Sample divided in four bins of L_d/L_⋆. The underlying SEDs are the Bayesian CIGALE estimates (UV–MIR CIGALE fit up to WISE 22 μm, UV–FIR CIGALE fit for longer wavelengths), and are normalised by the total stellar luminosity (from the short wavelength CIGALE fit). The markers show the median (per broadband) of each bin, and the shaded region is filled between the 16th and 84th percentile. These percentiles are taken for each Bayesian broadband flux. The CIGALE best models are used to interpolate between the broadband fluxes, in order to guide the eye. In the bottom right, the number of galaxies per bin and the median L_d/L_⋆ per bin are shown.