KL Divergence Comparison for MRI (averaged over 3 runs):
+---------------------------+---------------------+-----------------------+
| Method                    | Average KL (Prob)   | Average KL (Argmax)   |
+===========================+=====================+=======================+
| Original                  | 6.99e-02 ± 0.00e+00 | 1.22e-01 ± 0.00e+00   |
+---------------------------+---------------------+-----------------------+
| MCal_CE (Cross-Entropy)   | 4.75e-04 ± 0.00e+00 | 1.85e-02 ± 0.00e+00   |
+---------------------------+---------------------+-----------------------+
| PatchCutout-trained Model | 4.04e-04 ± 0.00e+00 | 4.24e-04 ± 0.00e+00   |
+---------------------------+---------------------+-----------------------+
| Platt Scaling             | 1.30e-01 ± 0.00e+00 | 1.33e-01 ± 0.00e+00   |
+---------------------------+---------------------+-----------------------+
| Temperature Scaling       | 2.99e-02 ± 0.00e+00 | 1.22e-01 ± 0.00e+00   |
+---------------------------+---------------------+-----------------------+
| Replace Mean              | 8.45e-02 ± 0.00e+00 | 1.51e-01 ± 0.00e+00   |
+---------------------------+---------------------+-----------------------+
| Arch Mod                  | 8.28e-02 ± 9.77e-04 | 1.36e-01 ± 1.88e-03   |
+---------------------------+---------------------+-----------------------+