Controllable Molecule Generation on QM9 (test)
0.1Alpha MAE (Bohr^3)QM9*
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| QM9*description=Lower bound; directly running classifier on ground-truth second half of QM92023.05 | 0.1 | 64 | 39 | 36 | 0.043 | 0.04 | |
| QM9*2023.12 | 0.1 | 64 | 39 | 36 | 0.043 | 0.04 | |
| Twigs2024.10 | 1.36 | 323 | 225 | 244 | 0.627 | 0.559 | |
| JODO2024.10 | 1.42 | 335 | 226 | 256 | 0.628 | 0.581 | |
| Twigs2024.10 | 1.46 | — | — | — | 0.712 | — | |
| JODO2024.10 | 1.52 | — | — | — | 0.717 | — | |
| TEDMol2024.10 | 2.24 | 443 | 279 | 412 | 0.84 | 0.847 | |
| TEDMol2024.10 | 2.27 | — | — | — | 0.809 | — | |
| GEOLDMdescription=Geometric Latent Diffusion Model modeling in latent space2023.05 | 2.37 | 587 | 340 | 522 | 1.108 | 1.025 | |
| GeoLDM2024.10 | 2.37 | 587 | 340 | 522 | 1.108 | 1.025 | |
| EQUIFM2023.12 | 2.41 | 591 | 337 | 530 | 1.106 | 1.033 | |
| EquiFM2024.10 | 2.41 | 591 | 337 | 530 | 1.106 | 1.033 | |
| EEGSDE2024.10 | 2.5 | 487 | 302 | 447 | 0.777 | 0.941 | |
| EEGSDE2024.10 | 2.61 | — | — | — | 0.855 | — | |
| EDMdescription=Diffusion model running on atomic features2023.05 | 2.76 | 655 | 356 | 584 | 1.111 | 1.101 | |
| EDM2023.12 | 2.76 | 655 | 356 | 584 | 1.111 | 1.101 | |
| EDM2024.10 | 2.76 | — | — | — | 1.158 | — | |
| EDM2024.10 | 2.78 | 671 | 371 | 601 | 1.123 | 1.065 | |
| Natomsdescription=Property prediction using only the number of atoms2023.05 | 3.86 | 866 | 426 | 813 | 1.053 | 1.971 | |
| Natoms2023.12 | 3.86 | 866 | 426 | 813 | 1.053 | 1.971 | |
| Random*description=Upper bound; random shuffling of property labels2023.05 | 9.01 | 1,470 | 645 | 1,457 | 1.616 | 6.857 | |
| Random*2023.12 | 9.01 | 1,470 | 645 | 1,457 | 1.616 | 6.857 |