Multi-property Molecular Optimization on BDPQ
65.13Success Rate (SR)DPO-LLM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DPO-LLMBackbone=Mistral, Protocol=DPO2026.04 | 65.13 | 43 | 5.65 | |
| DPO-LLMBackbone=LLaMA, Protocol=DPO2026.04 | 59.2 | 42 | 5.97 | |
| SFT-LLMBackbone=Mistral, Protocol=SFT2026.04 | 57 | 52 | 3.05 | |
| SFT-LLMBackbone=LLaMA, Protocol=SFT2026.04 | 46.15 | 50 | 3.36 | |
| Claude-3.5shot_count=5-shot2026.04 | 20.8 | 35 | 3.53 | |
| LlaSmolBackbone=Mistral2026.04 | 14 | 62 | 1.03 | |
| Mistralshot_count=5-shot2026.04 | 11 | 69 | 0.96 | |
| LLaMAshot_count=5-shot2026.04 | 9.6 | 54 | 3.45 | |
| Claude-3.5shot_count=0-shot2026.04 | 8 | 60 | 1.34 | |
| LLaMAshot_count=0-shot2026.04 | 5.2 | 80 | 0.62 | |
| GPT-4oshot_count=1-shot2026.04 | 3.4 | 70 | 0.61 | |
| Mistralshot_count=0-shot2026.04 | 3.2 | 77 | 0.87 | |
| GPT-4oshot_count=0-shot2026.04 | 1.8 | 71 | 0.39 | |
| ChemLLM2026.04 | 0 | 0 | 0 |