Discussion Same Brain but different behavior for the same prefabs.

Hi all! I'v just trained six same agents with a complicated tasks for 5000000 steps. However, different agents perform differently both in training process and after training. Some can behave successfully while others can't. I wonder why this happens. Is this because of the convergence problem that the policy still gives some bad random actions? In that case, to solve the problem, should I just make it train for more steps? Thank you!