Did you do hyper-parameter tuning for each of the three random seeds?
Or did you hyper-parameter tune using a particular seed and train the model using another three different random seeds to get the test set predictions?
We tune hyper-parameters using three particular seeds. And under the same seeds, we obtain test results with the best hyper-parameters on the validation set.
Did you do hyper-parameter tuning for each of the three random seeds? Or did you hyper-parameter tune using a particular seed and train the model using another three different random seeds to get the test set predictions?