We consider a generalization of the quantum state redistribution task, where pure multipartite states from an ensemble source are distributed among
an encoder, a decoder and a reference system. The encoder, Alice, has access to two quantum systems: system $A$ which she compresses and sends to the decoder,
Bob, and the side information system $C$ which she wants to keep at her site. Bob has access to quantum side information in a system $B$, wants to decode
the compressed information in such a way to preserve the correlations with the reference system on average.