Text-based de novo molecule generation on ChEBI-20 benchmark leaderboard · SOTA2 Research

Text-based de novo molecule generation on ChEBI-20

86.1BLEULAMOLT5

Evaluation Results

Method	Links
LAMOLT52025.02		86.1	32.5	14.685	89.9	81.9	76	0.352	59.6	96.1
MolReGPT2025.02		85.7	28	17.14	90.3	80.5	73.9	0.41	59.3	89.9
LAMOLT52025.02		85.6	32.8	15.666	89.2	81.6	75.4	0.371	59.3	96.2
T52025.02		85.4	27.9	16.721	82.3	73.1	67	0.401	55.2	90.2
MolT52025.02		85.4	31.1	16.071	83.4	74.6	68.4	0.385	55.4	90.5
LAMOLT52025.02		85.2	28.7	16.009	89.1	80.5	74.1	0.384	59.4	95
TGM-DLM2025.02		82.6	24.2	17	85.4	73.9	68.8	0.77	58.1	87.1
MolT52025.02		76.9	8.1	24.458	72.1	58.8	52.9	1.16	49.6	77.2
T52025.02		76.2	6.9	24.95	73.1	60.5	54.5	1.43	49.9	66
MolT52025.02		75.5	7.9	25.988	70.3	56.8	51.7	1.35	48.2	72.1
Text+Chem T52025.02		75	21.2	27.39	87.4	76.7	69.7	0.499	57.4	79.2
T52025.02		74.1	6.4	27.703	70.4	57.8	52.5	1.77	47.9	60.8
RNN2025.02		65.2	0.5	38.09	59.1	40	36.2	4.34	40.9	54.2
Transformer2025.02		49.9	0	57.66	48	32	21.7	16.03	27.7	90.6