Appendix for Learning to Predict Trustworthiness with Steep Slope Loss Y an Luo