Sensitivity to agreement model Clause Samples

Sensitivity to agreement model. We evaluated how sensitive GAM is to the choice of agreement model architecture. To assess this, we ran GAM multiple times, with a fixed classification model architecture and we various agreement model sizes. Figure 6 in Appendix C shows the test accuracy per co-train iteration for each of these models. The results indicate that the behaviour of GAM is stable with respect to the agreement model size, which suggests that the agreement model size is a parameter that does not require much tuning effort.