Embodiments may include storing a query result in a results set cache with the query result being generated from execution of a particular query on a plurality of data records. One or more probabilistic data structures may be generated based, at least in part, on a subset of the data records that is reflected in the query result. An indication may be received that the plurality of data records has been modified. It may then be determined whether to invalidate the query result in the results set cache based, at least in part, on the one or more probabilistic data structures and a representation of modified plurality of data records.