Dataset · HydroSophyTech
Molecular structure detoxification benchmark for MLLMs
HydroSophyTech/ToxiMol-benchmark
First benchmark for evaluating multimodal LLMs on molecule toxicity reduction.
What's inside
First comprehensive benchmark for molecular structure modification aimed at eliminating toxicity, focused on multimodal large language models (MLLMs). Accepted as an oral presentation at KDD 2026 (AI for Sciences track). The dataset contains fewer than 1000 examples in English and is formatted for tabular classification and regression tasks.
What you can do with it
- Evaluate the ability of multimodal LLMs to propose safe modifications for toxic molecules
- Test models for computational chemistry and drug design on molecular property classification and regression tasks
- Tasks
- Tabular classification, Tabular regression
- Languages
- English
- Size
- up to 1K rows
- Downloads this month
- 22.6K
- Likes
- 2
How to load it
from datasets import load_dataset
ds = load_dataset("HydroSophyTech/ToxiMol-benchmark")