A Statistical Framework for Low-bitwidth Training of Deep Neural Networks

Open paper

Benchmark	Model	Rank	Results
linguistic-acceptability-on-cola	PSQ (Chen et al., 2020)	#15	Accuracy: 67.5
natural-language-inference-on-multinli	PSQ (Chen et al., 2020)	#10	Matched: 89.9
natural-language-inference-on-qnli	PSQ (Chen et al., 2020)	#13	Accuracy: 94.5
natural-language-inference-on-rte	PSQ (Chen et al., 2020)	#20	Accuracy: 86.8
semantic-textual-similarity-on-mrpc	PSQ (Chen et al., 2020)	#9	Accuracy: 90.4
semantic-textual-similarity-on-sts-benchmark	PSQ (Chen et al., 2020)	#6	Pearson Correlation: 0.919
sentiment-analysis-on-sst-2-binary	PSQ (Chen et al., 2020)	#19	Accuracy: 96.2