Transformer-based WorkingMemoryforMultiagent ReinforcementLearningwithActionParsing