Forward-direction molecular property editing on MolEditRL 500 molecules (test)
94.4logP ScoreSLIM
Evaluation Results
| Method | Links | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SLIMBackbone=Mistral (Mi), Steering=SLIM2026.05 | 94.4 | 98.2 | 90.8 | 93 | 75 | 95.8 | 54.2 | 47.4 | 81.1 | — | — | — | — | — | — | — | |
| SLIMSimilarity Threshold (τ)=0.15, Number of candidates (n)=5, Base Model Selection=best result across four models2026.05 | 94.4 | — | — | 93 | 85.6 | 95.8 | 77.4 | 74.2 | 86.7 | — | — | — | — | — | — | — | |
| MolEditRLSimilarity Threshold (τ)=0.15, Number of candidates (n)=52026.05 | 91 | — | — | 85.6 | 76.4 | 82.8 | 82.6 | 84.2 | 83.8 | — | — | — | — | — | — | — | |
| SLIMBackbone=LLaMA-3 (L3), Steering=SLIM2026.05 | 89.6 | 97.2 | 88.8 | 89.8 | 64.2 | 94.8 | 47.6 | 42.2 | 76.8 | — | — | — | — | — | — | — | |
| SFTBackbone=LLaMA-3 (L3), Steering=None2026.05 | 88.2 | 95.8 | 88.2 | 86.8 | 55.8 | 94 | 45 | 28.8 | 72.8 | — | — | — | — | — | — | — | |
| SLIMBackbone=MolGen (MG), Steering=SLIM2026.05 | 86.4 | 76.2 | 83.6 | 87 | 85.6 | 49.4 | 55 | 40.8 | 70.5 | — | — | — | — | — | — | — | |
| SLIMBackbone=DrugAssist (DA), Steering=SLIM2026.05 | 61.2 | 77.6 | 50.6 | 46.8 | 41.2 | 67.6 | 77.4 | 74.2 | 62.1 | — | — | — | — | — | — | — | |
| SLIMBackbone=Mistral (Mi), Steering=SLIM2026.05 | 60.2 | 48.6 | 47.2 | 63.8 | 43.6 | 55.4 | 30.6 | 30 | 47.4 | — | — | — | — | — | — | — | |
| SFTBackbone=Mistral (Mi), Steering=None2026.05 | 58 | 71 | 52.4 | 51.4 | 38.6 | 65.4 | 49.4 | 37.4 | 52.9 | — | — | — | — | — | — | — | |
| SFTBackbone=DrugAssist (DA), Steering=None2026.05 | 55.8 | 77 | 47.6 | 41.8 | 36.4 | 67.6 | 50.6 | 49.6 | 53.3 | — | — | — | — | — | — | — | |
| SLIMBackbone=DrugAssist (DA), Steering=SLIM2026.05 | 51.4 | 61.8 | 42.6 | 38.6 | 33.8 | 55.6 | 56 | 55.2 | 49.4 | — | — | — | — | — | — | — | |
| SLIMBackbone=LLaMA-3 (L3), Steering=SLIM2026.05 | 48.6 | 34.4 | 30.4 | 48.4 | 29.8 | 35.4 | 27.2 | 23.8 | 34.8 | — | — | — | — | — | — | — | |
| SFTBackbone=MolGen (MG), Steering=None2026.05 | 46.2 | 42.4 | 37.2 | 48.2 | 43.2 | 44.4 | 42.6 | 33.8 | 42.3 | — | — | — | — | — | — | — | |
| SLIMBackbone=MolGen (MG), Steering=SLIM2026.05 | 46.2 | 36.6 | 33.4 | 46.4 | 35.4 | 31.2 | 32.4 | 18.2 | 35 | — | — | — | — | — | — | — | |
| SFTBackbone=Mistral (Mi), Steering=None2026.05 | 44 | 52.2 | 40.2 | 37 | 29.6 | 49.8 | 37 | 25.4 | 39.4 | — | — | — | — | — | — | — | |
| SFTBackbone=DrugAssist (DA), Steering=None2026.05 | 43.4 | 60.4 | 35.8 | 32.2 | 27 | 54.8 | 38.4 | 37.2 | 41.4 | — | — | — | — | — | — | — | |
| SFTBackbone=LLaMA-3 (L3), Steering=None2026.05 | 43.2 | 32.4 | 29.6 | 44.6 | 26.4 | 32.6 | 20.8 | 13.8 | 30.4 | — | — | — | — | — | — | — | |
| REINVENT4Similarity Threshold (τ)=0.15, Number of candidates (n)=52026.05 | 36 | — | — | 34 | 38.4 | 2.4 | 40 | 45.8 | 32.8 | — | — | — | — | — | — | — | |
| SFTBackbone=MolGen (MG), Steering=None2026.05 | 27.2 | 19.4 | 23 | 22.6 | 18 | 25.2 | 23 | 16.8 | 21.9 | — | — | — | — | — | — | — | |
| MolEditRLSimilarity Threshold (τ)=0.65, Number of candidates (n)=52026.05 | — | — | — | — | — | — | — | — | — | 57.8 | 40.4 | 39.2 | 62.8 | 48.4 | 58.2 | 51.1 | |
| REINVENT4Similarity Threshold (τ)=0.65, Number of candidates (n)=52026.05 | — | — | — | — | — | — | — | — | — | 11.4 | 6.8 | 11.2 | 1 | 19 | 26.8 | 12.7 | |
| SLIMSimilarity Threshold (τ)=0.65, Number of candidates (n)=5, Base Model Selection=best result across four models2026.05 | — | — | — | — | — | — | — | — | — | 60.2 | 63.8 | 43.6 | 55.6 | 56 | 55.2 | 55.7 |