Skip to main content
. Author manuscript; available in PMC: 2023 Mar 28.
Published in final edited form as: J Chem Inf Model. 2022 Mar 10;62(6):1376–1387. doi: 10.1021/acs.jcim.1c01467

Figure 4:

Figure 4:

Test results on USPTO_500_MT. We evaluated both accuracy and SMILES invalidity for top-k predictions on different tasks. Individual tasks and multi-task training schemes have comparable performance in accuracy for all three tasks. But combined training gives much lower invalid SMILES rate. Note that one may have a high top-k accuracy and high SMILES invalidity at the same time.