Table E.1.

Averaged shear response from model predictions $\widehat{R}$ against measurements R across four sets of simulations with varying observing conditions. The model is always trained on (0.2,6.0,0.6). The balanced Brier score losses (BBSL) of the detection emulator are shown in the last row. A BBSL of 0 indicates perfect classification, whereas uniformly random predictions in the range (0, 1) would result in BBSL ≈0.333.

	Rblend × 102	${\widehat{R}}_{\mathrm{blend}}\times {10}^2$	BBSL
(0.2,6.0,0.6)	1.0787 ± 0.0375	1.1401 ± 0.0014	0.062
(0.2,6.0,0.4)	0.9722 ± 0.0486	0.7870 ± 0.0018	0.105
(0.1,1.5,0.4)	0.8990 ± 0.0536	0.8386 ± 0.0018	0.088
(0.1,6.0,0.6)	0.5064 ± 0.0652	0.4944± 0.0023	0.061