I encountered a strange case where the maxLoss reported in the pareto front of one of the MvSR solution is much lower than any of the reported losses. Therefore the model achieves the lowest loss even though it is very bad. Below are two screenshots from the same pareto front. One is a well behaving solution and the other one shows the bug.
I can transfer the full pareto front if needed
I encountered a strange case where the maxLoss reported in the pareto front of one of the MvSR solution is much lower than any of the reported losses. Therefore the model achieves the lowest loss even though it is very bad. Below are two screenshots from the same pareto front. One is a well behaving solution and the other one shows the bug.
I can transfer the full pareto front if needed