Table 3. Prediction Accuracies (R²) for the Random Datasets of the SMC Reaction^a.

training:test	70:30	50:50	30:70	20:80	10:90	5:95	2.5:97.5	1:99
one-hot-RF	0.84 (0.01)	0.82 (0.00)	0.78 (0.01)	0.75 (0.01)	0.68 (0.01)	0.61 (0.03)	0.51 (0.04)	0.24 (0.08)
random-RF	0.84 (0.01)	0.82 (0.01)	0.78 (0.00)	0.75 (0.01)	0.68 (0.01)	0.60 (0.01)	0.46 (0.06)	0.27 (0.05)
yield-BERT	0.80 (0.01)	0.78 (0.06)	0.73 (0.01)	0.68 (0.01)	0.58 (0.02)	0.49 (0.02)	0.36 (0.06)	0.25 (0.04)
T5Chem	0.88 (0.01)	0.87 (0.01)	0.82 (0.01)	0.79 (0.01)	0.69 (0.01)	0.58 (0.03)	0.41 (0.05)	0.20 (0.11)
XGBoost	0.86 (0.01)	0.85 (0.00)	0.81 (0.00)	0.77 (0.01)	0.70 (0.01)	0.63 (0.01)	0.51 (0.04)	0.32 (0.07)
MPNN-transformer	0.89 (0.01)	0.86 (0.01)	0.82 (0.01)	0.78 (0.01)	0.67 (0.02)	0.56 (0.02)	0.41 (0.02)	0.09 (0.14)
MPNN-transformer(3.0m)	0.88 (0.01)	0.85 (0.01)	0.79 (0.01)	0.75 (0.01)	0.69 (0.02)	0.59 (0.01)	0.44 (0.03)	0.09 (0.14)

^aFor each model and dataset, the average (standard deviation) of the R² value for the Random datasets of the SMC reaction using the five ensemble models is reported. For each test dataset, the highest R² value is highlighted in bold.