Decreasing rhat for Black-Box Likelihood Model

Well, I’m estimating 4 parameters. The posterior density plots of 3 of them show two or more peaks, i.e., they’re not concentrated on a single value each. I’m again using uniform priors.