DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter

Benchmark Model Rank Results
semantic-textual-similarity-on-sts-b-glue-devDistilBERT#9Pearson/Spearman Correlation: 86.9
paraphrase-identification-on-mrpc-glue-devDistilBERT#10F1: 87.5
semantic-textual-similarity-on-mrpcDistilBERT 66M#10Accuracy: 90.2%
semantic-textual-similarity-on-sts-benchmarkDistilBERT 66M#11Pearson Correlation: 0.907
linguistic-acceptability-on-cola-glue-devDistilBERT#12Matthews Correlation: 51.3
question-answering-on-multitqDistillBERT#12Hits@1: 8.3Hits@10: 48.4
natural-language-inference-on-rte-glue-devDistilBERT#13Accuracy: 59.9
natural-language-understanding-on-glue-devDistilBERT#13GLUE (dev): 77.0
question-answering-on-quora-question-pairsDistilBERT 66M#13Accuracy: 89.2%
natural-language-inference-on-qnli-glue-devDistilBERT#15Accuracy: 89.2
natural-language-inference-on-multinli-devDistilBERT#17Matched: 82.2
question-answering-on-squad11-devDistilBERT#20EM: 77.7
question-answering-on-squad11-devDistilBERT 66M#44F1: 85.8
task-1-grouping-on-ocwDistilBERT (BASE)#21# Correct Groups: 49 ± 4Fowlkes Mallows Score (FMS): 29.1 ± .2…
natural-language-inference-on-wnliDistilBERT 66M#23Accuracy: 44.4
sentiment-analysis-on-imdbDistilBERT 66M#28Accuracy: 92.82
linguistic-acceptability-on-colaDistilBERT 66M#34Accuracy: 49.1%
natural-language-inference-on-qnliDistilBERT 66M#35Accuracy: 90.2%
sentiment-analysis-on-sst-2-binary-classificationDistilBERT 66M#54Accuracy: 91.3
natural-language-inference-on-rteDistilBERT 66M#64Accuracy: 62.9%