Capybara‑Preferences Dataset

Dataset Information

Features

• source: Data source, string type.
• chosen: Selected samples, containing:
  • content: Content, string.
  • role: Role, string.
• chosen_rating: Rating of selected samples, integer.
• chosen_model: Model used for selected samples, string.
• rejected: Rejected samples, containing:
  • content: Content, string.
  • role: Role, string.
• rejected_rating: Rating of rejected samples, integer.
• rejected_model: Model used for rejected samples, string.

Data Split

• train: Training set with 15 204 samples, size 150 554 506.82 bytes.
• test: Test set with 200 samples, size 1 980 459.18 bytes.

Dataset Size

• download_size: 79 362 069 bytes.
• dataset_size: 152 534 966.0 bytes.

Configuration

• config_name: default
  • data_files:
    • train: Path data/train-*.
    • test: Path data/test-*.