This document does not provide a detailed introduction, so I have some doubts about the training modes of BinM1 and BinLong. I am not sure how to determine their TOGGLE, EARLY, and ATE states, and how to update the third-order equilibrium parameters in these three states?