Question Answering On Squad20 Dev

Results

Performance results of various models on this benchmark

Model Name	EM	F1	Paper Title	Repository
ALBERT base	76.1	79.1	ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
RoBERTa (no data aug)	86.5	89.4	RoBERTa: A Robustly Optimized BERT Pretraining Approach
ALBERT large	79.0	82.1	ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
XLNet (single model)	87.9	90.6	XLNet: Generalized Autoregressive Pretraining for Language Understanding
RMR + ELMo (Model-III)	72.3	74.8	Read + Verify: Machine Reading Comprehension with Unanswerable Questions	-
SemBERT large	80.9	83.6	Semantics-aware BERT for Language Understanding
SpanBERT	-	86.8	SpanBERT: Improving Pre-training by Representing and Predicting Spans
SG-Net	85.1	87.9	SG-Net: Syntax-Guided Machine Reading Comprehension
TinyBERT-6 67M	69.9	73.4	TinyBERT: Distilling BERT for Natural Language Understanding
XLNet+DSC	87.65	89.51	Dice Loss for Data-imbalanced NLP Tasks
ALBERT xlarge	83.1	85.9	ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
U-Net	70.3	74.0	U-Net: Machine Reading Comprehension with Unanswerable Questions
ALBERT xxlarge	85.1	88.1	ALBERT: A Lite BERT for Self-supervised Learning of Language Representations

0 of 13 row(s) selected.