CASA-B: A Unified Framework of Model-Free Reinforcement Learning