How does GPT-2 Predict Acronyms? Extracting and Understanding a Circuit via Mechanistic Interpretability

  1. García-Carrasco, J.
  2. Maté, A.
  3. Trujillo, J.
Actes de conférence:
Proceedings of Machine Learning Research

ISSN: 2640-3498

Année de publication: 2024

Proceedings of the 27th International Conference on Artificial Intelligence and Statistics, AISTATS 2024

Volumen: 238

Pages: 3322-3330

Type: Communication dans un congrès