Why didn't the high-dimensional gaussian MI-NEE converge in the demo?

[nanguoyu]

Hi,

Thank you for the code. I really like your paper.

I am playing with the demo Higher dimensional Gaussian with MI-NEE. But the result didn't show the same trend in the figure 2(a) of the arxiv paper. It seems not converge. Is it somehow a typo or bug in this demo?

BR,