[Benchmark] Add verification logic

[sea212]

Currently the results of the benchmarks are not verified. Consequently, we have no way to know whether the expected benchmark behavior is equal to the actual benchmark behavior, eventually deriving weights based on incorrect benchmarks.