Metric Residual Networks for Sample Efficient Goal-Conditioned Reinforcement Learning | Read Paper on Bytez