OpenMol/ChemCoTBench

Tasks

mol_und

• fg-level
  • fg_count.json
    • 100 samples across 38 different functional groups detection
  • ring_count.json
    • 20 samples, 9 types of ring unit
• scaffold-level
  • Murcko_scaffold.json
    • 40 samples, using MurckoScaffold extraction
  • ring_system_scaffold.json
    • 60 samples, extract ring system as scaffolds
• SMILES-level
  • equivalence.json
    • 50 samples, each smiles -> mutate -> permutate, mutated smiles differs from original smiles
    • 50 samples, each smiles -> permutate, permutated smiles is same with original smiles

mol_edit

• add.json
  • 20 samples, covers 10 func groups addition
• delete.json
  • 20 samples, covers 10 func groups deletion
• sub.json
  • 60 samples, covers 37 func groups substitution

mol_opt

• drd.json
  • 100 samples, target-level
• gsk.json
  • 100 samples, target-level
• jnk.json
  • 100 samples, target-level
• logp.json
  • 100 samples, physicochemical-level
• qed.json
  • 100 samples, physicochemical-level
• solubility.json
  • 100 samples, physicochemical-level

reaction

• forward reaction prediction
  • fs.json
    • 100 samples, 100 rxn-cls (each has one sample)
• (single-step) retrosynthesis prediction
  • retro.json
    • 100 samples, 100 rxn-cls (each has one sample)
• reaction condition prediction/recommendation
  • rcr.json
    • 90 samples. 10 types of reaction. Each type has 3 samples for 'Catalyst' prediction, 3 for 'Reagent' prediction and 3 for 'Solvent' prediction
• Next elementary step product prediction (NEPP)
  • nepp.json
    • given former elementary steps description, predict next elementary step's product.
    • 85 rxn cls, each has 1 sample
• Mechanism Route Selection (MechSel)
  • mechsel.json
    • 100 samples

Links

• Our larger CoT dataset ChemCoTBench-CoT

Citation

If you find our work helpful, feel free to give us a cite.

@article{li2025chemicalqaevaluatingllms,
      title={Beyond Chemical QA: Evaluating LLM's Chemical Reasoning with Modular Chemical Operations}, 
      author={Hao Li and He Cao and Bin Feng and Yanjun Shao and Xiangru Tang and Zhiyuan Yan and Li Yuan and Yonghong Tian and Yu Li},
      year={2025},
      eprint={2505.21318},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2505.21318}, 
}