The gp meld method in sklearn interface

[Hintonein]

I have seen the code in core.py, and the gp meld method isn't related to the benchmark dataset. But I also found that I cannot use the sklearn interface to use the gpmeld method in sklearn.py. When I compared the operation mode of BenchmarkDataset class and the sklearn-like (X,y) in regression.py, I think there is no difference between the two except that the benchmark dataset specifies the set of symbols to be used. Does this mean that DSO cannot achieve good results in symbolic regression tasks other than the specified benchmark dataset? Or is there any way to run dso effectively with sklearn-like (X,y) data using gp-meld method?