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

Table 10

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

Models	Short span dataset		Long span dataset
	F1	EM	F1	EM
Human performance	93.56	85.34	90.47	82.56
PLM with short span MLM	25.03	23.93	13.09	10.79
PLM with long span MLM	24.74	23.31	13.28	11.06