Each square is one question, shaded by difficulty (darker means harder). ✓ right, ✗ wrong. The estimates on the right come from the same model as Fig. 6.