At the moment, I do not have a good understanding of multi-modality for high-dispersion spectra though, intuitively, high-res seems to be less problematic than low-res retrievals. Numpyro v0.7 can use JAXNS (Nested sampling). So, probably we can test the multi-modality using JAXNS? This is also suggested by an anonymous reviewer for Paper I.
At the moment, I do not have a good understanding of multi-modality for high-dispersion spectra though, intuitively, high-res seems to be less problematic than low-res retrievals. Numpyro v0.7 can use JAXNS (Nested sampling). So, probably we can test the multi-modality using JAXNS? This is also suggested by an anonymous reviewer for Paper I.