The paper, compositional learning for interleaving parallel automata, by Labbai, Groote, Hojjat and Mousavi has been accepted for FOSSACS 23. It describes an algorithm to learn behaviour of parallel components, without knowing the parallel structure.