Table 1:

Uncertainty quantification performances of BAL on protein docking based on η, the relative error in iRMSD; and $\widehat{P}$, the portion of confidence intervals from 100 runs encompassing the global optima. For η, means (and standard deviations in parentheses) are reported.

Dataset	1 – σ	0.99	0.95	0.90	0.85	0.80
Training	η	0.40 (0.23)	0.35 (0.18)	0.31 (0.17)	0.27 (0.15)	0.22 (0.10)
	$\widehat{P}$	0.97	0.91	0.84	0.79	0.75
Test a	η	0.43 (0.26)	0.39 (0.21)	0.28 (0.16)	0.25 (0.13)	0.21 (0.09)
	$\widehat{P}$	0.95	0.87	0.83	0.75	0.73
Test b	η	0.44 (0.22)	0.38 (0.16)	0.26 (0.14)	0.23 (0.10)	0.19 (0.08)
	$\widehat{P}$	0.91	0.84	0.80	0.74	0.71
CAPRI	η	0.43 (0.20)	0.35 (0.13)	0.27 (0.11)	0.22 (0.10)	0.20 (0.09)
	$\widehat{P}$	0.91	0.85	0.81	0.74	0.70