About multiple experiments

[Study-gsl]

Thank you for your wonderful work. I encountered a problem. When running multiple experiments, do we fix the random seed and run it again, or change the random seed and run it again? In the experiment, I found that the standard deviation of the results after changing the random seed was larger. Maybe I have a problem in the statistics. Is it the result of the last round or the best result of the 2000 pre-train rounds？ Thank you.