MolDeTox: Toxicity-Aware Molecular Editing Benchmark

MolDeTox is a benchmark dataset designed to evaluate toxicity-aware molecular editing capabilities of LLMs and VLMs. It is constructed based on the concept of toxicity cliffs, where structurally similar molecules exhibit opposite toxicity labels. The dataset was created by the MolDeTox organization and was last updated on May 5, 2026.

Use Cases

• Benchmarking toxicity-aware molecular editing models based on the toxicity cliff concept.
• Training models to learn minimal structural modifications for detoxification based on the dataset's design.
• Evaluating model performance in preserving physicochemical properties while altering toxicity labels.

Strengths

• Designed specifically for evaluating toxicity-aware molecular editing in LLMs and VLMs.
• Constructed using the concept of toxicity cliffs to enable learning of minimal structural changes.
• Last updated on May 5, 2026, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data source details are truncated; the full description must be accessed externally.

Provenance

Source

Derived from publicly available sources, as indicated by the truncated description.

Collection Method

Constructed based on the concept of toxicity cliffs.

Freshness

Last updated 2026-05-05 05:45:27