Are you fine tuning (re training) on them or using them in a tool to feed as a context to a larger model?