MTQA: Text-Based Multitype Question and Answer Reading Comprehension Model

Table 6

Decoding head comparative experiments.

Models	Accuracy
	EM	F1
MTQApre_Electra_Base	75.67	79.99
Remove segment prediction head	45.82	46.60
Remove arithmetic expression head	52.68	56.87
Remove counting head	69.69	73.69
Remove negation head	72.78	76.95
Remove basic decoding head	74.72	79.18