Fig. 5

Validation evaluation metrics for solvent (blue) and reagent (red) multi-label classification. Panels A and B depict the evolution of F1-scores across epochs using various cutoffs. A cutoff value of 0.3 yields the highest F1-score in the concluding epochs, leading to the adoption of this threshold for categorizing labels as positive or negative in the candidate generation model. Panels C and D showcase hamming loss and example-based precision, recall, and f1-score, all calculated using a 0.3 cutoff