Accuracy benchmark

Thank you for opening this repository to reproduce researchs. I have a question on the meaning of 2 metrics that usually been used in the main evaluation, which are $\overline{A}$ and $A\_{B}$. Is $\overline{A}$ calculated by taking average of the accuracy measurements from all of the tasks after the last session, and $A\_{B}$ is the top-1 accuracy in the last session?
