Comment by tgsovlerkhgsel

I'd be much more interested in how the biases of the models differ, and in which direction they're biased. Are there any metrics on that?