Retrosynthesis on SMolInstruct (EM/FTS/Validity)
47EMLiterature SOTA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Literature SOTAModel Category=Task-specific, non-LLM based models2026.06 | 47 | 77.5 | 99.7 | |
| SciReasoner-8BModel Category=General / molecular LLM baselines, Parameters=8B2026.06 | 46 | — | — | |
| BioMatrix-4B (SMILES)Parameters=4B, Line Notation=SMILES2026.06 | 45.16 | 74.88 | 99.93 | |
| BioMatrix-4B (SELFIES)Parameters=4B, Line Notation=SELFIES2026.06 | 42.9 | 73.63 | 99.93 | |
| BioMatrix-1.7B (SMILES)Parameters=1.7B, Line Notation=SMILES2026.06 | 42.69 | 73.81 | 99.77 | |
| SciReasoner-1.7BModel Category=General / molecular LLM baselines, Parameters=1.7B2026.06 | 41 | — | — | |
| BioMatrix-1.7B (SELFIES)Parameters=1.7B, Line Notation=SELFIES2026.06 | 39.32 | 71.97 | 99.93 | |
| LlaSMol MistralModel Category=General / molecular LLM baselines2026.06 | 32.9 | 70.4 | 100 | |
| BioMedGPT-Mol-8BModel Category=General / molecular LLM baselines, Parameters=8B2026.06 | 32.4 | 75.2 | 99.9 | |
| GPT-ossModel Category=General / molecular LLM baselines2026.06 | 8 | — | — | |
| Claude Opus 3Model Category=General / molecular LLM baselines2026.06 | 1.1 | 46.2 | 94.8 | |
| GPT-4Model Category=General / molecular LLM baselines2026.06 | 0 | 33.4 | 42.6 |