There are actually some psychophysical differences that are worth considering.

Light mode, because it is light, tends to cause more constricted pupils, leading to sharper vision. For people where visual acuity is even remotely a problem, light mode is superior. Note that this is everyone who scales their font to get as many vertical lines visible as possible, because you do it until you start stressing your visual acuity.

Dark mode, because it is dark, allows considerably broader discrimination between colors of thin lines: the signal of a dim red leaps out against a black background, but good luck telling dark (slightly reddish) from dark (slightly greenish) against the background of a white glare. Because of compensation for the brightness (and pupil constriction...), the red signal is considerably impaired.

This doesn't mean that you have to follow the psychophysics. For example, people plot single-cell RNAseq data consisting of tens of thousands of tiny colored dots, sometimes of dozens of different colors, on a white background. Why? It looks better. A dark mode UMAP would allow better determination of cell types, but eh, the spatial separation is supposed to do that anyway.