Table 6.

* Pearson correlation and associated p value between each score and aggregated uncertainty (the aggregated uncertainty value of each sampled collection equals to the value of hyper-parameter $\tau$ being set during the uncertainty sampling process, the test suite scores are averaged by taking the sum of all the instances scores and divided by the number of instances within each collection); ${}^{\wedge }$ The Pearson correlation and associated p value between each evaluation metric and per-instance prediction uncertainty

	Predictions	Pearson R	p value
Precision ${}^{\ast }$	Consistent	− 0.95	9e$-$08
	Type A	− 0.06	0.82
	Type B	− 0.85	7e$-$05
	Type C	− 0.70	3e$-$03
GOLen${}^{\wedge }$	Consistent	0.27	6e$-$28
	Type B	− 0.04	0.14
	Type C	0.16	3e$-$07
AlignRatio${}^{\wedge }$	Consistent	0.05	0.03
	Type B	− 0.09	9e$-$05
	Type C	0.16	2e$-$07
GEORatio${}^{\wedge }$	Consistent	− 0.13	4e$-$08
	Type B	− 0.04	0.09
	Type C	0.1	9e$-$04
%ContainRoman${}^{\ast }$	Consistent	0.79	5e$-$04
	Type B	− 0.26	0.35
	Type C	− 0.46	0.08
%ContainDigit${}^{\ast }$	Consistent	0.81	3e$-$04
	Type B	− 0.44	0.10
	Type C	− 0.65	9e$-$03
%ContainStop-OF${}^{\ast }$	Consistent	0.81	2e$-$04
	Type B	0.80	3e$-$04
	Type C	0.58	0.02