An OpenCompass wrapper for six molecule-oriented Mol-Instructions tasks covering molecular descriptions, design, reactions, properties, and retrosynthesis.
unassessed
| Category | domain |
|---|---|
| Subcategory | molecule-oriented biomolecular instruction evaluation |
| Page status | active |
| Metric | Tanimoto similarity, mean absolute error, or METEOR by task |
| Direction | higher_is_better |
| Unit | task-specific |
| Dataset size | 148400 |
| Dataset licence | CC BY 4.0 |
| Publisher | Zhejiang University NLP |
This benchmark evaluates instruction following for small-molecule chemistry. The molecule-oriented component contains molecule description generation, description-guided molecule design, forward reaction prediction, retrosynthesis, reagent prediction, and property prediction. Inputs use SELFIES in the OpenCompass configuration and outputs are molecular strings, numbers, or natural-language descriptions.
Zero-shot generation. OpenCompass formats each task with a chemistry system prompt and asks for SELFIES tags for molecular outputs, a boxed number for property prediction, or natural language for descriptions. It evaluates molecular outputs with Morgan-fingerprint Tanimoto similarity, numbers with MAE, and descriptions with METEOR.
No model card in ModelSpec reports this benchmark yet.