Chemical Reaction Prediction on USPTO50k
75.1Accuracy (%)RxnNano-0.5B (with AAM)
Evaluation results
| Method | Links | |
|---|---|---|
| RxnNano-0.5B (with AAM)Size=0.5B, ExtraData=No, Tokenizer=No, LLM Category=Finetuned LLM2026.02 | 75.1 | |
| RxnNano-0.5B (without AAM)Size=0.5B, ExtraData=No, Tokenizer=No, LLM Category=Finetuned LLM2026.02 | 69.8 | |
| RetroDFM-R-7BSize=7B, ExtraData=Chemical, Tokenizer=No, LLM Category=Finetuned LLM2026.02 | 59 | |
| ChemDual (Llama-3.1-8B)Size=8B, ExtraData=No, Tokenizer=Modified, LLM Category=Finetuned LLM2026.02 | 49.95 | |
| MCESetting=Offline2026.01 | 20 | |
| MCESetting=Online2026.01 | 20 | |
| ACESetting=Offline2026.01 | 18 | |
| MCE (w/o skills)Setting=Online2026.01 | 18 | |
| ChemDFM-v1.5-13BSize=13B, ExtraData=Chemical, Tokenizer=No, LLM Category=Finetuned LLM2026.02 | 17.9 | |
| GEPASetting=Offline2026.01 | 15 | |
| MIPROv2Setting=Offline2026.01 | 14 | |
| DCSetting=Online2026.01 | 14 | |
| ACESetting=Online2026.01 | 13 | |
| o1-miniSize=-, ExtraData=No, Tokenizer=No, LLM Category=Base LLM2026.02 | 12.2 | |
| GPT-4.1-miniSize=-, ExtraData=No, Tokenizer=No, LLM Category=Base LLM2026.02 | 11.3 | |
| DeepSeek-R1Size=671B, ExtraData=No, Tokenizer=No, LLM Category=Base LLM2026.02 | 11.2 | |
| ICLSetting=Offline2026.01 | 9 | |
| DeepSeek-V3Size=671B, ExtraData=No, Tokenizer=No, LLM Category=Base LLM2026.02 | 8.6 | |
| Base ModelSetting=Base2026.01 | 6 | |
| GPT-4oSize=-, ExtraData=No, Tokenizer=No, LLM Category=Base LLM2026.02 | 0.7 | |
| Qwen2.5-0.5BSize=0.5B, ExtraData=No, Tokenizer=No, LLM Category=Base LLM2026.02 | 0 |