instructions_anormales

  • Descriptif :

Ensemble de données décrit dans l'article : Unnatural Instructions : Tuning Language Models with (Almost) No Human Labor (2022). Contient des ensembles d'instructions en langage naturel, avec des contraintes facultatives / des reformulations générées par LLM.

Diviser Exemples
'train' 66 010
  • Structure des fonctionnalités :
FeaturesDict({
    'id': Text(shape=(), dtype=string),
    'instances': Sequence({
        'constraints': Text(shape=(), dtype=string),
        'input': Text(shape=(), dtype=string),
        'instruction_with_input': Text(shape=(), dtype=string),
        'output': Text(shape=(), dtype=string),
    }),
    'instruction': Text(shape=(), dtype=string),
    'reformulations': Sequence({
        'input': Text(shape=(), dtype=string),
        'instruction': Text(shape=(), dtype=string),
        'instruction_with_input': Text(shape=(), dtype=string),
        'output': Text(shape=(), dtype=string),
    }),
})
  • Documentation des fonctionnalités :
Caractéristique Classer Forme Dtype La description
FonctionnalitésDict
identifiant Texte chaîne Identifiant unique par exemple.
les instances Séquence
instances/contraintes Texte chaîne Contraintes spécifiques à la tâche.
instances/entrée Texte chaîne Entrée à alimenter dans des espaces réservés pour une instruction donnée.
instances/instruction_with_input Texte chaîne Instructions avec entrées fournies aux espaces réservés.
instances/sortie Texte chaîne Sortie cible pour une tâche donnée.
instruction Texte chaîne Instruction avec espace réservé pour les entrées.
reformulations Séquence
reformulations/entrée Texte chaîne Entrée à alimenter dans des espaces réservés pour une instruction donnée.
reformulations/instruction Texte chaîne Instruction avec espace réservé pour les entrées.
reformulations/instruction_with_input Texte chaîne Instructions avec entrées fournies aux espaces réservés.
reformulations/résultats Texte chaîne Sortie cible pour une tâche donnée.
  • Citation :
@misc{honovich2022unnatural,
      title = {Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor},
      author = {Honovich, Or and Scialom, Thomas and Levy, Omer and Schick, Timo},
      url = {https://arxiv.org/abs/2212.09689},
      publisher = {arXiv},
      year={2022}
}