ValidMol is an environment for evaluating an agent's ability to complete and fix corrupted molecular SMILES strings. Given a partial or corrupted SMILES notation, the agent must produce a valid, chemically stable molecule. Tasks apply four corruption strategies to real PubChem compounds: truncation, character deletion, ring opening, and bond corruption.
- Completing and repairing corrupted SMILES molecular representations
- Understanding chemical structure and bonding rules
- Reasoning about molecular stability and validity
- Working with cheminformatics concepts (ring closure, bond types)
Agents are given a standard environment with no sandbox or file system access.
There are two splits in this environment:
- train: 789 molecular completion tasks
- test: 142 molecular completion tasks
Tasks are generated from PubChem compounds (CIDs 2500-3500 for train, CIDs 4500-4700 for test) with four corruption types:
| Corruption Type | Frequency | Description |
|---|---|---|
| Truncation | 40% | Removes 20-50% of the SMILES suffix |
| Character deletion | 25% | Removes 1-3 random characters |
| Ring opening | 20% | Removes a closing ring number |
| Bond corruption | 15% | Removes or changes bond symbols |
This is a sparse, verifiable reward environment. The agent calls submit_completion with a SMILES string and receives a binary reward based on three-tier RDKit validation:
- Syntactic validity: RDKit parsing succeeds
- Structural validity: RDKit sanitization passes
- Chemical stability: No peroxides, hydrazines, or conjugated alkynes
- 1.0: All three tiers pass
- 0.0: Any tier fails
No LLM graders are used.
Task data is generated from PubChem compounds via the PUG-REST API. Quality filters include SMILES length 10-80, heavy atoms 5-50, and a maximum of 5 rings. Data is stored on the OpenReward platform.
| Tool | Description |
|---|---|
submit_completion |
Submit a completed/fixed SMILES string for three-tier chemical validation |
Single-turn. Agents receive a corrupted SMILES string and submit one completion.
[Put environment difficulty statistics here]
There are no further environment requirements; ValidMol works out of the box with the OpenReward endpoint.
Agents in ValidMol complete molecular structures from corrupted inputs. There is a dual-use concern in that improved molecular generation capabilities could be applied to both beneficial and harmful purposes.
@dataset{GRValidMol,
author = {General Reasoning Inc. Team},
title = {ValidMol},
year = {2026},
publisher = {OpenReward},
url = {https://www.openreward.ai/GeneralReasoning/ValidMol}
}