[Retracted] Analyzing the Effect of Masking Length Distribution of MLM: An Evaluation Framework and Case Study on Chinese MRC Datasets

Table 4

Statistics of the proposed multiple-choice cloze datasets.

	Short multiple-choice cloze			Long multiple-choice cloze
	Train set	Dev set	Test set	Train set	Dev set	Test set
Passages #	4500	1000	1000	4500	1000	1000
Blanks #	40,500	9000	9000	40,500	9000	9000
Max tokens in a passage #	1000	1000	1000	1000	1000	1000
Max answer tokens #	14	14	14	29	29	29
Min answer tokens #	7	7	7	17	17	17
Options #	9	9	9	9	9	9