---
dataset_info:
  features:
  - name: country
    dtype: string
  - name: capital
    dtype: string
  splits:
  - name: train
    num_bytes: 4776
    num_examples: 197
  download_size: 5587
  dataset_size: 4776
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---
A benchmark dataset for slms. Contains every country and capital pair.
The capitals.yaml file contains the config file for use in lm-evaluation-harness.

![image](https://cdn-uploads.huggingface.co/production/uploads/6325c1d65cf955bfbbde74b6/JD4fVr03Bd53eUhFSw9B4.png)