Multi-objective molecular optimization (BPQ) on MuMOInstruct seen instruction
47.1Success Rate (SR)SFT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SFT2026.03 | 47.1 | 24.4 | 11.5 | |
| RePO2026.03 | 27.4 | 50.9 | 13.9 | |
| GRPOinitialization=None2026.03 | 21.2 | 56.7 | 12 | |
| Distill-SFT2026.03 | 6.4 | 16.5 | 1.1 | |
| GRPOinitialization=SFT init2026.03 | 5.6 | 8.5 | 0.5 | |
| Base Model2026.03 | 5.2 | 19.4 | 1 |