A Experimental Details

Environments The episode length is 150 for RoboBin and RoboKitchen and 1000 for RoboYoga. We show all goals in Figure A.1. For both Walker and Quadruped, the success criterion is based on the largest violation across all joints. The global rotation of the Quadruped is expressed as the three independent Euler angles. Global position is not taken into account for the success computation.