TextBox 2.0)#1CIDEr: 12.98BLEU-4: 28.18SPICE: 33