New preprint on what common assumptions in neurosymbolic systems imply in terms of expressiveness and learnability.