Our work on what common assumptions in neurosymbolic systems imply in terms of expressiveness and learnability is accepted at ICML 2024.