Table 3.

Results of GuacaMol distribution-learning benchmarks

Metric	SMILES-based				Graph-based
	LSTM	VAE	AAE	ORGAN	GraphMCTS	Proposed
Validity	0.959	0.870	0.822	0.379	1.000	0.830
Uniqueness	1.000	0.999	1.000	0.841	1.000	0.944
Novelty	0.912	0.974	0.998	0.687	0.994	1.000
KLD	0.991	0.982	0.886	0.267	0.522	0.554
FCD	0.913	0.863	0.529	0.000	0.015	0.016