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

Table 11

The evaluation results of pretrained models and human performance on cloze datasets.

Models	Short cloze dataset		Long cloze dataset
	QAC	PAC	QAC	PAC
Human performance	97.45	92.87	95.24	90.71
PLM with short cloze MLM	64.63	17.10	65.82	17.60
PLM with long cloze MLM	63.28	15.20	67.02	18.60