BERT large finetune UDA

Reported on 5 benchmarks across 1 task · 1 paper · 2 SOTA

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

Natural Language Processing5 results

Sentiment AnalysisonYelp Binary classification
Error· 2019-04-29
2.05
best: 1.37 (XLNet)
SOTA
Unsupervised Data Augmentation for Consistency Training arXiv:1904.12848
Sentiment AnalysisonIMDb
Accuracy· uses extra data· 2019-04-29
95.8
best: 96.68 (RoBERTa-large with LlamBERT)
SOTA
Unsupervised Data Augmentation for Consistency Training arXiv:1904.12848
Sentiment AnalysisonAmazon Review Polarity
Accuracy· 2019-04-29
96.5
best: 97.37 (BERT large)
Unsupervised Data Augmentation for Consistency Training arXiv:1904.12848
Sentiment AnalysisonYelp Fine-grained classification
Error· 2019-04-29
32.08
best: 27.05 (XLNet)
Unsupervised Data Augmentation for Consistency Training arXiv:1904.12848
Sentiment AnalysisonAmazon Review Full
Accuracy· 2019-04-29
62.88
best: 65.83 (BERT large)
Unsupervised Data Augmentation for Consistency Training arXiv:1904.12848