UniLM + ASGen

Reported on 3 benchmarks across 1 task

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Natural Language Processing3 results

Question GenerationonSQuAD1.1
BLEU-4· uses extra data
23.7
best: 25.41 (ERNIE-GENLARGE (beam size=5))
Question GenerationonSQuAD1.1
METEOR· uses extra data
25.9
best: 26.73 (BART (TextBox 2.0))
Question GenerationonSQuAD1.1
ROUGE-L· uses extra data
52.3
best: 52.8 (ProphetNet + ASGen)