If you run into a negative beta with a positive correlation just flip the between-subject contrast and it will make more sense with the red correlation lines and red nodes (i.e., change from [1 -1] to [-1 1])
Best,
MG
Best,
MG