Learning Generalizable Robot Policies By Understanding Semantics And Logic From Task Demonstrations