BiDAF++ (single model)

Natural Language Processing7 results

• Question AnsweringonCoQA

  In-domain· 2018-09-27

  69.4

  best: 82.5 (BERT Large Augmented (single model))

  SOTA
  A Qualitative Comparison of CoQA, SQuAD 2.0 and QuACarXiv:1809.10735
• Question AnsweringonCoQA

  Out-of-domain· 2018-09-27

  63.8

  best: 77.6 (BERT Large Augmented (single model))

  SOTA
  A Qualitative Comparison of CoQA, SQuAD 2.0 and QuACarXiv:1809.10735
• Question AnsweringonCoQA

  Overall· 2018-09-27

  67.8

  best: 85 (GPT-3 175B (few-shot, k=32))

  SOTA
  A Qualitative Comparison of CoQA, SQuAD 2.0 and QuACarXiv:1809.10735
• Question AnsweringonSQuAD1.1

  EM

  77.573

  best: 90.622 ({ANNA} (single model))
• Question AnsweringonSQuAD1.1

  F1

  84.858

  best: 95.719 ({ANNA} (single model))
• Question AnsweringonSQuAD2.0

  EM

  65.651

  best: 90.939 (IE-Net (ensemble))
• Question AnsweringonSQuAD2.0

  F1

  68.866

  best: 93.214 (IE-Net (ensemble))