I don't think so. 'GroupBy' queries don't actually contain any records from the dataset. The best you can do is be aware of when changes to the dataset are made, and then recalculate your query.
There are ways to get around this though - for example, if your dataset contains a unique incremental field, like an AutoNumber, or transaction number, then you could remember the highest number, and only re-query for the new records, and add the results with the previous GroupBy query. If your GroupBy query takes a long time to compile, this sort of approach may give good performance.