Appendix of Decomposed Mutual Information Optimization for Generalized Context in Meta-Reinforcement Learning Y ao Mu The University of Hong Kong