Some Questions About Clone Detection experiments

[fly1ngpengu1ns]

Hi，
I was recently researching this work and I am very interested in it. However, I noticed that in the code clone detection experiments, the paper mentioned that except for GrapCodeBert's experiments in POJ-104, you directly taken the results from CodeXGLUE

However, as we know from microsoft/CodeXGLUE#63, the authors of CodeXGLUE found a problem with the code in which they originally calculated the results, and corrected it in their latest paper. But the results quoted in ContraBert are the ones before the update

So I would like to know:

1. Did you use the correct calculation method to get the correct results when you experimented with GraphCodeBert?
2. I would like to know if you processed the annotated information in the dataset and whether the metrics you got were Eval MAP or Test MAP, because I got the following results while the report MAP is 90.46

Eval MAP	Test MAP	comment
0.8451	0.8795	with comment
0.8463	0.8926	without comment

I am very much looking forward to your answers!