Table 2.

Summary of mean accuracy/$R^2$ (± standard deviation) for the baseline models when the seed and training set order is changed during training.

Dataset (Task)	Seed	Shuffle	Overall	SotA performance
KAIMRC (Classification)	82.576 ± 0.4668	83.381 ± 0.1811	83.12 ± 0.4779	83.22
KAIMRC (Regression)	0.5927 ± 0.0113	0.579 ± 0.0122	0.5858 ± 0.01326	n/a
BCW	92.185 ± 1.7315	91.5 ± 2.8319	91.843 ± 2.3172	99.04
Codon usage (Kingdom)	85.280 ± 1.8029	85.38 ± 1.1778	85.33 ± 1.4367	84.25
Codon usage (DNA)	99.268 ± 0.0950	99.166 ± 0.0921	99.217 ± 0.1033	99.15
MIMIC-IV	76.362 ± 2.5808	79.736 ± 1.8906	78.049 ± 2.7769	84.72

The state of the art (SotA) model performance is also reported to confirm the models are properly trained.