What's the SE-3-dev?

[mfxss]

I am doing research on SE-3 WSD. And I want do compare with your method.
So I want to use the same dev set as yours.
Thank you very much.

[orenmel]

We followed a leave-one-out approach on the entire train set. For our nearest-neighbor model, this is very simple - for every instance x in the train set T, we look for the nearest instance y in {T \ x} and then use the labels of y to classify x.