Significant Discrepancy in Robot Motion Between Simulations Despite Using the Same Policy

[zzHezzo]

I am encountering an issue during Sim-to-Sim transfer in the motion_tracking_controller project. The motion exhibited by the robot in the MuJoCo window is completely different from the robot's motion during Policy Evaluation in the whole_body_tracking project, even though both are using the same trained policy.
This discrepancy in manifested motion, despite starting from an identical policy, suggests a potential inconsistency in the simulation environments or the evaluation pipelines.