---
license: cc-by-4.0
tags:
- chemistry
- molecular-qa
configs:
- config_name: property
  data_files:
  - split: train
    path: property/train.parquet
  - split: validation
    path: property/validation.parquet
  - split: test
    path: property/test.parquet
- config_name: source
  data_files:
  - split: train
    path: source/train.parquet
  - split: validation
    path: source/validation.parquet
  - split: test
    path: source/test.parquet
- config_name: structure
  data_files:
  - split: train
    path: structure/train.parquet
  - split: validation
    path: structure/validation.parquet
  - split: test
    path: structure/test.parquet
- config_name: usage
  data_files:
  - split: train
    path: usage/train.parquet
  - split: validation
    path: usage/validation.parquet
  - split: test
    path: usage/test.parquet
- config_name: all
  data_files:
  - split: train
    path:
    - property/train.parquet
    - source/train.parquet
    - structure/train.parquet
    - usage/train.parquet
  - split: validation
    path:
    - property/validation.parquet
    - source/validation.parquet
    - structure/validation.parquet
    - usage/validation.parquet
  - split: test
    path:
    - property/test.parquet
    - source/test.parquet
    - structure/test.parquet
    - usage/test.parquet
---

# MoleculeQA, hub-native

`IDEA-AI4S/MoleculeQA` (`cd1bcb922c9127b57e0944cdb605baccaa98ef9d`), restructured into one subset per category with the unified columns `ground_truth`, `response` and `scoring_context` added beside every original one. See `scripts/datasets/README.md` in the AutoDataSci repository for what those columns mean.

Row counts: {"train": 49974, "validation": 5795, "test": 5786}.

```json
{
  "property": {
    "train": 4826,
    "validation": 698,
    "test": 731
  },
  "source": {
    "train": 11061,
    "validation": 1225,
    "test": 1343
  },
  "structure": {
    "train": 32171,
    "validation": 3314,
    "test": 3113
  },
  "usage": {
    "train": 1916,
    "validation": 558,
    "test": 599
  }
}
```